CN100487370C - Digital identification method of levelling measurement - Google Patents

Abstract

First, image of leveling staff in viewing field is obtained through telescope and image acquisition system. Characters are that following steps are included in the method: determining position of rectangular stripes in sampled image; based on spatial relation between rectangular stripes and digit labels, determining zones of digit labels, and separating the zones of digit labels from other image obtained so as to obtain image blocks containing each digits; carrying out correlation calculation between sampled image and reference image created by algorithm of generating image of standard digit; carrying out relevant error checking and correction for result of recognizing digits; combining recognized digits to obtain final recognized result-this is a reading of large number from leveling staff. The invention guarantees correctness and reliability of result of recognizing digits.

Description

A kind of digit recognition method of measurement of the level

Technical field

The present invention relates to a kind of digit recognition method of measurement of the level.

Background technology

Spirit-leveling instrument in using at present mainly contains two kinds of optical level and digital levels.

The spirit-leveling instrument that optics measurement of the level system is made up of the telescope that has level and graticule, self leveling device, and the levelling pole composition that has numeral and special delineation.Levelling pole is the striped of the first-class width of equally spaced printing on on-deformable tall and thin straight sheet material, and at the other numerical reference of striped.The measurement of the level of carrying out with optical level be with the partial image of the levelling pole of spirit-leveling instrument horizontal line of sight equal height, be imaged near the graticule of spirit-leveling instrument eyepiece, the surveyor is with level silk the position on levelling pole of eyes by telescopical eyepiece surveyor's level graticule, and according to the numerical reference on the levelling pole, obtain big reading (rice and decimetre or centimetre), and the geometric relationship of the level silk of graticule and levelling pole delineation obtain little reading (centimetre, millimeter and less than the part of millimeter), can obtain the reading of the levelling pole of spirit-leveling instrument horizontal line of sight equal height.

The spirit-leveling instrument that the digital leveling measuring system is made up of the telescope that has level and graticule, self leveling device, ccd sensor and computer system, and the levelling pole composition that has special delineation.Levelling pole is the non-striped that waits width in the printing of unequal interval according to certain rules on the on-deformable tall and thin straight sheet material, numerical reference not on the levelling pole.The measurement of the level of carrying out with digital level be with the partial image of the levelling pole of spirit-leveling instrument horizontal line of sight equal height, be imaged near the graticule and ccd sensor the spirit-leveling instrument eyepiece sampling of computer control CCD, processing.According to the image of the sampling of CCD and the relativeness of levelling pole delineation, obtain the reading of the levelling pole of spirit-leveling instrument horizontal line of sight equal height.

International optics measurement of the level system can satisfy the measurement of the level of each grade.The spirit-leveling instrument that China produces at present all is an optics measurement of the level system, and the spirit-leveling instrument of the various models of production and levelling pole can satisfy respectively registers measurement of the level except that primary levelling.And in long-term research and production, dropped into great amount of manpower and material resources, obtain many achievements and standard, and had many factories.The shortcoming of optics measurement of the level system is to realize that not digitizing, operating efficiency are lower than the digital leveling measuring system, and makes mistakes easily.

The digital leveling measuring system that China uses at present all is (the Japan of import, Switzerland and Germany), because its special measuring method, force China must carry out the research of the leveling norm that adapts with it, but also will drop into a large amount of funds development annexes supporting with it.

The investment of protection China, the preferred plan that makes full use of existing resources are exactly to realize the digitizing of optics measurement of the level system.It is one of gordian technique of realizing this goal that errorless knowledge and nothing refuse to know digit recognition.

Existing digital recognition methods at present all exists to know by mistake and refuse to be known, and the environment of measurement of the level is open-air, levelling pole is scratched unavoidably or dirt blocks, and measurement of the level does not allow to occur to know and refuse by mistake to know, and this makes existing digit recognition method can't satisfy the errorless knowledge of measurement of the level and the harsh requirement that nothing refuses to know digit recognition.

Summary of the invention

Purpose of the present invention just provides the measurement of the level digit recognition method that a kind of good reliability, errorless knowledge and nothing refuse to know, and promptly utilizes the digit recognition technology to realize the automatic identification of obtaining the numerical reference on the levelling pole automatically of the big reading of levelling pole.

The object of the present invention is achieved like this: a kind of digit recognition method of measurement of the level, at first telescope and the image capturing system by spirit-leveling instrument obtains the image of levelling pole in the range of telescope of spirit-leveling instrument, it is characterized in that also comprising the steps:

Separation number word image: the position of search rectangle striped in the view field image of sampling, geometric relationship according to rectangle striped and numerical reference, and the position of rectangle striped, determine the position of numerical reference in sampled images, from sampled images, isolate the numerical reference image, from the numerical reference image, separate the image of each bit digital;

Digit recognition is with synthetic: the sampled images of each bit digital that will separate carries out correlation computations with the reference picture that generates by standard digital image generating algorithm, the digital picture image height of reference picture image height=separation, reference picture numeral k=0-9, correlation computations is undertaken by following formula:

