WO2004112685A1

WO2004112685A1 - Drug auditing device

Info

Publication number: WO2004112685A1
Application number: PCT/JP2004/008937
Authority: WO
Inventors: Hiroyuki Yuyama; Gakuji Goto; Katsumasa Shirai; Ken Kodama; Yoshito Goto; Hiroyasu Hamada
Original assignee: Yuyama Mfg. Co., Ltd.
Priority date: 2003-06-19
Filing date: 2004-06-18
Publication date: 2004-12-29
Also published as: JPWO2004112685A1; TW200504620A; JP4857768B2; KR20060019606A

Abstract

A drug auditing device comprising a storing means (19) for storing in advance the images of a large number of drug containers (8) in association with the names of drugs, an auditing table (4) for mounting thereon a drug container (8) to be audited, an imaging means (5) for picking up the image of a drug container (8) on the auditing table (4), a judging means (18) judging whether or not an image obtained by the imaging means (5) can be found in images stored in the storing means (19). Accordingly, drugs can be audited quickly and easily, and an accurate collation can prevent an auditing error.

Description

Specification

Technical field .

The present invention relates to a medicine inspection device for automatically inspecting medicine such as an ampule. Background art

Auditing whether or not drugs, such as tablets, powders, and injections, to be administered to patients at hospitals and dispensing pharmacies fall under prescriptions is very important in preventing medical mistakes. Various devices have been proposed that perform this automatically. For example, a device for inspecting the quality of a tablet (for example, see Japanese Patent Application Laid-Open No. 2000-276583), a device for detecting the number of tablets (for example, Japanese Patent Application Laid-Open No. 7-200770, Japanese Patent Application Laid-Open No. 204253, JP-A-9-1299448, and JP-A-11-206855). Inspection work to check whether or not injections such as ampoules correspond to injection notes was manually performed visually by the auditor. In other words, the inspector visually checked the label attached to the surface of the dispensed ampule, read the name of the drug, and checked whether or not it was the same as the drug described on the injection note.

Such visual inspection was a very cumbersome task, using nerves to prevent audit errors. Also, for example, if 800 prescriptions a day (an average of 5 ampoules per prescription) are audited by six people, the number of supervisions per person is less than 700, which means a lot of time and labor. ·

In particular, injections have not only similar appearances but also similar names or cases with the same name but different contents (contents are included at the end of the name). . For example, very confusing names are used, such as Utemerin and Methenaline, Adalart and Adatat L, Amisalin and Alinamin, Evamir and Eberdale, and similar names. If the logos are similar, the matching rate could be judged to be the same with a probability of more than 90%. If the name is the same and the content is different, for example, 3 ml of medicine A and 8 ml of medicine A, When the matching rate was 98% or more, there was a problem that the two were judged to be the same. For this reason, if 8 ml of medicine A was dispensed by mistake when the order was 3 m1 for medicine A, there was a risk that the patient would be administered twice or more the amount of medicine. ■

Also, when a drug is co-injected into an infusion and provided to a patient, the co-infusion work has already been completed at the time of administration, so it is not possible to confirm that the co-infusion has been performed as prescribed and the history of co-infusion work cannot be known There was. Disclosure of the invention

The present invention has been made in view of the above-mentioned problems, and provides a drug monitoring device that can quickly and easily perform a drug monitoring operation and that can prevent an audit error by accurate collation. It is an issue.

Another object of the present invention is to provide a medicine inspection device capable of leaving a history of co-injection work.

In order to solve the above problems, the present invention provides

Storage means for storing in advance images of a large number of drug containers in association with their drug names,

An audit table equipped with drug containers to be audited,

Imaging means for capturing an image of a medicine container mounted on the monitoring stand;

Determining means for determining whether or not the image obtained by the imaging means includes the image stored in the storage means.

Here, the medicine container includes not only a medicine container having a circular cross section such as an ampoule but also a medicine container having a non-circular cross section such as an injection kit or an infusion bag. As the imaging means, a CCD camera is preferable.

With the above-described configuration, the present invention stores an image of each medicine container even if the medicine container has various forms, so that the memorized image is included in the image of the medicine container taken by the imaging means. By searching for a weak force, drug containers can be identified quickly and easily, and accurate collation can be performed.

Preferably, the storage means stores an image of a characteristic portion of the medicine container. Here, the characteristic part is a special outer part of the container or a special table of drug names. The part that can be distinguished from other drug containers, such as the current character or the character of a special logo. In this way, by storing the image of the characteristic portion of the medicine container in the storage means, it is possible to perform quicker and more accurate collation.

The storage means stores images of a plurality of characteristic portions of the medicine container,

It is preferable that the determination means determines whether or not the image obtained by the imaging means has any of a plurality of characteristic images stored in the storage means. In this way, by storing images of a plurality of characteristic portions of the medicine container, it is possible to quickly and accurately check even a misleading or medicine container.

The discriminating means, when judging whether or not the image obtained by the imaging means has any image of the characteristic part stored in the storage means, a threshold value different for each image of the characteristic part It is preferable to judge by. As a result, the matching rate is improved, and the threshold can be lowered.

It is preferable that the inspection table includes means for changing a posture of the medicine container. As a result, it is also possible to take an image of a part hidden from the viewpoint of the imaging means, and to perform an accurate audit.

The means for changing the posture of the medicine container is preferably a pair of rollers that support and rotate the medicine container. Thus, the entire circumference of the medicine container can be imaged only by mounting the medicine container on the inspection table.

Further comprising a light source for irradiating the medicine container with light,

Preferably, the light source is dimmable and has frosted glass for uniform brightness. Thus, there is no shade of light, and the medicine can be accurately discriminated. It is preferable that a white color plate is provided behind the medicine container so that the image of the medicine container has a white background. As a result, even if the transparent medicine container contains a transparent medicine, the background becomes white, and the name of the medicine container surface can be clearly imaged.

It is preferable that the imaging means captures a planar image of the medicine container and an image of another surface. Thus, accurate determination can be made based on both the planar image of the medicine container and the images of the other surfaces.

The image of the other surface of the medicine container is a back surface image of the medicine container, and the imaging unit It is preferable to capture the back surface image reflected by the mirror body in the imaging direction. Thus, the imaging unit can capture both the planar image and the back surface image of the medicine container, and can perform accurate determination.

Image acquisition means for acquiring an image of the medicine container stored in the storage means from the imaging means,

The image obtaining means cuts out an image of a characteristic portion selected from the obtained image of the medicine container, and filters the image of the cut out characteristic portion with a predetermined image adjustment threshold value. It is preferable that the image data of the characteristic portion obtained is stored in the storage unit in association with the medicine name together with the image adjustment threshold value. As a result, the stored image can be sharpened, and the reliability of the determination is improved.

