WO1992000004A1

WO1992000004A1 - Methods and apparatus relating to micropropagation

Info

Publication number: WO1992000004A1
Application number: PCT/GB1991/001043
Authority: WO
Inventors: Paul Frank Davis; Robin Deirdre Tillett; John Arthur Marchant
Original assignee: British Technology Group Ltd
Priority date: 1990-06-28
Filing date: 1991-06-27
Publication date: 1992-01-09
Also published as: EP0536287A1; GB9113891D0; GB2245810A; AU8198891A; GB9014387D0

Abstract

A method of cutting a plant comprising shoots and callus comprises the steps of generating an image signal by a video camera representing an image of the plant, and processing the image signal to locate voids between shoots of the plant and to define a cutting line extending from a selected void through the callus of the plant. The void is selected by evaluating for each void a weighted combination which varies linearly and positively with a measure of the area of the void or parts thereof; negatively with a measure of the horizontal distance of the void from a preferred location; and quadratically with a measure of the vertical distance of the bottom of the void from a preferred location and varies with respect to the size and placement of other voids. A pivot point is selected at the centroid of the void or of a lower part thereof. The cutting line passes through the pivot point and its slope is set by reference to a weighted combination of the effect of the shape and size of the void and the shape and size of the callus. The weighted combination also varies quadratically with respect to the slope of a tentative cutting line through the void and the slope of the callus.

Description

METHODS AND APPARATUS

RELATING TO MICROPROPAGATION

The present invention relates to methods and apparatus for use in micropropagation. The invention is concerned in particular with a method of selecting a cutting line or plane for cutting a plant comprising shoots growing from a base region of the plant, and apparatus for putting the method into effect.

Micropropagation of plants involves the use of the techniques of plant tissue culture and the application of these to the propagation of plants. At its simplest, micropropagation consists initially of excising small pieces of actively growing tissue, normally shoot tips or nodes cut from the stems of plants. Then, under sterile conditions, the pieces of tissue are transferred to a nutrient medium which supports plant growth. The plant material will finally develop into entire plantlets. These plantlets must then be weaned from the axenic conditions in which they have existed within the laboratory into viable, rooted plants capable of survival in conventional

horticultural or agricultural environments. Normally the step of cutting a plantlet into small pieces, referred to as ex-plants, for regrowth (the multiplication stage) is repeated several times before a batch of plantlets is grown into viable plants.

The growth of plants from tissue culture,

micropropagation, is a technique which can produce large numbers of genetically identical plants, perhaps possessing a desirable quality such as disease resistance, in a short time. The tasks of dissecting and transplanting such plants are labour-intensive and repetitive, and the gains in speed, sterility and labour costs which could be

achieved by the use of robots make automation an attractive prospect for the fast-expanding micropropagation industry. However, automation is difficult, requiring methods of robot guidance which can deal with the natural variability of biological objects. Vision processing is a method of sensory control which has been applied to similar problems in other areas of agriculture, such as tomato sorting, fruit harvesting, and plant identification. However these techniques are difficult to apply in micropropagation because of the nature of the plant material and the

features, such as nodes of the plant, which need to be located.

In known, manual, methods of micropropagation, four of the particularly important operations which are repeated frequently are (i) removing a plant from a container, (ii) cutting a required portion of plant tissue from the donor plant, (iii) transferring the cut portion of plant

material, and (iv) placing the plant portion in a soft nutrient medium in such a manner that it stands upright. As performed at present, the cutting operation normally consists of an operator holding the plant material by forceps on a sterilised card by one hand, and cutting the required portion of the plant by strokes of a scalpel, by the other hand. Commonly, the cuts are required to cut from the donor a shoot tip, or a node from which a side shoot will develop from an anxillary bud. The cut portion is then transferred by the forceps to the soft nutrient medium, which is in the nature of a gel, and the cutting is then manoeuvred so as to stand upright with the stem part of the cutting in the soft nutrient medium. The

manoeuvring is normally carried out with the use of forceps.

In the micropropagation of some species, for example syngoniums, a callus of plant material is grown and small shoots grow from it. At intervals during its growth it is divided into several more pieces such that each retains one or more strong shoots. The larger leaves on the longest shoots are cut off to encourage the formation of more shoots. After a number of cycles of dissection and

replanting, the many pieces of callus obtained are allowed to grow into normal plants. The present invention is concerned with the automation of such processes with the aid of robotic, vision analysis guided, apparatus. It is an object of the present invention to provide a method of selecting a cutting line for cutting a plant comprising shoots growing from a base region of the plant, particularly, but not exclusively, suitable for use in the multiplication stage of micropropagation, where the main requirement is to cut the plant material into portions each of which has sufficient base region and sufficient shoots, to provide regrowth.

According to the present invention in a first aspect, there is provided a method of selecting a cutting line for cutting a plant, comprising shoots growing from a base region of the plant comprising generating an image signal representing an image of the plant, and processing the image signal to carry out the steps of locating voids in the image of the plant, selecting a void suitable to include the starting point of a cutting line for cutting the base region of the plant, and defining a cutting line extending from the selected void through the base region of the plant the said void being selected by reference to: (i) the area of the void or one or more selected parts of the void, alone or in combination with the area of one or more neighbouring voids or selected parts thereof; and/or (ii) the horizontal location of the void; and/or (iii) the vertical location of a bottom region of the void; and/or (iv) the slope of a proposed cutting line determined for the void. Where references are made to horizontal and vertical, these are to be taken as references to horizontal and vertical directions on a normal presentation of the image of the plant on a monitor screen with the base region of the plant at the lower part of the image and the shoots at the upper part. The terms vertical and horizontal when used with regard to the image signal do not necessarily relate to the orientation of the plant itself at the actual work station.

By the term void is meant an area of the image of the plant which is bounded at least partially by plant

material, and may be a totally closed void, or may be a void which is closed at the bottom but open at the top. By the term gap is meant a space between shoots in the actual three dimensional plant being examined, which space, when viewed in the two dimensional image of the plant, may be divided by plant material into more than one void in the image of the plant.

Preferably the method includes evaluating for each void at least two of the following functions, namely a function related to the area of the void or one or more selected parts of the void, alone or in combination with the area of one or more neighbouring voids or selected parts thereof; a function related to the horizonal location of the void; a function related to the vertical location of a bottom region of the void; and a function related to the slope of a proposed cutting line determined for the void; and selecting that void which has a preferred weighted combination of said at least two functions.

Preferably the void is selected by reference to the area of the void, or, where the vertical depth of the void is greater than one or more predetermined values, with reference to the area or areas of parts of the void bounded by the bottom of the void and extending up the void to the said predetermined value or values of vertical depth. Thus there is also an improvement in giving different weightings to the importance of the average width of a void over several different regions of the void. Also preferably the said void is selected by reference to selected portions of neighbouring voids diminished by factors relating to the linear and angular displacement between at least one reference point in the void and in each of the neighbouring voids. Conveniently the void is selected by reference to the sums of products over neighbouring voids while in each sum such a product is that of a variable factor with the area of a portion of a neighbouring void. In such

arrangements the relative positions of voids in the

silhouette of the plant are examined in order to discern whether separate voids are likely to be partially obscured parts of the same gap between stems. In accordance with another preferred feature, the said void is selected by reference to the horizontal location of the void relative to the centroid of the base region of the plant. In accordance with a further preferred feature, the said void is selected by reference to the horizontal location of the void relative to a location where, if the void were located there, on average there would be an even distribution of the image of plant material on either side of the void. Preferably the void is selected by reference to a measure of the difference in area of plant material shown in the image on either side of the void, or a

selected part thereof. Conveniently the void is selected by reference to the difference between the total area of plant material on one side of the void and that on the other side of the void, divided by the vertical depth of the void or, where the void has a depth greater than a predetermined vertical depth, the difference up to that depth divided by the said predetermined vertical depth. In accordance with another preferred feature, the said void is selected by reference to the height of the bottom region of the void relative either to the bottom of the base region of the plant, or to a reference frame of the image. Also preferably the void is selected by reference to the relationship of the height of the bottom region of the void to the height of the centroid of the base region of the plant. Conveniently the said void is selected by reference to the square of the height of a reference point of the void from the bottom of the base region of the plant, the reference point of the void being the centroid of the void, or where the vertical depth of the void is greater than a predetermined depth, the centroid of the part of the void bounded by the bottom of the void and extending up to the said predetermined vertical depth.