$C_k(i_0,j_0)=\frac{1}{m\×n}\underset{i=0}{\overset{m}{\mathrm{\Σ}}}\underset{j=0}{\overset{n}{\mathrm{\Σ}}}{\mathrm{CY}}_{i-i_o,j-j_o}\×{\mathrm{CK}}_{k,i,j}$

CY _{I, j}: the i of sampled images is capable, j row pixel,

CK _{I, j}: the i of reference picture is capable, j row pixel,

C _k: with respect to correlation computations value reference picture and sampled images of k=0-9,

i ₀: with respect to the capable zero migration of sampled images,

j ₀: with respect to the row zero migration of sampled images,

K: corresponding to numeral 0-9;

Capable null drift i according to different sampled images ₀, the row null drift j of sampled images ₀, calculate correlation C _{K, i, j}, with maximum correlation value C _{K, i, j}As C _kWith C ₀, C ₁, C ₂, C ₃, C ₄, C ₅, C ₆, C ₇, C ₈, C ₉In maximal value as final recognition result;

Recognition result error checking and correction (ECC): a plurality of numerical references in the levelling pole sampling are discerned, and it is relevant up and down to carry out the homonymy of recognition result the check that offside is relevant up and down, C ₁Recognition result on the=offside-1, C ₂Recognition result under the offside+1, C ₃Recognition result on the=homonymy-2, C ₄Recognition result under the=homonymy+2, this position recognition result of C=is worked as C ₁=C or C ₂=C or C ₃=C or C ₄During=C, recognition result is correct; As find mistake, and can carry out error correction according to correlativity, finally obtain the digit recognition result that errorless knowledge and nothing refuse to know; Merge the numeral of identification respectively, both obtained the big reading of measurement of the level.

Described digit recognition is with synthetic, by with digital sample picture altitude of separating and actual numeral height, generate reference picture by different ratios, make the reference picture of generation have identical size with sampled images.

Described recognition result error checking and correction (ECC), numerical relation according to adjacent numeral, the difference of the numerical reference that the levelling pole homonymy is neighbouring is 2 or 4, and the difference of the numerical reference that offside is neighbouring is 1 or 2, carries out adjacent digit recognition result's check and error correction.

Existing digit recognition scheme all has two indexs of misclassification rate and reject rate, and just numerical values recited has difference, and this programme recognition result is the digit recognition result that errorless knowledge and nothing refuse to know.

The present invention uses the relevant error checking and correction (ECC) scheme of correlated digital identification and recognition result, digit recognition result's correctness and reliability have been guaranteed, this obtains the digit recognition result that errorless knowledge and nothing refuse to know for the measurement of the level of carrying out under the lowered in field environment, seems particularly important.

Description of drawings

Fig. 1 is a levelling pole sampling synoptic diagram in the visual field;

Fig. 2 is an overview flow chart;

Fig. 3 is digital picture separation process figure;

Fig. 4 is the digit recognition process flow diagram;

Fig. 5 is a recognition result error checking and correction (ECC) process flow diagram;

1. visual fields among the figure; 2. levelling pole; 3. rectangle striped; 4. numerical reference; In the level silk.

Embodiment

A kind of digit recognition method of measurement of the level, as shown in Figure 1, at first telescope and the image capturing system by spirit-leveling instrument obtains the image of levelling pole in the range of telescope of spirit-leveling instrument.The position of search rectangle striped in the image that obtains; According to the geometric relationship of rectangle striped and numerical reference, and the position of rectangle striped, determine that the numerical reference image block is isolated to finish from the image that obtains in the zone at numerical reference place.Every digital separation in the isolated numerical reference image block is opened, obtained the image block of each bit digital of separation.The sampled images of each bit digital of separating is carried out correlation computations with the reference picture that generates by standard digital image generating algorithm, and correlation computations is undertaken by following formula.

$C_k(i_0,j_0)=\frac{1}{m\×n}\underset{i=0}{\overset{m}{\mathrm{\Σ}}}\underset{j=0}{\overset{n}{\mathrm{\Σ}}}{\mathrm{CY}}_{i-i_0,j-j_0}\×{\mathrm{CK}}_{k,i,j}$

CY _{I, j}: the i of sampled images is capable, j row pixel.

CK _{I, j}: the i of reference picture is capable, j row pixel.

C _k: with respect to correlation computations value reference picture and sampled images of k=0-9.

i ₀: with respect to the capable zero migration of sampled images.

j ₀: with respect to the row zero migration of sampled images.

K: corresponding to numeral 0-9.