It is preferable that the image picked up by the image pickup means is filtered with the same threshold value as the predetermined image adjustment threshold value. This allows the image to be audited to be as sharp as the stored image, improving the speed and reliability of the judgment.

It is preferable that the image adjustment threshold can be changed. As a result, the threshold value can be adjusted according to the imaging environment, the stored image can be made clearer, and the reliability of determination can be further improved.

It is preferable that the determination unit tilts the storage image by a predetermined angle and compares the tilted storage image with the storage image. Even if the drug to be monitored is placed diagonally on the inspection table, accurate determination can be made, and the reliability of the determination is further improved.

The discriminating means includes: a candidate point search for performing an approximate correlation calculation to search for a plurality of catch points similar to the storage image; and a final search for performing a correlation calculation for the candidate points to narrow down an image that matches the storage image. Is preferably performed. By doing so, the speed and reliability of the search are improved.

It is preferable that the threshold value of the candidate point search is smaller than the threshold value of the final search. It is preferable that, when performing the candidate point search, the discriminating means use a simplified storage image reconstructed by removing pixels from the storage image at a constant pitch. As a result, the search speed can be further improved.

The storage means divides images of a large number of medicine containers into a plurality of areas in advance and stores them in association with the medicine names, The determination unit determines a degree of coincidence between the image obtained by the imaging unit and an image of each area stored in the storage unit, calculates a difference in the degree of coincidence of the image of each area, It is preferable to determine whether or not the image obtained by the imaging means includes an image stored in the storage means by comparing the disparity with a predetermined threshold value. By doing so, it is possible to reliably determine even if a part of the image is different.

It is preferable that the apparatus further includes a display unit that displays a history of the determination by the determination unit. Thus, when the drug is administered to a patient, for example, after the drug is monitored, the history can be checked to confirm that the drug has been co-injected as prescribed. BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 is a schematic configuration diagram of a medicine inspection system according to a first embodiment of the present invention. Figure 2 is a side view of the drug inspection device.

Figure 3 is a front view of the audit table.

FIG. 4 is a flowchart showing an operation of storing a characteristic image of a medicine container.

FIG. 5 is a plan view of a medicine container whose shape is a feature image.

Figure 6 is a plan view of a medicine container with character images.

FIG. 7 is a flowchart showing the operation of identifying a medicine container.

FIG. 8 is a front view showing another example of the monitoring stand.

FIG. 9 is a schematic configuration diagram of a medicine inspection system according to the second embodiment of the present invention. FIG. 10 is a perspective view of the monitoring device.

Figure 11 is a cross-sectional view of the lower case of the inspection device.

FIG. 12 is an exploded perspective view of a light source.

Fig. 13 is a cross-sectional view of the audit device at the time of the ampoule audit.

Fig. 14 is a cross-sectional view of the inspection device at the time of inspection of the injection kit.

Figure 15 is a perspective view of the injection kit.

FIG. 16 is a flowchart showing an operation of storing a characteristic image of a medicine container.

FIG. ¹⁷ is a flowchart showing the operation of identifying a medicine container.

FIG. 18 is an image recognition live of the drug inspection system according to the third embodiment of the present invention. The figure which shows the positioning of a rally.

Figure 19 is a diagram showing a panel screen using the image recognition library.

FIG. 20 is a flowchart of image registration.

FIG. 21A is a diagram showing an example of a file name and a feature image when a feature image is captured. FIG. 21B is a diagram showing a form at the time of registration of a feature image.

FIG. 22 is a diagram illustrating a change in the feature image A when the threshold value is changed.

FIG. 23 is a diagram illustrating a change in the feature image B when the threshold value is changed.

FIG. 24 is a diagram illustrating a change in the feature image C when the threshold value is changed.

Figure 25 is a diagram showing the data structure of a feature image.

FIG. 26 is a diagram showing an example in which a feature image is divided.

Figure 27 shows the flow of image inspection.

FIG. 28 is a diagram showing a flow of image determination.

FIG. 29 is a diagram showing a flow of obtaining prescription data judgment results.

FIG. 30 is a diagram showing directions of image search.

FIG. 31 is a diagram showing another example of an image search.

FIG. 32 is a diagram showing a relationship between a stored image and dots of an image to be inspected.

FIG. 33 is a diagram showing an example of a pattern of a stored image.

FIG. 34 is a diagram showing an example of a pattern of a simple storage image.

FIGS. 35A and 35B are diagrams each showing an example of a real image and a determination result thereof.

FIGS. 36A and 36B are diagrams each showing an example of a solid-filled image and its determination result. FIGS. 37A and 37B are diagrams showing a real image partly missing and a determination result thereof. Figure 38 is a diagram showing the initial screen of the mixed injection monitoring system.

Figure 39 shows the menu screen of the mixed injection monitoring system.

Figure 40 shows the mixed injection audit monitor screen.

Figure 41 is an enlarged view of the patient information column on the mixed injection audit monitor screen.

Figure 42 is an enlarged view of the mixed injection information column on the mixed injection monitoring screen.

Figure 43 is an enlarged view of the detail information column on the mixed injection audit monitor screen.

Figure 44 Figure 4 is an enlarged view of the detailed information column displaying the audit results.

Figure 45 shows the message of the infusion audit by barcode reading. Figure 46 shows various environment setting screens.

FIG. 47 is a diagram showing a drug determination threshold setting screen.

FIG. 48 is a diagram showing an initial drug determination threshold setting screen.

FIG. 49 is a view showing a final drug determination threshold setting screen.

FIG. 50 is a diagram showing a master maintenance screen. BEST MODE FOR CARRYING OUT THE INVENTION

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

1. First Embodiment

FIG. 1 shows a medicine inspection system according to a first embodiment of the present invention, wherein 1 is an inspection device, and 2 is a personal computer.

As shown in FIG. 2, the monitoring device 1 has an inspection table 4 in a rectangular box-shaped case 3, and a camera 5 as an imaging unit above the monitoring table 4. A half mirror 6 is arranged between the audit table 4 and the camera 5 at an angle of 45 °. The half mirror 6 reflects the light from the lamp 7 of the light source installed behind the half mirror 6 downward to irradiate the medicine container 8 mounted on the inspection table 4. As shown in FIG. 3, the inspection table 4 is composed of a pair of rollers 9a and 9b which are arranged in parallel at regular intervals so as not to contact with each other and are rotatably supported. The rear end of a is connected to the motor 11 via gears 10a and 10b. Between the pair of rollers 9 a and 9 b of the inspection table 4, a medicine container 8 to be introduced from an opening 12 formed below the front of the case 3 can be mounted. The camera 5 is supported by a lift guide 14 via an arm 13 so as to be able to move up and down. The camera 5 has an interchangeable lens 15, a distance sensor 16 for measuring the distance to the medicine container 8 mounted on the inspection table 4, and a focus corresponding to the distance measured by the distance sensor 16. And a focus control device 17 for control.