Also conveniently the void is selected by reference to the product of the height of a reference point of the void with the height of the centroid of the base region of the plant measured from the bottom of the base region, the reference point of the void being the centroid of the void, or where the vertical depth of the void is greater than a

predetermined depth, the centroid of the part of the void bounded by the bottom of the void and extending upto the said predetermined vertical depth. In accordance with a yet further preferred feature the void is selected by reference to the slope of a proposed cutting line through the base region of the plant, which cutting line has its angle of inclination determined by reference to the shape and/or size of the void, and/or to the shape and/or size of the base region of the plant.

Preferably the said void is selected by reference to the square of a measure of the inclination to the vertical of a proposed cutting line for the void. Also preferably the said void is selected by reference to the relationship of the slope of a proposed cutting line extending from the void through the base region of the plant, to a measure of the slope of the base region of the plant. Finally, it is preferred that the said void is selected by reference to the product of a measure of the inclination to the vertical of a proposed cutting line for the void with that of a line, through the centroid of the image of the base region of the plant, having a least sum of squares fit to the horizontal locations of plant pixels in the image of the base region of the plant. In such arrangements those voids which would result in a highly inclined choice of cutting plane are regarded as less suitable for selection.

In one particular form, the method includes evaluating for each void ten items which are functions of the size, shape and location of the void, namely:- two parts of the area of the void; two sums such that the first and second sum are of the first and second parts, respectively, of the areas of neighbouring voids, but with each part of a neighbouring void weighted according to its relative position with respect to the original void; a measure of the imbalance between the amount of plant material on one side of the void and that on the other side of it; the horizontal displacement of a reference point in the void from the centroid of the base region of the plant; the square of the height of the reference point above the bottom of the base region of the plant; the product of that height with the height of the centroid of the base region of the plant; the square of the slope of a proposed cutting line extending to the void; and, lastly, the product of that slope with a measure of the slope of the base region of the plant. A sum of the ten specified functions weighted with empirically chosen parameters forms a discriminant function by which a void, with a cutting line extending to it, is selected by choosing the void with the largest discriminant function. In particular the discriminant function is partly comprised of the following three sums, namely:- a weighted sum of functions of the horizontal position of the void which has its greatest value at what may be regarded as a preferred horizontal position; a weighted sum of the functions of the height of the void which is a quadratic function with a maximum value at what may be regarded as a preferred height; and a weighted sum of the functions of the slope of the proposed cutting line which is a quadratic function with a maximum at what may be regarded as a preferred slope.

In a preferred aspect of the invention, there may be provided a method of selecting a cutting plane for cutting a plant comprising shoots growing from a base region of the plant, the method comprising rotating the plant through a series of angular positions; viewing the plant at each said position and generating an image signal representing an image of the plant at that angular position; processing the image signal from each angular position by a method in accordance with the invention as set out hereinbefore to select for each different image a void suitable to include the starting point of a cutting line for cutting the base region of the plant, and further processing the image signals to carry out the steps of comparing the selected voids in the different images to select from the different images the most suitable void to include the starting point of a cutting line for cutting the base region of the plant, and defining a cutting plane for cutting the plant as being a plane containing the cutting line of the selected void and containing the direction of viewing of the plant which gave rise to the image containing the selected void. Conveniently, the selection of a void from voids occurring in different images taken at different angular positions, is made, alone or in combination with other selection procedures, in accordance with any one or more of the selection procedures set out hereinbefore. Preferably the parameters used when comparing voids from different images of the plant taken at different angular positions, are different from the parameters used when comparing voids from the same image taken at a single angular position.

Most preferably the selection of a void from voids occurring in different images taken at different angular positions, is selected, alone or in combination with other selection procedures by reference to:-

(i) the width of the base region of the plant in the image containing the void being considered;

and/or

(ii) the height of the centroid of the base region of the plant in the image containing the void being considered.

In a particularly preferred arrangement, the selection of a void from voids occurring in different images taken at different angular positions, is selected, alone or in combination with other selection procedures by reference to:-

(i) a function related to the ratio of the width of the base region of the plant to the height of its centroid; and/or

(ii) a function related to the height of the centroid of the base region of the plant; and/or (iii) a function related to the ratio of the square of a difference function to the width of the base region of the plant, where the difference function is the difference between the total area of plant material on one side of the void and that on the other side of the void divided by the vertical depth of the void or, where the void has a depth greater than a predetermined vertical depth, divided by the said

predetermined vertical depth.

In accordance with another aspect of the invention, there may be provided in a broad aspect a method of

selecting a cutting plane for cutting a plant comprising shoots growing from a base region of the plant, comprising: rotating the plant through a series of angular positions; viewing the plant at each said position and generating an image signal representing an image of the plant at that angular position; and processing the image signals to carry out the steps of locating voids in the images of the plant, selecting a void suitable to include the starting point of a cutting line for cutting the base region of the plant, defining a selected cutting line extending in the image of the selected void from the selected void through the base region of the plant, and defining a cutting plane as being a plane containing the selected cutting line and containing the direction of viewing of the plant which gave rise to the image containing the said selected void.

In accordance with the invention in another aspect, there is provided a method of selecting a cutting line for cutting a plant comprising shoots growing from a base region of the plant, comprising generating an image signal representing an image of the plant, and processing the image signal to carry out the steps of locating a selected void in the image corresponding to a gap between shoots of the plant, defining a cutting line extending from the selected void through the base region of the plant, and determining the angle of inclination of the cutting line relative to a predetermined direction; the angle of

inclination of the cutting line being determined by

reference to: the shape and/or size of the void, and/or the shape and/or size of the base region of the plant.

Preferably the method includes the steps of selecting a pivot point located in the void, setting the cutting line to pass through the pivot point, and setting the slope of the cutting line relative to the vertical by reference to the size and/or shape of the void and/or the base region of the plant. Preferably the pivot point is located at the centroid of the void, or where the vertical depth of the void is greater than a predetermined value, at the centroid of part of the void bounded by the bottom of the void and extending up the void to the said predetermined value of depth. The predetermined depth is such that the pivot point is a reference point of the bottom region of the void. Conveniently, the slope of the cutting line is set as the weighted sum of the slopes of two lines both passing through the pivot point, the slope of the first line being set by reference to the size and/or shape of the void, and the slope of the second line being set by reference to the size and/or shape of the base region of the plant.

Preferably the said weighted sum of the slopes gives more weight to the second line than to the first line.

In one convenient arrangement, the method includes formulating for the void a line which minimises a measure of the horizontal distances on either side of the line, of the component pixels of the void or a selected part

thereof, and setting the slope of the cutting line with reference to the said line. Preferably the said minimising step comprises minimising the sum of the squares of the horizontal distances on either side of the line, of the component pixels of the void or the selected part thereof. Also conveniently, the method includes formulating for the base region of the plant a line which minimises a measure of the horizontal distances on either side of the line, of the component pixels of the base region of the plant or a selected part thereof, and setting the slope of the cutting line with reference to the said line. Preferably the said minimising step comprises minimising the sum of the squares of the horizontal distances on either side of the line, of the component pixels of the base region of the plant or the selected part thereof.

The invention also provides a method of cutting a plant comprising shoots growing from a base region of the plant, comprising the steps of selecting a cutting line in accordance with the method of any of the preceding

paragraphs, generating a control signal related to the cutting line, and cutting the plant along the cutting line by robotic cutting means under the control of the said control signal.