According to the image height of sampled images CY and different numerals (0-9),, generate different reference picture CK according to the standard picture generating algorithm of numerical reference.According to the result of correlation computations, adjust image line zero migration and row zero migration, with the value of the correlation computations of best matching result-maximum recognition result as specific numeral.Capable null drift i according to different sampled images ₀, the row null drift j of sampled images ₀, calculate correlation C _{K, i, j}, with maximum correlation value C _{K, i, j}As C _kWith C ₀, C ₁, C ₂, C ₃, C ₄, C ₅, C ₆, C ₇, C ₈, C ₉In maximal value as final recognition result.Because the adjacent numerical reference of levelling pole is correlated with, the numerical reference on levelling pole right side is odd number (even number), and then the numerical reference in levelling pole left side just is even number (odd number), or both sides are even number, and this is relevant about being.The difference of the numerical reference that homonymy is neighbouring is 2 or 4, and the difference of the numerical reference that offside is neighbouring is 1 or 2, in case selected levelling pole, the relation of the numerical reference that it is adjacent has just been determined.A plurality of numerical references in the levelling pole sampling are discerned, and it is relevant up and down to carry out the homonymy of recognition result the check that offside is relevant up and down, C ₁Recognition result on the=offside-1, C ₂Recognition result under the offside+1, C ₃Recognition result on the=homonymy-2, C ₄Recognition result under the=homonymy+2, this position recognition result of C=is worked as C ₁=C or C ₂=C or C ₃=C or C ₄During=C, recognition result is correct; As find mistake, and can carry out error correction according to correlativity, finally obtain the digit recognition result that errorless knowledge and nothing refuse to know.Merge the numeral of identification respectively, both obtained the big reading of measurement of the level.

When digit recognition is carried out in practice the present invention's utilization, at first levelling pole 2 is vertically placed, made clear picture by focus control, and make the image of levelling pole be positioned at visual field 1 central authorities.By the sampled images of levelling pole in the image capturing system acquisition visual field, in the sampled images that obtains, determine the position of rectangle striped 3; According to the spatial relationship of rectangle striped and numerical reference 4, determine the zone at numerical reference place, the numerical reference image is separated from the image that obtains.And each bit digital is separated from isolated numerical reference image block, obtained the image block of each bit digital of separation.The sampled images of each bit digital of separating is carried out correlation computations with the reference picture that generates by standard digital image generating algorithm, to the result of the digit recognition error checking and correction (ECC) of being correlated with, merge the numeral of discerning, both obtained the big reading of correct digit recognition result-levelling pole.Digit recognition is handled and the error correction flow process is seen Fig. 2, Fig. 3, Fig. 4 and Fig. 5.

Claims (3)

1. the digit recognition method of a measurement of the level at first by the telescope and the image of image capturing system acquisition levelling pole in the range of telescope of spirit-leveling instrument of spirit-leveling instrument, is characterized in that also comprising the steps:

Separation number word image: the position of search rectangle striped in the view field image of sampling, geometric relationship according to rectangle striped and numerical reference, and the position of rectangle striped, determine the position of numerical reference in sampled images, from sampled images, isolate the numerical reference image, from the numerical reference image, separate the image of each bit digital;

Digit recognition is with synthetic: the sampled images of each bit digital that will separate carries out correlation computations with the reference picture that generates by standard digital image generating algorithm, the digital picture image height of reference picture image height=separation, reference picture numeral k=0-9, correlation computations is undertaken by following formula:

$C_k(i_0,j_0)=\frac{1}{m\×n}\underset{i=0}{\overset{m}{\mathrm{\Σ}}}\underset{j=0}{\overset{n}{\mathrm{\Σ}}}{\mathrm{CY}}_{i-i_0,j-j_0}\×{\mathrm{CK}}_{k,i,j}$

CY _{I, j}: the i of sampled images is capable, j row pixel,

CK _{I, j}: the i of reference picture is capable, j row pixel,

C _k: with respect to correlation computations value reference picture and sampled images of k=0-9,

i ₀: with respect to the capable zero migration of sampled images,

j ₀: with respect to the row zero migration of sampled images,

K: corresponding to numeral 0-9;

Capable null drift i according to different sampled images ₀, the row null drift j of sampled images ₀, calculate correlation C _{K, i, j}, with maximum correlation value C _{K, i, j}As C _kWith C ₀, C ₁, C ₂, C ₃, C ₄, C ₅, C ₆, C ₇, C ₈, C ₉In maximal value as final recognition result;

Recognition result error checking and correction (ECC): a plurality of numerical references in the levelling pole sampling are discerned, and it is relevant up and down to carry out the homonymy of recognition result the check that offside is relevant up and down, C ₁Recognition result on the=offside-1, C ₂Recognition result under the offside+1, C ₃Recognition result on the=homonymy-2, C ₄Recognition result under the=homonymy+2, this position recognition result of C=is worked as C ₁=C or C ₂=C or C ₃=C or C ₄During=C, recognition result is correct; As find mistake, and can carry out error correction according to correlativity, finally obtain the digit recognition result that errorless knowledge and nothing refuse to know; Merge the numeral of identification respectively, both obtained the big reading of measurement of the level.

2. the digit recognition method of a kind of measurement of the level according to claim 1, it is characterized in that: digit recognition is with synthetic, by with digital sample picture altitude of separating and actual numeral height, generate reference picture in different ratios, make the reference picture that generates have identical size with sampled images.

3. the digit recognition method of a kind of measurement of the level according to claim 1, it is characterized in that: the recognition result error checking and correction (ECC), numerical relation according to adjacent numeral, the difference of the numerical reference that the levelling pole homonymy is neighbouring is 2 or 4, the difference of the numerical reference that offside is neighbouring is 1 or 2, carries out adjacent digit recognition result's check and error correction.

Images