As shown in FIG. 1, the personal computer 2 has a central processing unit (CPU) 18, a hard disk 19, a memory 20, and a monitor 21. Injection note data is input from the host computer 22 to the personal computer 2, and image data from the camera 5 of the inspection device 1 is input to the personal computer 2. A drive signal is output from the personal computer 2 to a motor drive circuit 23 that drives the motor 21 of the monitor 4, and the drive signal is output to a camera elevating circuit 24 for adjusting the depth of field of the power camera 5. An elevating signal is output. The hard disk 19 stores identification software, and also stores in advance characteristic images of the various medicine containers 8 in association with the names or IDs of the medicine containers, and a depth of field corresponding to the size of each medicine container 8. Is stored in association with the drug name or ID.

Next, an inspection procedure of the medicine container 8 (hereinafter, referred to as an ampoule) by the inspection system will be described.

First, the adjustment of the camera 5 will be described. Prepare a plurality of standard size ampules 8 and place them on tray 25. Mount the first one ampule 8 on the audit stand 4 and start the identification software. Subsequently, the pair of rollers 9 a and 9 b of the monitor 4 is rotated to capture an image of the ampule 8 on the rollers 9 a and 9 b, the captured image is confirmed on the monitor 21, and the ampoule 8 is mounted. Recognize the size. Adjust the position and tilt of camera 5, raise and lower camera 5, and adjust

Stop at a position where body depth is reduced. This stop position is stored in association with the mounting size. The same operation is performed for amplifiers 8 of all sizes.

Memorize feature image of sample>

The procedure for taking the characteristic images of the various ampules 8 into the hard disk 19 of the personal computer 2 will be described with reference to the flowchart of FIG. 4. In step 101, the pair of rollers 9a and 9b of the audit table 4 are rotated. When the amplifier 8 is placed on the monitor 4, the ampule 8 is imaged in step 102, this ampule image is displayed on the monitor 21 in step 103, and in step 104, the Select a certain image. In this case, as shown in Fig. 5, if the outer shape of the ampoule 8 has a characteristic, a part of the outer shape is used as a feature image and displayed on the surface of the ampoule 8 as shown in Fig. 6. Characters, for example, those that are characterized by “strong” use that character as the feature image. When a feature image is selected in this way, the feature image is captured in step 105 and associated with the name or ID of the sample 8 in step 106. Store them on hard disk 19. If there is the next amplifier 8 in step 107, the same steps are repeated, and if there is no amplifier 8, the process is terminated.

In the ampoule identification audit, first, in step 201, the injection note data is received, and the ampules 8 described in the injection note data are prepared and collected in the tray 25. Confirm that sample 8 has been prepared in step 202. With this confirmation, it is preferable to provide guidance such as "Begin." In step 203, the rollers 9a and 9b are rotated, and the amplifier 8 to be identified first is mounted on the pair of rollers 9a and 9b of the audit table 4. In step 204, the image of the mounted sample 8 is taken, and in step 205, the mounted size of the mounted sample 8 is recognized, and the depth corresponding to the mounted size is adjusted so that the subject depth is suitable for the mounted size. Stop camera 5 at the position stored in hard disk 19. In this case, when the identification monitoring is started from the small-sized ampule 8, the automatic adjustment of the camera 5 is performed smoothly, and the recognition speed is increased. In step 206, the image data of the mounted sample 8 is imported. At step 207, a pattern recognition is performed by searching whether or not a characteristic image has a force in the image data of the mounted sample 8. If there is no characteristic image, there is no amplifier 8 or the mounting position is poor. Therefore, the force for mounting the ampoule 8 is changed, and the process returns to step 204. If there is a special image, the recognition confirmation is displayed on the monitor 21 in step 208. In this case, it is preferable to perform the notification by voice at the same time. If there is the next amplifier 8 in step 209, the process returns to step 203 and the same steps are repeated. If there is no ampoule 8, record the recognition confirmation display in step 210. This record is recorded on electronic medical records, receipts, etc., and can be used as important evidence that the audit was properly performed in the event of a drug accident.

In the above embodiment, the inspection table 4 is constituted by a pair of rollers 9 a and 9 b, and the mounted medicine container ⁸ is rotated to take an image of the entire circumference of the medicine container 8. The back surface may be imaged without rotating the container 8. For example, as shown in FIG. 8, the inspection table 4 is composed of two parallel wires 26a and 26b, and the medicine container 8 mounted on the pair of wires 26a and 26b. An image of the back surface of the medicine container 8 is reflected by the mirrors 27a and 27b from the lower left and lower right of the container. By doing this, There is no need to rotate the medicine container 8, and the structure is simplified. 2. Second Embodiment

FIG. 9 shows a drug monitoring system according to a second embodiment of the present invention. In this drug monitoring system, the same portions as those in the first embodiment are denoted by the same reference numerals, and description thereof is omitted.

As shown in FIG. 10, the inspection device 31 includes a rectangular box-shaped lower case 32 a and an upper case 32 b rising upward from the rear end of the lower case 32 a and extending further forward. And a case consisting of

The second part of the upper surface of the lower casing 3 2 a, the opening ³ 3 is provided, the openings ³

A first監查table 3 4 exposed from 3, second監查stand ³⁵ located around the opening 3 3 is provided. The first monitoring stand 34 includes a pair of rollers 9a and 9b which are arranged in parallel and rotatably supported, as in the above-described embodiment. As shown in FIG. The rear end of the roller 9a is connected to the motor 11. As shown in FIG. 11, between the pair of rollers 9a and 9b of the first inspection table 34, a medicine container 8 having a circular cross section such as an ampule can be mounted. On the second inspection table 35, a drug container 8a having a non-circular cross section such as an injection kit or an infusion bag shown in FIG. 15 can be placed. The first half of the upper surface of the lower case 32b is a take-out stand 36 inclined downward and forward as shown in FIG. A chute 37 having a shallow groove is formed at the center of the take-out table 36 to prevent the medicine container 8 extruded from the second inspection table 35 by a pushing piece 46 described later from rolling.

As shown in FIG. 13, a first mirror 38 is arranged at an angle of 45 ° below the first monitoring stand 34 inside the lower case 32 a, as shown in FIG. A first camera 39, which is an imaging means, is arranged at the rear. The first mirror 38 reflects the light beam from the medicine container 8 supported by the pair of rollers 9 a and 9 b of the first

One camera leads to 39. As shown in FIG. 11, light is applied to the medicine container 8 supported by a pair of rollers 9 a and 9 b of the first monitoring table 34 adjacent to one of the first mirrors 38. A first light source 40 is disposed, and a first medicine detection device for detecting whether or not there is a medicine container 8 between a pair of rollers 9 a and 9 b of the first monitoring table 34 adjacent to the other. Outgoing sensor 41 is arranged. Further, inside the lower case 32 a, a second medicine detection sensor 42 for detecting whether or not the medicine container 8 a has a force on the second monitoring table 35 is arranged.