It is particularly to be noted that where a feature of the invention has been set out with regard to a method, there is also provided in accordance with the invention an apparatus incorporating that feature, and vice versa. In particular, there is provided in accordance with the invention apparatus for use in micropropagation, comprising means for generating an image signal representing an image of a plant comprising shoots growing from a base region of the plant; and signal processing means for processing the image signal to locate voids in the image of the plant, to select a void suitable to include the starting point of a cutting line for cutting the base region of the plant, and to generate an output signal containing information as to the location of a cutting line extending from the selected void through the base region of the plant; the signal processing means being arranged to select the said void by reference to: (i) the area of the void or one or more selected parts of the void, alone or in combination with the area of one or more neighbouring voids or selected parts thereof, and/or (ii) the horizontal location of the void; and/or (iii) the vertical location of a bottom region of the void; and/or (iv) the slope of a proposed cutting line determined for the void. In a preferred aspect of the invention the apparatus may further comprise means for rotating a plant comprising shoots growing from a base region of the plant, through a series of angular positions; and means for viewing the plant at each said position and for generating an image signal representing an image of the plant at that angular position; the said signal processing means being arranged to process the image signal from each angular position to select for each different image a void suitable to include the starting point of a cutting line for cutting the base region of the plant, and to compare the selected voids in the different images to select from the different images the most suitable void, said output signal containing information defining a cutting plane as being a plane containing the selected cutting line and containing the direction of viewing of the plant which gave rise to the image containing the said selected void.

In accordance with one main, broad, aspect of the invention, there may be provided apparatus for use in micropropagation, comprising: means for rotating a plant comprising shoots growing from a base region of the plant, through a series of angular positions; means for viewing the plant at each said position and for generating an image signal representing an image of the plant at that angular position; and signal processing means for processing the image signals to locate voids in the images of the plant, to select a void suitable to include the starting point of a cutting line for cutting the base region of the plant, and to generate an output signal containing information as to the location of a cutting line extending in the image containing the selected void from the selected void through the base region of the plant, said output signal containing information defining a cutting plane as being a plane containing the selected cutting line and containing the direction of viewing of the plant which gave rise to the image containing the said selected void.

In accordance with another aspect of the invention, there is also provided apparatus for use in

micropropagation, comprising means for generating an image signal representing an image of a plant comprising shoots growing from a base region of the plant; and signal

processing means for processing the image signal to locate a selected void in the image corresponding to a gap between shoots of the plant, to define a cutting line extending from the selected void through the base region, and to generate an output signal containing information as to the location of the said cutting line; the signal processing means being arranged to determine the angle of inclination of the cutting line by reference to:- the shape and/or size of the void; and/or the shape and/or size of the base region of the plant.

The apparatus may further comprise, in either aspect, means for presenting the plant at a workstation, and robotic cutting means for cutting the plant along the cutting line under the control of the said output signal.

Embodiments of the invention will now be described by way of example with reference to the accompanying drawings in which:- Figure 1 is a diagrammatic representation of apparatus embodying the invention for cutting plants during

micropropagation; Figure 2 is a flow chart of a routine for processing a visual image signal in an embodiment of the invention;

Figures 3a and 3b show visual images of two syngonium plantlets with suitably selected cutting lines for

micropropagation purposes;

Figures 4a and 4b show visual images of syngoniums with poorly selected cutting lines during micropropagation; and

Figure 5 is a diagrammatic representation of a system of boundary labelling of a silhouette of part of a plant, which may be used in carrying out an embodiment of the invention.

The micropropagation of plants is a labour intensive process that can benefit from automation. Seedling size plants are grown from selected portions of plants in a special propagating medium, and selected portions of these are then grown into more small plants. For some types of plant special care is required in selecting the portions for reasons which vary with the type of plant, and it is more difficult to automate their micropropagation. For example, a certain amount of care is needed in dividing syngoniums during micropropagation in order to achieve a good rate of multiplication of the plants. The pieces of syngonium each form a callus with shoots on it but not much root when grown in the special propagating medium. Rapid growth and multiplication is maintained by dividing each callus after only a few weeks of growth into several more pieces which should all be about the same size, and such that each retains one or more strong shoots. The pieces of callus are placed in fresh propagating medium. Normally a callus is cut into two or three pieces, and the larger leaves on long stalks are removed to encourage the growth of fresh shoots. Often there are two or more distinct clusters of shoots on a callus but sometimes the shoots are somewhat entwined around each other. The cutting planes should be chosen to pass between the shoots and, if

possible, not through any of them. This process of growing pieces of callus and cutting them is repeated several times before the pieces are allowed to grow into normal plants.

To automate syngonium micropropagation several

consecutive processes have to be automated. The seedling size plants are grown in sealed containers each holding several plants, and each plant has to be extracted from its container. The plant then has to be orientated suitably for dissection, and cut into selected pieces. Lastly, the pieces have to be transferred to freshly prepared

propagating medium in other containers.

The embodiment of the invention will be described in the context of automatic operations on plant tissue culture during micropropagation, for example the cutting of

syngonium calluses. In Figure 1 a conveyor belt 11 carries plant material beneath a cartesian robotic system indicated generally at 13. The plant material can be viewed from the side by a solid state video camera 12. A carriage 14 can be moved to any selected location over the conveyor belt 11 by means of motors 15 and 16 controlled by a control device 17. The speed of the conveyor 11 is controlled by a control device 18 so that the position of plant material observed by the camera 12 can be correlated with the subsequent position of the material beneath the carriage 14. The carriage 14 carries a tool 19 for manipulating, rotating and cutting plant material, which is operated under control of a control device 20. The control devices 17, 18 and 20 are all controlled by a main microcomputer 21, which constitutes a signal processing means programmed to carry out the method embodying the invention.

The video camera 12 feeds a visual image signal to the signal processing means 21 which embodies the invention and which produces output signals containing information as to a selected cutting line and its location relative to the plant material observed by the camera 12. This information is processed to produce control signals fed to the control devices 17, 18 and 20 which then operate the components already described, to move the cutting parts of tool 19 relative to the plant when it has been brought to the work station beneath the carriage 14, by the conveyor 11.

Figure 2 illustrates in diagrammatic form a flow chart setting out the main steps of the processing of the image signal in the signal processing means 22. This chart will be described in detail below, after the general manner of operation of the embodiment has been described.

The image processing algorithm described herein works on a side view of the plant. It seems best to view each plant from the side so that the shape of the callus and the bottom of each cluster of shoots is most clearly visible. In a modification, not shown, the plant may be fixed in a rotatable holder, which can be rotated about an axis perpendicular to the viewing axis of the camera 12, that is to say about a vertical axis. The angle of rotation determines one of the parameters of the cutting plane, and the remaining parameters are determined by a calculated cutting line on a two dimensional image, as will be

described in accordance with the invention. The most appropriate angle of rotation for each plant may be

selected manually, for example being observed on a video monitor 23 connected to the camera. Alternatively means may be provided for automatically selecting the angle of rotation, for example by comparing a series of images which are recorded as each plant is automatically presented at a series of small equal angle increments. The image of a silhouette of the plant is generally adequate for selecting a suitable angle of rotation and automatically analyzing the image. The object during the rotation of the plants is to obtain images which most clearly distinguish the

clusters of shoots. At the most suitable angle of

rotation, it is then assumed that the most suitable cutting plane is perpendicular to the viewing plane. This is not always the case, but is a good approximation. The task of the automatic image processing method of the present invention is to select a cutting line on the image, and this line then corresponds to the intersection of the cutting plane with the viewing plane. Figure 3a shows an image of a plant with two very distinct clusters of shoots and the selected cutting line between them.

Figure 3b shows a more complicated image in which the clustering of the shoots is less obvious, but nevertheless the selected cutting line separates two distinct clusters of shoots. The inclination of the cutting line in Figure 3b ensures that a reasonable size piece of callus remains attached to the smaller cluster.