An elongated slide member 43 is provided on the side of the first inspection table 34 so as to be slidable on the support plate 44 in the front-rear direction of the device. On one side of the slide member 4 3 rack 4 5 is formed, the other of the pair of rollers on the side extend horizontally the first audit table ³ 4 9 a, 9 b pushing piece rises to between 4 6 Is protruding. A pinion gear 48 provided on the drive shaft of the motor 47 is engaged with the rack 45 of the slide member 43. When the motor 47 is driven by this, the slide member 43 slides in the front-rear direction. And move. When the slide member 43 moves forward, the medicine container 8 supported between the pair of rollers 9a and 9b of the first monitoring table 34 is pushed out and the chute of the discharge table 36 is pushed. It is discharged on 37.

As shown in FIG. 13, an opening 49 facing the first monitoring table 34 and the second monitoring table 35 of the lower case 32 a is formed on the lower surface of the upper case 32 b. I have. Inside the upper case 32 b, a second mirror 50 is arranged at an angle of 45 °, and behind the second mirror 50, a second camera 51, which is imaging means, is arranged. As shown in FIG. 14, the second mirror 50 reflects the light beam from the medicine container 8 a placed on the second inspection table 35 and guides the light beam to the second camera 51. A white color plate 52 having a hole 52a through which a light beam passes is provided between the second mirror 50 and the second camera 51. The color plate 52 only needs to be white on the side facing the second mirror 50. A second light source 53 for irradiating the medicine container 8a mounted on the second inspection table 35 with light is arranged adjacent to the second mirror 50. As shown in FIG. 10, a power switch 54 and a display lamp 55 are arranged on the upper inclined surface of the upper case 32b.

As shown in FIG. 12, the first light source 40 of the lower case 32b and the second light source 53 of the lower case 32a are provided with a board 57 on which a number of light emitting diodes 56 are arranged and arranged. Is housed in an elongated rectangular case 58, and an opening 59 of the case 58 is closed with a force par 60 made of ground glass so as to face the light emitting diode 56. The light from many light emitting diodes 56 can be In both cases, dimming can be performed by a control circuit (not shown). Further, the first light source 40 and the second light source 53 are held by a holding bracket 61 attached to the case 32, and the holding bracket 61 is attached to the mounting bracket 62 with screws 63. The screw hole 64 of the mounting bracket 62 through which the screw 63 passes is extended in the horizontal direction so that the irradiation position can be changed.

The monitoring procedure of a medicine container 8 (hereinafter, referred to as an ampule) having a circular cross section and a medicine container (hereinafter, referred to as an injection kit) 8a having a non-circular cross section by the medicine inspection system having the above configuration will be described. Note that the adjustment of the first camera 39 and the second camera 51 is the same as the adjustment of the camera in the first embodiment, and thus the description is omitted.

When the power switch 54 is turned on, the first light source 40 and the second light source 53 are turned on, and the first and second inspection tables 34 and 35 are irradiated. The ampoule 8 is placed between a pair of rollers 9a and 9b of the first audit table 34 as shown in FIG. A transparent drug is contained in a transparent ampoule 8 and is placed on the surface of the ampoule 8. When the drug name is displayed in black or blue, the white color of the color plate 52 is Since the image is taken as the background of the ampoule 8 via the mirror 50, the medicine name of the ampule 8 can be clearly taken.

FIG. 16 shows a procedure for loading characteristic images of various ampules 8 into the hard disk 19 of the personal computer 2. In step 100, when the first drug detection sensor 41 is turned on to detect that the ampule 8 is present in the first inspection table 34, a pair of rollers 9a of the first inspection table 34 is detected in step 101. , 9b are rotated, the ampule 8 is imaged in step 102, and this ampoule image is displayed on the monitor 21 in step 103, and in step 104-1, the first image in the sample image is displayed. Select a distinctive image and select Step 1 05— :! Captures the first feature image. Similarly, in step 104-4, a second characteristic image in the sample image is selected, and in step 1055-2, the second characteristic image is fetched. When there are a plurality of drugs having similar names, the common part is referred to as a first characteristic image, and the different part is referred to as a second characteristic image. For example, if there are two confusing drugs named “Decadon Injection 2 mg” and “Decadron Injection 8 mg”, the “Decadin Injection” part is the first feature. The image with “2 mg” and “8 mg” is the second characteristic image. Next, the first and second characteristic images are stored in the hard disk 19 in step 106 in association with the name or ID of the amplifier 8. If there is the next sample 8 in step 107, the same steps are repeated, and if there is no sample 8, the process is terminated. Since the pair of roller rollers 9a and 9b do not need to be constantly rotated, step 10

0 can be omitted.

Identification of the ampoule>

Figure 17 shows the procedure for auditing the identification of ampoule 8. First, in step 201, the injection note data is received, and the amplifier 8 described in the injection note data is prepared and collected in the tray 25. In step 202, confirm that ampoule 8 has been prepared. Based on this confirmation, it is preferable to provide guidance such as "Begin." In step 202-1, the comparison image data corresponding to the injection note data is acquired from the hard disk 19 injection. In step 202-2, it is determined whether or not the number of loops is three. If not, in step 202-3, the first drug detection sensor 41 is turned on and the first inspection table 3 is turned on. When the presence of the ampule 8 in 4 is detected, the pair of rollers 9 a and 9 b of the audit stand 34 is rotated in step 203. Since the pair of rollers 9a and 9b do not need to be constantly rotated, step 203 can be omitted. In step 204, the ampule 8 is imaged, and in step 205, the mounting size of the mounted amp 8 is recognized, and the hard disk corresponding to the mounting size is adjusted so that the depth of field is suitable for the mounted size. The first camera 39 is stopped at the position stored in 19. In this case, if the identification audit is started from the small-sized ampule 8, the automatic adjustment of the camera 39 is performed smoothly, and the recognition speed is increased. If the camera 39 uses a lens with a large depth of field, the automatic adjustment of the depth of field in step 205 can be omitted.

Next, in step 206, the image data of the mounted sample 8 is fetched. Step two

At 07-1, the first feature image is searched for the presence or absence of a force in the image data of sample 8 to recognize the pattern. If there is no first feature image, the indicator lamp 55 is lit in blue in step 207-3 because the force without the amplifier 8 and the mounting position are poor. Here, it is preferable to make a voice notification such as "unrecognizable". Step 2 0 7 One 4 drives the motor 4 7, push the ampoule 8 in the pushing piece 4 6 urged by changing the mounting position of the ampoule ^8, stearyl by adding the number of loops in Step 2 0 7 ⁵ --up 2 0 Return to 2-2. If there is a first feature image in step 207--1, in step 207-2, search is made to see if there is a second feature image in the image data of sample 8, and pattern recognition and weaving are performed. I do. If the second feature image does not exist, there is no ampoule 8 or the mounting position is bad, so go through step 2 07-3, step 2 07-4, step 2 07-5 again, and step 2 02-2. Return to

To be more specific, if the whole image of “Decadron Injection 2 mg” was used as the feature image, compared to “Decadron Injection 8 mg”, there was only a difference in the power of “2”. Rates are very low and are often misidentified. Therefore, it is necessary to increase the threshold. However, in this embodiment, “Decadron Injection 2 mg” has two feature images of “Decadron Injection” and “2 mg” as the feature images! /, So the first feature image “ The “decad mouth injection” has a collation rate of 90% or more, and the “2 mg” part of the second feature image also has a collation rate of 60%, so that it is not erroneously recognized. Therefore, the threshold can be lowered.

Next, if there are the first and second feature images in Steps 207—1 and 207—2, a recognition confirmation is displayed on the monitor 21 in Step 209, and the display lamps 5 5 Lights up in green. In this case, it is preferable to simultaneously perform notification by voice such as “recognized”. When the recognition of the amplifier 8 is completed, the image data is stored in step 208-1, and then the motor 47 is driven in step 208-2, and the amplifier ⁸ is pushed out by the extruded piece 46. If there is the next amplifier 8 in step 209, the process returns to step 202-2 and repeats the same steps. If there is no sample 8, the recognition confirmation display is recorded in step 210. This record can be found on electronic medical records, receipts, etc. and can be used as important evidence that the audit was performed properly in the event of a drug incident.

If the number of loops is three in Step 2 0 2—2, that is, if the first and second feature images cannot be collated even if the mounting position of the ampoule 8 is changed three times, Step 20 In step 2—4, prompt the operator to perform a forced audit visually. This allows the operator to visually determine whether the extruded ampoule has the first and second feature images. Press the OK button (not shown) if there is a feature image, and press the NG button (not shown) if there is no feature image. If it is determined in step 202-2-6 that the OK button or NG button has been operated, the image data is saved in step 202-6, and the process proceeds to step 209.

In the case of the injection kit 8a, it is placed on the second inspection table 35 as shown in Fig. 14 with the side with the drug name displayed as shown in Fig. 15; The procedure for capturing images and the procedure for identification and auditing of a are the same as those for ampoule 8 in the flowcharts shown in Figs. 16 and 17, except that rollers 9a and 9b are not rotated. Description is omitted.

3. Third Embodiment

Hereinafter, a medicine inspection system using an image recognition library and a mixed injection monitoring system according to a ^third embodiment of the present invention will be described.

FIG. 18 shows the master file and software stored on the hard disk 19 and FIG. 9 except that the first camera 39 is connected to the CPU 18 via the image processing board 74 b. Is the same as the system configuration of FIG. The hard disk 19 stores a master file including a drug master file, a comment master file, a usage master file, and an image master file. Further, the hard disk 19 stores a mixed injection audit system 70 and an image recognition library 71. The image recognition library 71 is positioned separately from a host application (hereinafter referred to as a host AP) 72 and a precision system (hereinafter referred to as an OS) 73 such as Windows 2000 and XP. I have. This image recognition library 71 uses a recognition driver manufactured by First Co., Ltd. The vision library 74a is software associated with the image processing board 74b. The image recognition library 71 is software that performs the following processing.

(1) Acquire an image from the imaging means.

(2) Determine the image to be monitored and the registered image by determining the image from the imaging means and the registered image.

(3) After image discrimination, the highest score and drug code To the upper AP. 3.1 Image correction

Prior to acquiring an image using a CCD camera as an imaging means, it is necessary to eliminate the three problems of barrel distortion, chromatic aberration, and vignetting. To correct barrel distortion, a lattice-shaped calibration image is taken, and the lens distortion coefficient is acquired and stored by the barrel distortion correction function of this software. After that, the geometrical correction based on the lens distortion coefficient is automatically performed on the captured image. Chromatic aberration, a red component image, a green component image, to create a blue component image, among these by creating a ^binary image by good image, eliminated. Perimeter dimming is eliminated by taking a pure white image (image for white balance), preserving the luminance distribution of this image, and adding the difference from the maximum luminance to the subsequent photographed image, to achieve uniform luminance. Image can be obtained (shading process).

The procedure of the image correction will be described based on the panel screen 75 of FIG. After startup, drop the white balance image on the left side of the panel screen 75, and select “White balance” 76 6 “Set current image on whiteboard” in the menu. Drop the captured image on the left side of the panel screen, and operate the slider control (distortion coefficient A, distortion coefficient B, focal length) on the right side to adjust the image to a distortion-free image. For the captured image corrected for the barrel shape, “Execute brightness correction with white port” of “White Balance” 76 of the menu is selected. As a result, an image for collation is obtained. For this image for collation, select “Extract Reference Image” from the menu “Create DB” 7 8, enclose the required reference characters with the mouse, and press the ΞFemale key. As a result, the part of the area surrounded by the mouse is displayed on the right side interruption. Then, select “Create reference image” of “Create DB” 78 in the menu. 3.2 Image registration

FIG. 20 shows a flow of registering a medicine image by the image recognition library 71. First, the whole image of the medicine taken in step 301 is taken in, and in step 302 a medicine code is obtained. In step 303, a feature image is cut out of the entire image, and in step 304, image data of the feature portion is created in a bitmap format. The image data is registered in 2005. In step 310, it is determined whether the number of registrations per drug is 4 or less. If the number of registrations is less than 4 and the registration is not completed, the process returns to step 303 to store the next feature image. If the number of registrations is less than 4 and the registration is to be terminated, the process ends at that point. If the number of registrations exceeds 4, no more registrations can be made and the process ends.

The procedure for registering a drug image will be described more specifically. First, a medicine to be registered in the image master is photographed. The photographing of the medicine is performed using the inspection devices 1 and 31 described above. The captured image of the drug is displayed on the panel screen 75 as shown in FIG. FIG. 19 is an image of the saline solution kit H100 mL.

One or more features are extracted from the captured image and selected. In this example, as shown by the dashed lines, the characteristic part A of the “saline solution kit H”, the characteristic part B of “100”, and the characteristic part C of “Nipro Pharma Co., Ltd.” are selected. . As shown in Fig. 21A, each characteristic part A, B, C is taken in association with the drug ID of the drug master, and given the file names of 2pro.bmp, 100.bmp, and saline solution.bmp, respectively. I do.