Two examples of appropriate cutting lines have been shown in Figures 3a and 3b. Two examples of cutting lines considered inappropriate are shown in Figures 4a and 4b.

Figure 4a shows a plant with the cutting line which starts from an inappropriate void and cuts through a cluster of shoots. Figure 4b shows a plant with a vertical cutting line which, if it had been suitably inclined instead, could have divided the main part of the callus into nearly equal portions. There will now be described the basic principles of the algorithm embodying the invention which is used for defining the cutting line. Next there will then be described the main steps of the flow chart shown in Figure 2. Finally there will be described a number of techniques which may conveniently be used in processing the visual image signal, to put into effect the algorithm.

There will first be described the factors which are taken into account in the algorithm described in the present embodiment, which takes the form of a discriminant function.

The top of a cutting line should be located within some part of the top of the callus, but the top of the callus is not always easy to discern when there are many shoots growing from it. However, by referring to the height of the centroid of the callus many unsuitable choices for its location can be avoided. Among other items the function discriminates on a quadratic function of the difference between the height of any plausible location and the height of the centroid of the callus. To be plausible the top of the cutting line should also be in the bottom of a gap (also referred to as a void) in the silhouette, and this gap should appear to be between two clusters of shoots. The bottom of such a gap is signalled in the image signal by a new black-to-white boundary section followed by a new white-to-black section. Not all such pairs of boundary section begin at the base of clusters of shoots; some are associated with roots, some with leaves and some begin higher up clusters of shoots where shoots divide or cross each other. New boundary sections in the highest or lowest parts of an image are unlikely to belong to a gap that should be chosen. The gaps between clusters of shoots are generally, but not always, wider and taller than the gaps between the shoots in a cluster, and so the gaps with a large area are often more important than those with a small area. Above a certain height the shapes of the gaps tend to depend on the shapes of the leaves and stems, and so another input parameter is used to limit the depth considered within each gap. In a 64×64 pixel image frame 16 pixels is found to be a suitable depth value. Parts of the area of a gap, thus limited, are other items in the discriminant function used for selecting the top of the cutting line.

A cutting line that divides the callus into

approximately equal portions is preferable to one which divides it into a large and small portion. A cutting line starting in a more central gap generally has a more

balanced mass of shoots on each side of it. Another item included in the discriminant function is the difference between the sum of the widths of all dark areas of image on each side of the gap averaged over the previously limited depth of the gap. An amount proportional to the absolute difference is subtracted from the function in order to discriminate against the placement of a cutting line at the edge of the callus. The averaging mentioned above, amounts to dividing the total difference in image area of plant material on either side of the gap by the depth of the gap. Incidentally, this results in a measure of a horizontal distance of the gap from a preferred location for it with respect to the distribution of plant material above the callus. Thus information about both a preferred vertical and horizontal location is used. The division of the difference in area by the depth of the gap, enhances the relative importance of deep narrow gaps often found between large clusters of shoots such as those in Figure 3b.

Although in Figure 3b the gap selected to contain the top of the cutting line is narrow and slightly to the left of centre of the image it happens to be deep compared to a wider but less suitable gap nearby to the right of it.

A preferred formula for the discriminant function used is as follows:-

where n is the number of terms in f, and in the preferred arrangement n = 10; p_i is one of a set of n empirically chosen parameters; g_i is one of a set of n functions of the size and location of the void;

The functions g_i will shortly be defined for the preferred arrangement, but firstly here is a summary of their properties as identified in that arrangement. g₁ and g₂ only affect the preferred horizontal

location of the void; g₃ and g₄ only affect the preferred vertical location of the void; g₅ and g₆ affect the preferred slope for the proposed cutting line; g₇ and g₈ are areas of lower parts of the void which affect the choice of void; g₉ and g₁₀ are indicators of whether some neighbouring voids are parts of the same gap between the stems of the syngoniums. In the preferred arrangement the functions g_i are defined as follows, but not necessarily in the order given: g₁ is the absolute value of the horizontal

displacement of the pivot point of the void from the estimated position of the centroid of the image of the callus. The more favourable voids are often nearly

vertically aligned with the centroid. The sign of g could be changed if positive p values are desired. g₂ is the absolute value of an average difference, per horizontal scan, between the sum, over every scan averaged, of the pixels in all parts of the plant image encountered on one side of the void and the sum of those encountered on the other side. Here the average is taken from the bottom of the void to the highest of the prescribed levels in the definitions of g₇ and g₈, or to the top of the void, whichever is lower. Incidentally, the resulting average difference may be regarded as the horizontal distance, in pixel widths, to a site for the void which would be

preferred with respect to the consequent distribution of the image of plant material on either side of the void. g₃ is the square of the height of the pivot point of the void from the estimated position of the bottom of the callus. Here again the sign of g could be changed if positive p values are desired. g₄ is the product of the height described immediately above, with the height of the centroid of the callus measured from the bottom of the callus. The maximum value of the sum of the products of this and the preceding g value each with its respective empirical parameter occurs at a height to be preferred for a void selected according to that height. g₅ is the square of the slope, i.e. tangent, of the inclination to vertical of a tentative cutting line through the pivot point of the void. If positive p values are desired the sign of this g value can be changed in order for it to decrease for less favourable voids. g₆ is the product of the slope described above with that of a line, through the centroid of the image of the callus, having a least sum of squares fit to the horizontal locations of dark pixels in the image of the callus. The maximum value of the sum of the products of this and the preceding g value each with its respective empirical parameter, occurs at a cutting line slope to be preferred for a void selected according to that slope. g₇ and g₈ or, more generally, g₇ to g_m+6 , where m=2 is preferred, are areas of portions of the void including both that between the bottom of the void and the first of m prescribed levels and also those between the remaining levels. On a 64×64 pixel image frame suitable levels are at 8 and 16 pixels above the bottom of the void. If the void cannot accommodate all the levels the portion between the highest level to be accommodated and the top of the void becomes the highest portion to be accommodated, and higher portions then have zero area. g₉ and g₁₀ or, more generally, g_m+7 to g_m+km+6, where k=1 is preferred, are each comprised of sums of products over neighbouring voids while, in each sum, such a product is that of a variable factor with the area of a portion of a neighbouring void. The portions correspond to those defined with g₇ and g₈, but belong to the other voids. There is one variable factor in the preferred arrangement, but more generally there could be several as here denoted by k. Thus with m portions and k factors mk sums of products can be obtained. The factors are functions of the positions of the neighbouring voids relative to that of the original void, and such that f may be augmented when a neighbouring void is a slightly higher separated part of the image of the same gap between stems as the original void. With k=1 the factor in the preferred arrangement is set to zero if the height of such a portion from the bottom of the original void is not approximately within the range of that void's prescribed levels mentioned above,

regardless of the height of the original void. If

non-zero, that factor is r as follows: r = y³/ (x² + y² )². Here x and y are respectively the horizontal and vertical displacements of the neighbouring void from the original void as measured between the pivot points selected for each void. The preferred discriminant function will now be described further in terms of a formula for a void number j. The formula for the discriminant function for the void includes functions of the size and proximity of

neighbouring voids, and therefore the void number index j is introduced in order to distinguish between it and any neighbouring void numbered with a different index k.

In preferred arrangements the discriminant function f_j of void j can also be described as follows:

where Z_j = Y_j - Y_B Z_C = Y_C - Y _B q_1jk and q_2jk are functions of X_jk, Y_jk and h₁ and h₂

which are defined below;

if y_jk < h₂ then q_{lj k} = r_jk, otherwise q_ljk = O if y_jk < h₁ then q_2jk = r_jk, otherwise q_2jk = O

Xjk ^{= X}k ^{- X}j Y_jk = Y_k - Y_j

V_lj, V_2j are the areas of the lower and upper selected portions of the void, respectively; h₁, h₂, are the lower and upper levels respectively, which specify V₁ and V₂ and in a 64×64 pixel frame suitable levels are 8 and 16 pixels above the bottom (not bottom region) of the void; X_j, Y_j are the coordinates of the pivot point (with respect to the image frame);

X_C, Y_C are the coordinates of the callus centroid; Y_B is the height of the base of the callus;

S_j is the slope of the proposed cutting line for void j;

S_C is the slope of the callus; w_j = ( a_2j - a_lj) / D_j D_j is either the depth of the void, or h₂ if the depth exceeds h₂; a_lj is the total area of plant image over that depth on the left of the void; a_2j is the total area of plant image over that 2_j depth on the right of the void.

Before the use of the discriminant function suitable values have to be assigned to the empirical parameters.

For this purpose, in one example, an entirely separate set of 62 plant images were examined and the values of the mathematical functions in the terms of the discriminant functions of selected voids were recorded. A record was also kept of a set of comparisons between pairs of these voids as to which of each pair was humanly judged to be a more suitable location for the top of a cutting line.

These records were then used as input data for a computer program which thereby selected suitable values for the empirical parameters. The method used by the computer program was to minimise what is known as the Perceptron Criterion for the data patterns compared.

In the preferred discriminant function, there are no explicitly assigned preferred horizontal or vertical locations for a void. Nevertheless, for any particular image, the sum of certain terms in the discriminant

function reaches a maximum value at a particular vertical location, while that of other terms reaches a maximum value at a particular horizontal location. These locations, although related to the values of the empirical parameters, are not explicitly listed in any such input data, and they vary from one image to another image. However, the points where these maxima occur, may be regarded as implicitly preferred horizontal and vertical locations which,

incidentally, depend on the size and position of the callus. In the same sense there is an implicitly preferred value for the slope of the tentative cutting line

associated with the void, and this preferred value depends on the slope of the callus.

In the preferred arrangement, the image processing can also automatically classify a callus as too small to cut so that no void is then selected. An alternative formula for the discriminant function which may be used is as follows:

where v is the area of the void considered, subject to the depth limit imposed, w is the averaged difference in total width of dark parts on either side, y is the height of the bottom of the void, h is the most likely height at which to expect the top of the cutting line, and p and q are two positive parameters chosen by

experiment. By way of example, after trials on 24 training images, the value 8 was chosen for both p and q when using a 64x64 pixel image frame. Other values may be appropriate in other circumstances.

The void with the largest value of f is chosen to contain the top of the cutting line. There will now be described the factors taken into account in calculating the inclination of the cutting line.

The cutting line is constrained to pass through a pivot point near the bottom of the chosen void. If the pivot point is too far above the bottom of the void, an inclined cutting line may cut through the base of some of the shoots. However, if the pivot point is placed in the lowest pixel of the void, it is rarely sited close to the centre of the gap between the shoots. A compromise is obtained by siting the pivot at the centroid of a region at the bottom of the void, for example three pixels deep (if the void is that deep) in a 64x64 pixel resolution image. In Figure 3a that region is approximately the portion of the central void below the emergence of the right hand shoot bordering the void. In experience, it is found that in some cases the region may be not deep enough, and in other cases it may be slightly too deep. The effect of the compromise may be observed in Figure 3a. The cutting line in Figure 3a passes slightly to the right of the lowest point of the void and thereby cuts the base of the void more centrally than if it had started at the lowest point.

The slope of the cutting line through the pivot point is the weighted sum of the slopes of two lines which are both constrained to pass through the pivot. Here the slopes are unconventionally defined as the tangents of the inclinations of the lines with respect to a vertical direction instead of a horizontal direction. The direction of cutting line is usually not far from vertical, and for that reason the calculation of the slope of each of the two lines through the pivot is simply calculated as that slope which gives the least sum of squares of certain horizontal deviations from such a line. The first such line is obtained by a least squares fit to the horizontal locations of all the white pixels in the lower 16 pixel depth limited part of the void, and the second line is a similar fit to all the dark pixels below the void. This limited depth corresponds to nearly the length of the cutting line in Figure 3a or somewhat more than 60% of the depth of the central void. A cutting line of suitable slope for most plants is obtained by adding 20% of that of the first line to 80% of that of the second line. In Figure 3b it may be seen how important it is for the dark pixel locations in the callus to predominate in determining the slope of the cutting line in order for it to suitably divide the callus. It is clear that if the shape of the void had predominated then the cutting line would have been much more nearly vertical. If the selected void in an image is only one pixel deep at the 64x64 pixel resolution then the first line is assumed to be vertical with zero slope, and in that case the slope of the cutting line is determined entirely by the shape of the callus.

Thus, to summarise, the method embodying the invention which has been described, includes the steps of selecting a void, selecting a pivot point located in the void, setting the cutting line to pass through the pivot point, and setting the slope of the cutting line relative to the vertical by reference to the size and shape of the void and the callus, by the steps that:- (i) the pivot point is located at the centroid of the void, or where the vertical depth of the void is greater than a predetermined maximum, at the centroid of part of the void bounded by the bottom of the void and extending up to the said predetermined maximum vertical depth;

(ii) the slope of the cutting line is set at the weighted sum of the slopes of two lines both passing through the pivot point, the slope of the first line being set by reference to the size and shape of the void, and the slope of the second line being set by reference to the size and shape of the callus;

( iii) formulating for the void a line which minimises the sum of the squares of the horizontal distances on either side of the line of the component pixels of the void or a selected part thereof, and setting the slope of the cutting line with reference to the said line; and (iv) formulating for the callus a line which minimises the sum of the squares of the horizontal distances on either side of the line of the component pixels of the callus or a selected part thereof, and setting the slope of the cutting line with reference to the said line.

There will now be described the sequence of steps carried out in the flow chart of Figure 2.

The sequence begins at step A in the diagram when a video signal from a monochrome video camera image of the plant's silhouette is digitised and placed in RAM by a Frame Grabber card. A clear distinction between the silhouette and its background is obtained by converting the pixel grey levels to black below a certain threshold and white above it. Subsequent image processing is speeded up by reducing the resolution of the image but, in order to retain essential details, the reduction must not be less than a 64×64 pixel image frame.

Next, at step B in the diagram, the edges of the silhouette are located in the reduced resolution image.

The position of the edges are recorded by run-coding them, but other methods, such as chain-coding may be used in an alternative arrangement. To facilitate further image processing a cross-reference of the items in the run-code is constructed in a manner which conveniently indicates the structure of the image, as explained elsewhere in this document. If another method of edge location, such as chain-coding, is used an alternative form of cross

referencing will be required for the same purpose in the alternative arrangement.

At step C the image of the callus is located and analysed. The height of the base of the callus is first determined either from the image or, depending on the method of support, by reference to that of a support for the plant. The callus is then approximately identified as an unbroken region of significant width or average width, and depth which forms the base of the plant. If such a region is not found then, as indicated by step D, no cutting line is sought because the plant is considered to be too small to cut. A second measure of the location of the callus is then calculated. The preferred measure is the position of the centroid of the image of the region as described by its horizontal and vertical co-ordinates. The slope of the callus is identified as that of a line through the calculated location e.g. the centroid, which minimises a measure of the displacement from the line, of pixels within the image of that region. The preferred measure to be minimised is the sum of squares of the horizontal distances of the pixels from the line. Steps E, F and G involve a search of the edges of the image to locate voids in it which are either open or closed at the top, but closed at the bottom. As each void is located details of it are examined to help determine the values of terms in its corresponding discriminant function. The areas of the selected portions of the void are

evaluated in the preferred arrangement. The process of evaluating the areas of portions of void is equivalent to integrating the products of area increments with functions which are constant within the portions and zero outside them. In alternative arrangements the integrals of

products of area increments with other suitable sets of functions can be evaluated instead. The two areas

corresponding to the distribution of plant material on either side of it are also evaluated, the position of the pivot point for a tentative cutting line is found and that line is also found. Some of the terms in the discriminant function can be evaluated at this stage if their empirical parameters have been determined, but if a search is to be conducted for suitable values of the empirical parameters their g factors are recorded instead for later use with various trial values.