The image of the first characteristic part A is binarized in black and white, and filtered with a 20% image adjustment threshold (hereinafter simply referred to as threshold) and displayed. As shown in Fig. 22, this image is displayed with a lot of black noise. The threshold is changed to 30%, 40%, 60%, and 80% with the threshold adjustment bar 79 on the panel screen 75 of FIG. When the threshold is 30%, the recognition level as an image is improved, but black noise still remains. When the threshold value is 40%, the recognition level is remarkably improved, and the recognition level can be sufficiently used. At the thresholds of 60% and 80%, part of the image is missing and characters are difficult to see. If the threshold value is exceeded, the black portion decreases, and it cannot be recognized as a character at all. Therefore, the threshold is set to 40%.

Similarly, the image of the characteristic portion B is displayed, and the threshold is adjusted to 35% by the threshold adjustment bar 79,

As you change to 45%, 60%, and 90%, the image changes as shown in Figure 23. Set the highest recognition level of 45% as the threshold. Finally, the image of the characteristic part C is displayed, and the threshold is adjusted to 20%, 25%, 40%, 60%, 9 As you change to 0%, the image changes as shown in Figure 24. The threshold with the highest recognition level of 40% is set. After setting the threshold, the images of the characteristic parts A, B, and C are registered together with the threshold and the drug name, as shown in Fig. 21B. FIG. 25 shows a specific data structure 80 of the stored image. Product comparison data structure ⁸¹ includes a comparison agent number 82, consisting of comparing the master first image storing Adoresu ⁸ 3. The comparison master image storage address 83 has a comparison master image storage address from drug A to drug N. When each comparison master image storage address is specified, the comparison master image structure 84 is called. The comparison master one image structure 84 contains the drug code 85, the number of master image data 86, the set threshold values 8 to 7 for the master 1 to 4 images, and the storage addresses 8 to 8 for the master 1 to 4 images. Have. When each storage address 8 8 of the master 1 image is specified, each entity 8 9 of the 4 images from the master 1 image is called.

FIG. 26 shows an example in which the extracted image of the characteristic portion is further divided into a plurality of regions and stored. The feature image is divided at a fixed ratio in the vertical and horizontal directions according to the size and size. A black dividing line is displayed so that the divided area can be identified, but this dividing line is not necessarily required. The number of divisions is preferably set in advance in the vertical and horizontal directions. For example, according to the length of one side being 20, 50, 80, and 12 Omni, the division is made into two divisions, three divisions, four divisions, and five divisions. If the cut out feature is 18 mm long and 76 mm wide, divide it into two by 9 mm vertically and four by 18.5 mm horizontally. The number of divisions is preferably two to eight. When the number of divisions increases, the accuracy of pass / fail judgment improves, but the processing speed decreases. Figure 26 shows a feature image called “UNICALIC L” divided into eight regions by dividing the vertical into two parts and the horizontal into four parts. It is preferable to calculate and store the number of white pixels and black pixels for each divided region. The divided regions in FIG. 16 are equally spaced both vertically and horizontally, but may be unevenly spaced. The shape of the divided area is not limited to a rectangle, but may be a continuous triangle, hexagon, or circle. Furthermore, the collation at the time of monitoring may be performed only within the continuous circular divided area, instead of the entire image of the extracted characteristic portion.

3.3 Drug audit FIG. 27 shows a flow of drug inspection by the image recognition library 71. In step 311, initialize the image processing board 7 4b, and in step 312

Start '. In step 3 13, a registered image corresponding to the prescription data is acquired.If image acquisition is successful in ³ 14, it is determined in step ³ 15 whether or not the number of medicines is 20 or less. And an error. Agent number 2 0 less der lever, Step ³ 1 6 with chemicals code, obtains the threshold information and the like, performs the images determined in step ³ 1 ⁷ and ends.

FIG. 28 shows a flow of the image judgment of the audit flow of FIG. 27 described above. In this flow, in step 3 21, the upper tier ALL calls “prescription data judgment start”, and here acquires the registered images of all the medicines in the prescription data. In step 3 2 2, the upper AP “Prescription data judgment result acquisition” is called, an image of the medicine to be inspected is acquired here, and the acquired image is matched with the registered image. The degree of matching is scored, and the highest score and drug code are notified to the upper AP. In step 3 23, it is determined whether or not the determination has been completed for all the drugs, and if not, the step processing is performed for all the drugs. When the judgment is completed for all the drugs, the upper AP calls “prescription data judgment end” in step 3 2 4 and ends. FIG. 29 shows the flow of the pattern search process in the matching of “prescription data judgment result acquisition” in FIG. In this flow, the grayscale image memory area of the acquired image is searched to find the coordinate position where the same pattern as the stored image data pattern exists. As shown in FIG. 30, the search direction moves from the upper left corner to the right edge by a predetermined pixel, returns to the left edge below the predetermined pixel, and repeats in the same manner. FIG. 30 shows that the same pattern as the stored image with blood was present at the B1 position and the B2 position, respectively. In order to improve the image coincidence, as shown in FIG. 31, for example, a search can be performed by tilting the stored image at a pitch of 1 degree within a range of 3 degrees. FIG. 31 shows that the same pattern exists in the stored images at the B3 position and the B4 position. The search may be completed with a full search, or may be terminated when the same pattern is found. As shown in Fig. 32, the algorithm of the search process is the matching of the pattern image P (M, N) to be searched and the partial image Β (Μ, Μ) of the image B (X, Y) in the search area. Ne phase showing degree The relation number C ij is obtained by the following equation, and the coordinate value at which the value becomes larger than a predetermined threshold value is obtained.

First, in step 331 in Fig. 29, the first feature image is obtained, and in step 332, an approximate correlation calculation is performed, and an image similar to the stored image is searched for based on the initial judgment threshold, and Set as candidate points (candidate point search). In step 33, a correlation calculation is performed on the candidate points (final search). In step 33, it is determined whether or not the same pattern as the stored image has been found with a final determination threshold value larger than the initial determination threshold value. . If you do not find the same pattern, return to step 3 and repeat the procedure. If it is determined in step 3 3 4 that the same pattern has been found, it is determined in step 3 3 5 whether or not all feature images have been searched. If all feature images have not been searched, step 3 3 Return to 2 and repeat the same procedure to end. In this way, since the search is performed in two stages, the candidate point search and the final search, the search speed and reliability are ensured.