Steps H, I and J indicate the final steps in the evaluation of the discriminant function of each void.

After all the voids have been located it is possible to evaluate the factor g in the discriminant function of each void which arises from the proximity to it of other voids, then the complete discriminant function for the void can be found if the empirical parameters have been specified.

At step K the void with the largest discriminant function is chosen and, if the callus material below it is of sufficient width, or average width, and depth, the proposed cutting line already calculated for that void is chosen as the most suitable cutting line for the plant. Finally at step L the processor waits for another image on which to repeat the sequence of steps A to K.

Although the steps of the method according to the invention may be put into effect by various software, the following techniques are found convenient in processing the visual image signal which is derived from the camera 12. See for example the labelling with numbers illustrated in Figure 5.

In order to simplify and speed up the automatic image processing the images may be simplified. The use of a silhouette is the first stage of simplification, and a threshold grey level is chosen at which to divide the scene into dark and light areas. With a suitable threshold,

"noise" is excluded from the background to the silhouette, but the image processing is not adversely affected when a slight amount of noise remains. The size of the image is reduced in order to reduce the number of pixels to be examined, but the degree of reduction for the most

acceptable compromise between fast and accurate location of cutting lines has to be established empirically. In the analysis described in this section of the specification the image frame is shrunk from 512×512 pixels to 64x64 pixels because this has appeared in practice to be the minimum resolution that would be useful, but it sets a limit to be accuracy with which the cutting plane can be located. By way of example, the resolution of Figures 3a and 3b is 256×256 pixels. The amount of data required to describe an image is further reduced by run-coding each image. This consists of recording the distances between the silhouette boundaries, i.e. transitions from white to black or vice versa, in each of the 64 horizontal scans of the image.

The most important parts of a plant in the image analysis are the callus and the lower parts of the shoots, but with conventional software the scanning starts at the top of the image frame. In order that the scanning should start at the bottom of each callus, the images are inverted

beforehand. However, in the remainder of this section, the scan is conceptually regarded as proceeding up the plant image rather than down the image frame.

After the simplification of the image data the next step is to identify various components of the image. The general principles have already been discussed. The essence of the technique is as follows. The clusters of stems branching out from the callus are like branches growing from the top of the trunk of a pollarded tree, and it is convenient to supplement the run-code with a

tree-like array of labels in order to identify, by its label, the part of the image to which any item in the run code belongs. See for example the labelling with numbers illustrated in Figure 5.

Further processing begins with the automatic removal of very minor shoots and other unimportant protuberances from the stored image of the plant. An explanation and details of this are given below. Afterwards, as explained above, one of the visible gaps between the remaining shoots is automatically selected to contain the top end of a cutting line through the callus, and the cutting line is then chosen. Only one cutting line per image is chosen, not only because that is simpler than choosing several lines at once, but also because it is often best to view any cut pieces at other angles before choosing another cutting plane. The areas of prescribed lower parts of the gaps between shoots are only one of the items taken into account in selecting such a gap to contain the top end of the cutting line. Some gaps were located in more

favourable sites for the top of a cutting line than others, and it is necessary to weigh all the evidence in favour of a particular gap. A ten term discriminant function, (described above) is evaluated, and the gap with the maximum discriminant function is chosen to contain the top end of the cutting line. The cutting line, when

extrapolated, passes through a pivot point near the bottom of the chosen gap and, as explained above, its slope is determined by both the shape of the gap and the shape of the callus with the relative importance of the shape of the callus set by a parameter value. The various empirical parameter values required by the algorithm, were chosen, in a particular case, during trials on 62 images but trials on a larger number of images is recommended. An automatic method of selecting the

parameter values was employed.

Although the run code forms a complete description of an image it is not immediately obvious from the run code which element of any of the silhouette boundaries found in one horizontal scan is joined to an element in a

neighbouring scan. In some scans new sections of boundary appear where, for instance, several shoots appear on the top of a callus, while in other scans pairs of such

sections end at the tops of shoots. Thus by labelling the items in the run code with labels for these boundary sections, the shoots and the gaps between them can be identified easily. This labelling process is part of the cross referencing referred to previously in step B of the flow chart of Figure 2. A set of rules are applied in order to determine how the items in the run code should be labelled.

Implementation of the rules is somewhat complicated, but in essence the rules are simple. A computer program is designed to look for continuations of the boundaries of a silhouette in successive horizontal scans of its image.

These continuations have to satisfy the rules. Firstly the coding has to ensure that a white-to-black transition is not regarded as a continuation of a black-to-white

transition or vice versa, and secondly it has to ensure that any continuations are unique and genuine with no intervening gaps. A way of interpreting the rules is given below, but an implementation of the rules may be programmed in several ways.

The following interpretation of the rules used for labelling the items in the run code may be helpful. A boundary element of the silhouette found in a horizontal scan may be regarded as a continuation of a similar sort of transition in the previous scan if there are no intervening elements of other boundaries in either scan between or, in some circumstances, at their horizontal locations in the image. The circumstances in which another element has to be counted as being in the intervening interval considered can be clarified as follows. An interval may include or exclude its end points, and if it includes one of its end points then that end of the interval is mathematically regarded as closed. In order to ensure uniqueness of postulated boundary continuations between elements, the intervening horizontal intervals must be regarded as being open at only one end. For white-to-black transitions the intervals need to be regarded as closed on the left and open on the right in order to correspond to the labelling process programmed, while for black-to-white transitions they need to be regarded as open on the left and closed on the right.

Every time a horizontal scan encounters edges of a silhouette which are not continuations of those in the previous scan, they are assigned to new boundary sections. These sections are labelled by a sequence of numbers, and a negative sign is attached to the numbering of every

black-to-white boundary in order to distinguish it from a white-to-black boundary. Figure 5 illustrates how a silhouette image is labelled. A record is compiled of the numbering of the boundary sections in each scan in the same sequence as the distances between the boundaries in each scan in the run code. Thus identification of elements of the run code with the boundaries of different parts of the plants is facilitated.

As an additional aid to the identification of

different parts of the plants, an extra array of

information is compiled about the beginnings and endings of each boundary section. This forms an additional part of the cross referencing referred to instep B of the flow chart of Figure 2. Not only are the starting and finishing points of each boundary section recorded in the array, but also the numbering of the boundary sections linked to them at each end. The array acts as a pointer from the labels of boundary sections to their terminations. This will now be illustrated by reference to Figure 5. In the

arrangement described, the computer software automatically counts array elements from zero instead of one, and so the 64 horizontal scans are numbered from 0 to 63. For Figure 5 boundary section 1 is recorded as linked to those

labelled -2 and -5, and it starts at scan 0 transition 0 and ends at scan 8 transition 0. The second boundary section is linked section is linked to 1 and 3, and it starts at scan 0 transition 1 and ends at scan 1 transition 1. The remaining boundary sections are recorded in the same sort of way.

Next, there will be discussed the presence of very minor shoots and other unimportant protuberances in the image which makes it more difficult to select a suitable position for the top end of the cutting line. Minor shoots or pimples between more major shoots divide the gaps between the major shoots into smaller gaps which may then be too minute to be assigned the top end of the cutting line. By removing the very minor shoots and pimples, the gaps between the more important shoots can be detected more easily. A computer program may be written to remove pimples and unimportant shoots, and it should include in its input data, a parameter which specifies the maximum height of shoots and pimples to be removed. After testing the effect of the several trial values, a height of 1 pixel in the 64x64 image frame has been chosen in a particular case. The information compiled about the beginnings and endings of each section of the boundaries of the silhouette allows the minor shoots and pimples to be identified and removed from the run code easily. Afterwards the remaining sections of the run code are relabelled.

There will now be described a number of general techniques, and factors to be taken into account when putting the invention into operation, and also some

improvements and refinements in the techniques which can be used.

The selection of a suitable cutting line on an image is only appropriate if the plants are viewed from an angle at which the image plane is perpendicular to a suitable cutting plane. Ways of automatically selecting the angle of rotation of the plants, in order to achieve this may be considered.

The slope estimated for a cutting line is nearly always acceptable, even if not ideal, with the simple arrangements described above. However, the location of pivot points can be improved in some voids by using more of the context of the location in positioning the pivot points. In some of the wider gaps between clusters of stems, the pivot point has been found to be too far from the centre of the gap, and it would be advantageous to allow the region of void used in calculating the pivot point to be deeper in at least some of the wide gaps.

The accuracy achieved in locating the cutting line is limited by the coarseness of the resolution of the image in a 64x64 pixel frame. Use of a 128×128 pixel image frame allows a more detailed view of the image and more precision in the placement of the cutting line, but it doubles the size of the run code extracted and the subsequent

processing time required to calculate the cutting line. In particular, placement of the pivot point can be improved by this means, if it does not make the operating rate too slow. In the embodiments which have been described,

reference has been made mainly to the selection of a cutting line for cutting a plant. However, it has been explained that, when the cut is made in respect of a three dimensional plant, the cut is made along a cutting plane. The cutting plane consists of a plane containing the cutting line, and containing the direction of viewing of the plant which has given rise to the image in which the cutting line has been selected. As has been mentioned, in a modified form of the invention, there may be provided arrangements for rotating the plant through a series of angular positions, and viewing the plant at each said position to generate an image signal representing an image of the plant at that angular position. In a preferred form, the modified method then proceeds by selecting a void suitable to include the starting point of a cutting line, not only by comparing voids within a single image, but also by comparing voids from different images taken at different angular positions of the plant. The selected void can be selected in

accordance with any one or more of the tests set out hereinbefore for selecting a void from a single image, but in addition, further tests may be applied which involve reference to parameters of the plant, when viewed at the different angular positions. Where tests are applied to select voids from different images, the same tests may be used as for voids from the same image, but are preferably weighted differently. That is to say that, in the use of the formula set out hereinbefore, the constants P₁, P₂ and so on will be given different values.

In a simple method of selecting a suitable cutting plane, the plant may be rotated manually, and a suitable viewing angle can be selected by an operator. However it is much preferred that there is provided automatic

selection of the viewing angle of the plant. In such a case the signal processing apparatus may be arranged to process a number of images, for example selected by

rotating the plant by, say, approximately 10º between the generation of image signals. The signals are then

processed automatically by the processing means, to select the most suitable void from a number of images, in

accordance with the tests set down above and additional tests referred to below. An example of the modified method of automatically selecting the video camera viewing angle in order to completely specify the cutting plane, will now be described in detail with reference to syngonium plants. In the modified method of the invention, each

syngonium callus is rotated in front of the video camera in order to capture a number of views of it, and for each view a cutting line on its image is automatically selected as set out before. The features already used in selecting that cutting line are used again, together with other features, as inputs to a linear discriminant function for selecting a suitable cutting plane. The width of an image of a callus is often a good guide as to whether that image is the most suitable view of the callus, and in a simple system this width can be used as the only additional feature in selecting the most suitable image. However, it is not possible to select the best view of many plants by using only the callus width and the ten original features set out hereinbefore with regard to the unmodified method. The ratio of the width of the callus to its centroid height is a more suitable feature, but other additional features may be included as factors. The cutting line discriminant function set out above with regard to the unmodified method, already contains a factor labled w_j which indicates whether the two clusters of shoots on either side of the selected void j contain unequal amounts of plant material. However, for the viewing angle discriminant function, the ratio of the square of w_j to the width of the callus is also included in order to emphasize the selection of views with more centrally placed cutting lines.

Preferably, for selecting the viewing image giving the best void, the following factors, additional to the first ten factors described in the unmodified method, are as follows:-

Factor 11 Ratio (multiplied by ten) of callus

width to its centroid height.

Factor 12 Callus centroid height. Factor 13 Ratio (multiplied by ten) of the square of w_j to the callus width.

Claims

1. A method of selecting a cutting line for cutting a plant comprising shoots growing from a base region of the plant, comprising

generating an image signal representing an image of the plant, and

processing the image signal to carry out the steps of locating voids in the image of the plant, selecting a void suitable to include the starting point of a cutting line for cutting the base region of the plant, and defining a cutting line extending from the selected void through the base region of the plant;

the said void being selected by reference to:

(i) the area of the void or one or more selected

parts of the void, alone or in combination with the area of one or more neighbouring voids or selected parts thereof; and/or

(ii) the horizontal location of the void; and/or (iii) the vertical location of a bottom region of the void; and/or

(iv) the slope of a proposed cutting line determined for the void.

2. A method according to claim 1 including evaluating for each void at least two of the following functions, namely a function related to the area of the void or one or more selected parts of the void, alone or in combination with the area of one or more neighbouring voids or selected parts thereof; a function related to the horizonal location of the void; a function related to the vertical location of a bottom region of the void; and a function related to the slope of a proposed cutting line determined for the void; and selecting that void which has a preferred weighted combination of said at least two functions.

3. A method according to claim 1 in which the void is selected by reference to the area of the void, or, where the vertical depth of the void is greater than one or more predetermined values, with reference to the area or areas of parts of the void bounded by the bottom of the void and extending up the void to the said predetermined value or values of vertical depth.

4. A method according to claim 1 in which the said void is selected by reference to selected portions of

neighbouring voids diminished by factors relating to the linear and angular displacement between at least one reference point in the void and in each of the neighbouring voids.

5. A method according to claim 4 in which the void is selected by reference to the sums of products over

neighbouring voids while in each sum, such a product is that of a variable factor with the area of a portion of a neighbouring void.

6. A method according to claim 1 in which the said void is selected by reference to the horizontal location of the void relative to the centroid of the base region of the plant.

7. A method according to claim 1 in which the said void is selected by reference to the horizontal location of the void relative to a location where, if the void were located there, on average there would be an even distribution of the image of plant material on either side of the void.

8. A method according to claim 1 in which the void is selected by reference to a measure of the difference in area of plant material shown in the image on either side of the void, or a selected part thereof.

9. A method according to claim 8 in which the void is selected by reference to the difference between the total area of plant material on one side of the void and that on the other side of the void, divided by the vertical depth of the void or, where the void has a depth greater than a predetermined vertical depth, divided by the said

predetermined vertical depth.

10. A method according to claim 1 in which the said void is selected by reference to the height of the bottom region of the void relative either to the bottom of the base region of the plant, or to a reference frame of the image.

11. A method according to claim 1 in which the void is selected by reference to the relationship of the height of the bottom region of the void to the height of the centroid of the base region of the plant.

12. A method according to claim 1 in which the said void is selected by reference to the square of the height of a reference point of the void from the bottom of the base region of the plant, the reference point of the void being the centroid of the void, or where the vertical depth of the void is greater than a predetermined depth, the

centroid of the part of the void bounded by the bottom of the void and extending up to the said predetermined

vertical depth.

13. A method according to claim 1 in which the void is selected by reference to the product of the height of a reference point of the void with the height of the centroid of the base region of the plant measured from the bottom of the base region, the reference point of the void being the centroid of the void, or where the vertical depth of the void is greater than a predetermined depth, the centroid of the part of the void bounded by the bottom of the void and extending up to the said predetermined vertical depth.

14. A method according to claim 1 in which the void is selected by reference to the slope of a proposed cutting line through the base region of the plant, which cutting line has its angle of inclination determined by reference to the shape and/or size of the void, and/or to the shape and/or size of the base region of the plant.

15. A method according to claim 1 in which the said void is selected by reference to the square of a measure of the inclination to the vertical of a proposed cutting line for the void.

16. A method according to claim 1 in which the said void is selected by reference to the relationship of the slope of a proposed cutting line extending from the void through the base region of the plant, to a measure of the slope of the base region.