In order to improve the speed of the candidate point search, a simple stored image can be used. The simple stored image is an image reconstructed by extracting 50% of the pixels from all the pixels of the stored image. For example, from the stored image (200 pixels / let) of the character "y" shown in Fig. 33, 50% of the pixels are alternately extracted for each pitch, and the determination of the 100 pixels shown in Fig. 34 is made. A simple stored image having an invalid part “one” is reconstructed. When the simple memory image in Fig. 34 and the real image pattern shown in Fig. 35A are compared at a position shifted two pixels to the left, as shown in Fig. 35B, out of 100 pixels, NG "X" Are 28, the black OK “Hata” and the white OK “〇” are 35 and 37, respectively, for a total of 72, with a match of 72%. If the threshold in the search for a catch point is 55%, it can be determined that the pattern is the same even if it is shifted two pixels to the left from the real image pattern. In addition, the pattern of the simple storage image shown in FIG. 34 and the solid black real image shown in FIG. Matching the patterns, the NG out of 100 pixels is shown in Figure 26B.

"X" is 51, black OK "Hata J is 49, and the degree of coincidence is 49%. Similarly, comparing the simple stored image with the white real image pattern, The number of NG "X" is 49, and the number of black OK "bon" is 51, and the matching degree is 51%. Therefore, even if the ^{5 5%} threshold of the candidate point search, it is possible to determine the pattern of solid black and Shiro抜Kino real image to be mismatched. In the final search, all the pixels of the stored image shown in FIG. 33 are used instead of the simple stored image of FIG. If, as shown in FIG. ^{3 7} A, as chipping P image was present in the real image pattern, the result of the collation, as shown in FIG. 37 B, black "OK" 98/101, white "ΟΚ" is 99 99, about 97% for black and 100 for white. /. Thus, the match can be determined even if the final search threshold is 93%.

As a modified example of the embodiment, as described above, the image stored in the image memory can be divided into a plurality of regions, and the degree of coincidence between the image in each region and the real image image can be determined. In this case, in each region, pixel-by-pixel matching is performed as shown in Figure 37 37.

"NG" force Pass / fail of "OK" is determined. By calculating the disparity in the degree of coincidence between the images of the respective areas, and comparing the disparity in the degree of coincidence with a predetermined disparity threshold, the determination of “NG” power or “OKj” is performed. When the degree of coincidence LA between the real image and the image in the area _A is 95% and the degree of coincidence LB between the real image and the image in the area B is 60%, the threshold value is 55% or more. Also, the difference between the two, L _A — LB (95-60 = 35%), is very large, and is larger than a predetermined gap threshold (for example, 10%). Conversely, if the disparity is small and smaller than a predetermined disparity threshold, it is determined to be “OK”. This modification is particularly effective when discriminating images that are partially different because partial differences between images, such as “8” and “3”, appear as differences in the degree of matching between images in each region. The judgment accuracy is improved.

3.4 Co-injection audit system

Next, a screen operation procedure of the medicine monitoring system by the personal computer 2 will be described.

The desktop shown in FIG. 18 displayed on the monitor 21 of the personal computer 2 Clicking on the lock icon 81 will display the mixed injection system menu screen 82 shown in Figure 39. On this menu screen 82, each button of the mixed injection monitoring monitor 83, the Ph threshold value setting 84, the printer setting 85, the environment setting 86, and the operation end 87 is displayed. Here, the operation procedure when the button of the mixed injection monitoring monitor 83 and the environment setting 86 is clicked is explained, and the Ph threshold setting 84 and the printer setting 85 are not directly related to the present invention, so they are explained. Is omitted.

When you click the button of the mixed injection audit monitor 83 on the menu screen 82, the login screen is displayed. Enter the user ID and password and press the OK button. As a result, the mixed injection audit monitor screen 88 shown in FIG. 40 is displayed. To end the audit work, press the work end button 87 on the menu screen 82 of Fig. 39.

Mixed injection監查monitor screen 8 8 patient information field 8 9, mixed injection information column 9 0, detailed information column ⁹ 1, message temporary檷9 2,監查mode column 9 3, the communication information field 9 4, function keys It consists of nine and five.

As shown in Fig. 41, the patient information column 89 is composed of a ward code, a department code, a patient ID, a patient name, age, gender, height, weight, body surface area, special instructions, a room number, and a doctor name. I have. Enter the ward code and patient ID to display other patient information. If the patient is a pregnant woman, the special instructions will read “Pregnant woman”.

As shown in FIG. 42, the co-injection information column 90 is composed of an implementation date, a prescription category, a completion, an application unit, and a drug recipe (Recipe) number (R p). Select the implementation date to be monitored from the pull-down menu of the implementation date, and select the prescription category to be monitored from the pull-down menu of the prescription category, and the audit information of the audit subject on the implementation date is displayed in a list. In the completion column, if the audit is not completed, 〇 will be displayed, and if completed, the hit will be displayed. The application unit indicates the order of implementation. If the same recipe has a different recipe number, it is displayed on a different line. The drug recipe number is the recipe number of the co-administered drug. When you select the line to be monitored, the details are displayed in the details information.

As shown in FIG. 43, the detailed information 91 includes a drug name, a dose, a unit, a procedure, a Ph value, a co-injection group, and the number of completed Z tubes. The last line of the specification information shows the total dose. The order of the statement display can be selected from the pull-down menu in the statement display column. "Drug name" is the full name of the co-administered drug Is done. “Dose” is the amount of co-administered drug and “unit” is the unit of the amount. “Technique” indicates a method such as infusion. “Mixed injection group” indicates the group of drugs to be audited.

In “Completed / Number”, “Number” indicates the number of medicines to be audited, and “Completed” indicates the number of completed audits. If there is prescription detail additional information in the bottom column of the detail information, “there is a description item” is displayed.

Message 檷 92 displays the operation guide ゃ error contents.

The audit mode column 93 includes buttons for selecting an automatic or manual audit mode, and the button displays the selected audit mode. Press the button labeled Manual to switch the audit mode to automatic, and press the button labeled Automatic to switch to the manual mode.

In the communication information brochure 94, a communication state of normality> no is displayed.

On the mixed injection audit monitor screen 88, when starting the audit, if the monitoring mode is manual, select the number cell of the row to be audited from the list of detailed information. When the audit mode is automatic, audits are monitored in order, so there is no need to select the target rows.

The medicine is placed on the inspection table of the inspection devices 1 and 31. If the result of the audit is NG, “NG Gj” is displayed in the detailed information display area as shown in Fig. 44.

“OK” is displayed. The display of the results of the audit is not limited to characters, and the lines may be highlighted in an appropriate color. For example, if the result of the audit is NG, highlight the row in red; if it is Ο, highlight the row in blue.

The co-injection group can be divided and released if there is a problem such as a change in Ph or a loss of medicinal effect when processed in the same group. To divide the co-injection group, select the cell of the drug name in the row of the drug displayed in the detailed information No. 91 and press F4 of the function key 95. To release the co-injection group, select the drug name cell in the row for the drug displayed in the detailed information column 91 and press F5 of the function key. Select the group name before splitting. To perform a compulsory audit of a drug, select the drug name cell in the row for the drug displayed in the detailed information column 91, and press F1 of the function key 95. At this time, the row of the compulsory audited drug is displayed in green. When “Number of completed” is 1-3, In other words, when one of the three drugs is automatically audited and the result is "OK", F1 can be pressed and the remaining two can be visually inspected. At this time, every time F 1 is pressed, 1 is added to “Complete” of “Complete Ζ Number”.

The results of the above-mentioned audit, that is, the results of the pass / fail of the mandatory audit and image audit, and the captured images are recorded as an audit history in a time series on the host computer 22 so that they can be displayed or printed as appropriate. I'm familiar. In this way, if a medical accident or the like occurs, the audit history of the infusion co-administration work is investigated to clarify under what circumstances the accident occurred, or not just the accident but the sudden change in the patient's condition. Become.

When a drug is bar-coded with an infusion bag, monitoring may be performed by reading the bar-code. In this case, as shown in Fig. 45, it is preferable to display a message such as "Read the barcode of the infusion solution to be audited". Clicking on the button of the menu screen 82 preferences 8 6 3 9, various environmental settings screen ₉ 6 shown in FIG. 4 6 is displayed. On the various environment setting screens 96, the buttons for drug judgment threshold setting 97 and master maintenance 98 are displayed.

Click the button of drug judgment threshold 9 7 on various environment setting screen 9 6

47 The drug determination threshold setting screen 99 shown in 7 is displayed. On the medicine determination threshold setting screen 9 9, buttons for initial medicine determination threshold setting 100 and final medicine determination threshold setting 101 are displayed. Clicking the button for initial drug judgment threshold 100 Clicking the button of 100 sets the initial drug judgment threshold screen 102 shown in Fig. 48. Change the initial drug judgment threshold with the up / down key and use the setting button to set it. Press, and press the Back button if you do not want to set. Clicking the button for final drug determination threshold 1 0 1 will display the final drug determination threshold setting screen 103 in Fig. 49, and use the up / down key to change the final drug determination threshold and set it to Press the setting button, and if not, press the back button.

Clicking the button of the master maintenance 98 of the various environment setting screens 96 will display the master maintenance screen 104 of Fig. 50. The master maintenance screen 104 displays the following buttons: medicine master 105, comment master 106, usage master 106, and image master 107. If you click the button of the drug master 105, a drug master maintenance screen (not shown) is displayed, and the drug master file is displayed. You can add or change the name of the drug, drug ID, dose, contraindicated information, etc., stored in. Clicking the button of the comment master 106 displays a comment master maintenance screen (not shown), in which comments such as the infusion method, precautions, and side effects stored in the comment master file can be added or changed. When the button of the usage master 107 is clicked, a usage master maintenance screen (not shown) is displayed, and it is possible to add or change the usage of the drug stored in the usage master file, such as a usage amount, a limit amount, and a caution. By clicking the button of the image master 108, the image master maintenance screen is displayed, and it is possible to add / change the image data of the medicine stored in the image master file.

Claims

The scope of the claims

1. A storage means for storing images of a large number of drug containers in advance in association with their drug names,

An audit table equipped with drug containers to be audited,

Imaging means for capturing an image of a medicine container mounted on the inspection table,

A medicine inspecting device comprising: a judging means for judging whether or not the image obtained by the imaging means includes an image stored in the storage means.

2. The medicine inspection device according to claim 1, wherein the storage means stores an image of a characteristic portion of the medicine container.

3. The storage means stores images of a plurality of characteristic portions of the medicine container,

The apparatus according to claim 1, wherein the determination unit determines whether or not the image obtained by the imaging unit has any of the plurality of characteristic portion images stored in the storage unit.

4. The discriminating means, when discriminating whether or not any image of the characteristic part stored in the storage means is included in the image obtained by the imaging means, the image of the characteristic part 4. The drug monitoring device according to claim 3, wherein the determination is performed using a different threshold value for each.

5. The medicine inspection device according to claim 1, wherein the inspection table includes means for changing a posture of the medicine container.

6. The medicine inspection device according to claim 5, wherein the means for changing the posture of the medicine container is a pair of rollers for supporting and rotating the medicine container.

7. A light source for irradiating the medicine container with light is further provided.

The light source is dimmable and has frosted glass to make the brightness uniform throughout. 2. The drug monitoring device according to claim 1, wherein:

8. The medicine inspection device according to claim 1, wherein a white color plate is provided behind the medicine container so that the image of the medicine container has a white background.

9. The medicine inspection apparatus according to claim 1, wherein the imaging means takes an image of a plane of the medicine container and an image of another surface.

10. The image of the other surface of the medicine container is a back surface image of the medicine container, and the imaging means images a back surface image reflected in an imaging direction by a mirror body. Drug inspection device.

11. An image acquisition unit for acquiring an image of the medicine container stored in the storage unit from the imaging unit,

The image obtaining means cuts out an image of a characteristic portion selected from the obtained image of the medicine container, and filters the image of the cut out characteristic portion with a predetermined image adjustment threshold value. 2. The medicine inspection device according to claim 1, wherein the image data of the characteristic portion obtained is stored in the storage unit in association with the medicine name together with the image adjustment threshold value.

12. The medicine inspection device according to claim 11, wherein the image picked up by the image pickup means is filtered with the same threshold value as the predetermined image adjustment threshold value.

13. The medicine inspection device according to claim 11, wherein the image adjustment threshold is changeable.

14. The drug monitoring device according to any one of claims 11 to 13, wherein the determination unit tilts the stored image by a predetermined angle and compares the tilted storage image with the stored image.

15. The discriminating means performs an approximate correlation calculation to search for a plurality of candidate points similar to the stored image, and performs a correlation calculation on the candidate points to narrow down images matching the stored image. The method according to any one of claims 11 to 14, wherein a final search is performed.

16. The medicine inspection device according to claim 15, wherein a threshold value of the candidate point search is smaller than a threshold value of the final search.

17. The request according to claim 15, wherein the discriminating means uses a simplified storage image reconstructed by removing pixels from the storage image at a fixed pitch when performing the candidate point search. Drug inspection equipment.

18. The storage means divides a large number of medicine container images into a plurality of areas in advance and stores them in association with the medicine names,

The determination unit determines a degree of coincidence between the image obtained by the imaging unit and an image of each area stored in the storage unit, calculates a difference in the degree of coincidence of the image of each area, The medicine inspection device according to claim 1, wherein the difference is compared with a predetermined difference threshold to determine whether or not the image obtained by the imaging unit includes an image stored in the storage unit. .

19. The drug monitoring device according to any one of claims 1 to 18, further comprising display means for displaying a history of the determination by the determination means.