17. A method according to claim 1 in which the said void is selected by reference to the product of a measure of the inclination to the vertical of a proposed cutting line for the void with that of a line, through the centroid of the image of the base region of the plant, having a least sum of squares fit to the horizontal locations of plant pixels in the image of the base region of the plant.

18. A method according to claim 1 including defining a cutting line for the selected void and determining the angle of inclination of the cutting line relative to a predetermined direction by reference to the shape and/or size of the void, and/or to shape and and/or size of the base region of the plant.

19. A method according to claim 18 including the steps of selecting a pivot point located in the void, setting the cutting line to pass through the pivot point, and setting the slope of the cutting line relative to the vertical by reference to the size and/or shape of the void and/or the base region of the plant.

20. A method according to claim 19 in which the pivot point is located at the centroid of the void, or where the vertical depth of the void is greater than a predetermined value, at the centroid of part of the void bounded by the bottom of the void and extending up the void to the said predetermined value of depth.

21. A method according to claim 19 in which the slope of the cutting line is set as the weighted sum of the slopes of two lines both passing through the pivot point, the slope of the first line being set by reference to the size and/or shape of the void, and the slope of the second line being set by reference to the size and/or shape of the base region of the plant.

22. A method according to claim 21 in which the said weighted sum of the slopes gives more weight to the second line than to the first line.

23. A method according to claim 18 including formulating for the void a line which minimises a measure of the horizontal distances on either side of the line, of the component pixels of the void or a selected part thereof, and setting the slope of the cutting line with reference to the said line.

24. A method according to claim 23 in which the said minimising step comprises minimising the sum of the squares of the horizontal distances on either side of the line, of the component pixels of the void or the selected part thereof.

25. A method according to claim 18 including formulating for the base region of the plant a line which minimises a measure of the horizontal distances on either side of the line, of the component pixels of the base region of the plant or a selected part thereof, and setting the slope of the cutting line with reference to the said line.

26. A method according to claim 25 in which the said minimising step comprises minimising the sum of the squares of the horizontal distances on either side of the line, of the component pixels of the base region of the plant or the selected part thereof.

27. A method of selecting a cutting plane for cutting a plant comprising shoots growing from a base region of the plant, the method comprising

rotating the plant through a series of angular

positions;

viewing the plant at each said position and generating an image signal representing an image of the plant at that angular position; and

processing the image signal from each angular position by a method in accordance with any preceding claim to select for each different image a void suitable to include the starting point of a cutting line for cutting the base region of the plant,

further processing the image signals to carry out the steps of comparing the selected voids in the different images to select from the different images the most

suitable void to include the starting point of a cutting line for cutting the base region of the plant, and defining a cutting plane for cutting the plant as being a plane containing the cutting line of the selected void and

containing the direction of viewing of the plant which gave rise to the image containing the selected void.

28. A method according to claim 27 in which the selection of a void from voids occurring in different images taken at different angular positions, is made, alone or in

combination with other selection procedures, by reference to:

( i ) the area of the void or one or more selected

(ii) the horizontal location of the void; and/or (ϋi) the vertical location of a bottom region of the void; and/or

(iv) the slope of a proposed cutting line determined for the void.

29. A method according to claim 27 in which the parameters used when comparing voids from different images of the plant taken at different angular positions, are different from the parameters used when comparing voids from the same image taken at a single angular position.

30. A method according to claim 27, in which the selection of a void from voids occurring in different images taken at different angular positions, is selected, alone or in combination with other selection procedures by reference to:-

and/or

31. A method according to claim 30 in which the selection of a void from voids occurring in different images taken at different angular positions, is selected, alone or in combination with other selection procedures by reference to:-

(ii) a function related to the height of the centroid of the base region of the plant; and/or

(iii) a function related to the ratio of the square of a difference function to the width of the base region of the plant, where the difference function is the difference between the total area of plant material on one side of the void and that on the other side of the void divided by the vertical depth of the void or, where the void has a depth greater than a predetermined vertical depth, divided by the said

predetermined vertical depth.

32. A method of selecting a cutting plane for cutting a plant comprising shoots growing from a base region of the plant, comprising:

rotating the plant through a series of angular

positions;

processing the image signals to carry out the steps of locating voids in the images of the plant, selecting a void suitable to include the starting point of a cutting line for cutting the base region of the plant, defining a selected cutting line extending in the image of the

selected void from the selected void through the base region of the plant, and defining a cutting plane as being a plane containing the selected cutting line and containing the direction of viewing of the plant which gave rise to the image containing the said selected void.

33. A method of selecting a cutting line for cutting a plant comprising shoots growing from the base region of the plant, comprising

generating an image signal representing an image of the plant, and

processing the image signal to carry out the steps of locating a selected void in the image of the plant,

defining a cutting line extending from the selected void through the base region of the plant, and determining the angle of inclination of the cutting line relative to a predetermined direction;

the angle of inclination of the cutting line being determined by reference to:

the shape and/or size of the void, and/or

the shape and/or size of the base region of the plant.

34. A method of cutting a plant comprising shoots growing from a base region of the plant, comprising the steps of selecting a cutting line in accordance with the method of claim 1, generating a control signal related to the cutting line, and cutting the plant along the cutting line by robotic cutting means under the control of the said control signal.

35. Apparatus for use in micropropagation, comprising

means for generating an image signal representing an image of a plant comprising shoots growing from a base region of the plant; and

signal processing means for processing the image signal to locate voids in the image of the plant, to select a void suitable to include the starting point of a cutting line for cutting the base region of the plant, and to generate an output signal containing information as to the location of a cutting line extending from the selected void through the base region of the plant;

the signal processing means being arranged to select the said void by reference to:

( i ) the area of the void or one or more selected

(iv) the slope of a proposed cutting line determined for the void.

36. Apparatus according to claim 35 further comprising: means for rotating a plant comprising shoots growing from a base region of the plant, through a series of angular positions; and

means for viewing the plant at each said position and for generating an image signal representing an image of the plant at that angular position;

the said signal processing means being arranged to process the image signal from each angular position to select for each different image a void suitable to include the starting point of a cutting line for cutting the base region of the plant, and to compare the selected voids in the different images to select from the different images the most suitable void, said output signal containing information defining a cutting plane as being a plane containing the selected cutting line and containing the direction of viewing of the plant which gave rise to the image containing the said selected void.

37. Apparatus for use in micropropagation, comprising:

means for rotating a plant comprising shoots growing from a base region of the plant, through a series of

angular positions;

means for viewing the plant at each said position and for generating an image signal representing an image of the plant at that angular position; and

signal processing means for processing the image signals to locate voids in the images of the plant, to select a void suitable to include the starting point of a cutting line for cutting the base region of the plant, and to generate an output signal containing information as to the location of a cutting line extending in the image containing the selected void from the selected void through the base region of the plant, said output signal containing information defining a cutting plane as being a plane containing the selected cutting line and containing the direction of viewing of the plant which gave rise to the image containing the said selected void.

38. Apparatus for use in micropropagation, comprising means for generating an image signal representing an image of a plant comprising shoots growing from a base region of the plant; and

signal processing means for processing the image signal to locate a selected void in the image corresponding to a gap between shoots of the plant, to define a cutting line extending from the selected void through the base region of the plant, and to generate an output signal containing information as to the location of the said cutting line;

the signal processing means being arranged to

determine the angle of inclination of the cutting line by reference to:- the shape and/or size of the void; and/or

the shape and/or size of the base region of the plant.

39. Apparatus according to claim 37 in which the said signal processing means is arranged to select a void by reference to:

(i) the area of the void or one or more selected parts of the void, alone or in combination with the area of one or more neighbouring voids or selected parts thereof; and/or

(ii) the horizontal location of the void; and/or

(iii) the vertical location of a bottom region of the void; and/or

(iv) the slope of a proposed cutting line determined for the void.

40. Apparatus according to claim 35 further comprising means for presenting the plant at a work station, and robotic cutting means for cutting the plant along the cutting line under the control of the said output signal,