JP2787151B2 - Article quantity inspection method and device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[Industrial applications]

[0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an article inspection method and apparatus for inspecting a group of articles to determine whether the type and quantity are in accordance with a desired number of members. For example, when a predetermined article is stored in a transport container such as a bucket and handled, the present invention relates to a method suitable for inspecting whether or not a predetermined number of articles are stored in the transport container.

[0003]

2. Description of the Related Art For example, in the field of logistics, in order to carry out efficient delivery, various products produced in factories are once collected at a delivery center with good transportation, and the delivery request from many dealers is accepted. The form that responds is increasing.

[0004] In the field of manufacturing, the mode of production is shifting from conventional low-mix high-volume production to high-mix variable-volume production, and many different products and materials must be handled in the same factory. I have.

[0005] In such a case, it is important to quickly identify many varieties and check that a desired quantity exists without excess or shortage.

In the field of physical distribution, a person takes out (in the following, referred to as picking) articles of a type and quantity according to a request from a store or the like from an accumulation place, packs them, and ships them. In this case, it is indispensable not only to improve the efficiency of picking and packing but also to provide services to the dealers, etc., in order to strictly satisfy the requirements from the dealers.

[0007] On the other hand, in the manufacturing industry, erroneous types and quantities of necessary articles not only hinder the work in the downstream process, but also lower the quality.

Therefore, conventionally, a desired article has a desired amount,
In order to check whether or not it is present in a specified container, a bar code or the like specific to the article is attached, and a method of optically reading and checking the information, and registering in advance with image information taken by a TV camera For example, a method of comparing information described in Japanese Patent Application Laid-Open Nos. Sho 62-52694 and 63
-123660, JP-A-3-290781, etc.).

[0009]

However, according to the identification method using a bar code or the like, there is a problem that a predetermined bar code or the like must be attached or printed on all the articles. Also,
There have been problems such as the necessity of strict positioning for reading the data and the limitation on the size and shape of the article.

On the other hand, according to the method of identifying articles based on image information and weighing the number of articles, a bucket on a transport line is photographed from directly above by a television camera, and the image information is processed and registered in advance. The identification is performed by comparing with the image information of the article in question. However, this has the following problem.

(1) As the number of articles to be inspected increases, the processing time increases, and it cannot be used in real time.

(2) A large storage device is required to pre-register the image information (pattern) of each article.

(3) The procedure for registering a new article is complicated and requires many man-hours.

(4) Variation in subject depth, that is,
Inspection accuracy varies due to variations in the distance from the camera to the article.

That is, in the conventional method of identifying the type of an article by pattern matching, the outer peripheral portion of the article to be identified is cut out by the density differential processing (the degree of change in the density) of each pixel, and then the article is considered. For all areas, the type of article is identified by comparing it with the registered pattern. In the process of performing such an identification process, in order to compare the article image with the registered pattern, it is necessary to perform an enlargement / reduction process to match the size, or to perform a process such as rotation to adjust the position. .

Therefore, as the number of target articles increases, the processing time increases. Also, if there are many articles in one screen, the number of pixels for one article is reduced, so that the resolution is reduced and the recognition accuracy is reduced. Further, when the shapes of the articles are various, the distance between the articles and the camera is not uniform, and the lens focus cannot be optimally set for all the articles, so that the accuracy of recognizing fine characters and figures decreases. .

For this reason, instead of the method of identifying an article based on image information as described above, even if the number of articles is large or the attitude of the article is not constant, the method is simple, high-speed, and high. Method and apparatus capable of identifying an article type with high accuracy, counting an article quantity based on the identification result, and counting and inspecting the quantity of each type for a plurality of kinds of articles. The realization of has been desired.

A first object of the present invention is to solve the above-mentioned problems and to perform a simple, high-speed, and high-precision inspection of whether or not the type and quantity of a group of articles are of a desired number. An object of the present invention is to provide a method and apparatus for inspecting an article.

It is a second object of the present invention to provide a method for identifying an article which can be easily, quickly and accurately identified.

Further, a third object of the present invention is to provide a simple,
An object of the present invention is to provide an article counting method capable of counting the number of identical articles at high speed and with high accuracy.

[0021]

The present invention is characterized by the following means in order to achieve the above object.

In order to achieve the first object, the present invention provides a method for determining whether or not the quantity of each type of article has a predetermined number of members based on image information obtained by photographing a plurality of articles of a plurality of types to be inspected. in quantity inspection method of an article of checks, may be registered goods color information consisting of the area of the characteristic color and the characteristic color of the article surface for each type of pre-inspected and ing article, the inspected article When a plurality of types and the number of members for each type are given, based on the registered item color information, the area of the registered characteristic color for the item is multiplied by the number of members for each type of item, and the registered characteristic color is registered. The reference color area is determined, the color of each unit area of the image is analyzed based on image information obtained by photographing a plurality of types of articles to be inspected, and the registered color that matches the analysis color within a certain allowable range. The actual measured color area for each characteristic color is integrated, and Based integrated value of the measured color area of each recording characteristic color and the degree of coincidence between the reference color area, and judging whether there membered appointment quantities for each type of article.

In this case, it is preferable to set at least two pieces of article color information for each article. That is, if there is an article having the same characteristic color using only one characteristic color, it cannot be distinguished.

In order to represent the characteristic color, three color components of light (for example, red, green, and
Blue) and their brightness.

That is, a unit area to be subjected to color analysis (for example,
For example, the color of each pixel is separated into R (red), G (green), and B (blue), and each brightness gradation is calculated and normalized to relatively represent the gradation. It is represented by a combination of lightness r, g, b.

When there are two or more characteristic colors that match the analysis color of each unit area of the image within a certain allowable range, it is impossible to determine which of the articles is the object color. It is preferable to avoid the judgment by not adding.

When the article color information is set to two or more for each article, the quantity for each article type is scheduled based on the highest degree of coincidence between the integrated value of the actually measured color area and the reference area. It is determined whether the number of members exists.

The apparatus for carrying out the article quantity inspection method of the present invention includes a television camera for photographing a plurality of kinds of articles to be inspected from a predetermined direction, and information of the photographed image. and effective area cutout means cuts out an effective area that is present, double of the article to be present in the effective area
A storage means for storing inspection information of the number of types and the number of members for each type, and article color information including the characteristic color of the article surface and the area of the characteristic color are registered in advance for each type of article to be inspected. The reference color area of the registered characteristic color is obtained by multiplying the area of the registered characteristic color of the article by the number of members for each type of the article based on the article color information storage unit and the article color information and the inspection information. An inspection reference color information generation unit, an image for analyzing a color of each unit area of an image existing in the effective area, and integrating an actually measured color area for each of the registered characteristic colors that matches the analysis color within a certain allowable range; Inspection for determining whether or not the quantity of each type of article has a predetermined number of members based on the color information classifying means and the degree of coincidence between the reference color area and the integrated value of the actually measured color area for each of the registered characteristic colors. Can be realized by providing .

In this case, when the degree of coincidence is higher than a predetermined value, the inspection judging means divides the integrated value of the actually measured color area by the area per article of the corresponding characteristic color, and The quantity is calculated, and if the calculated quantity matches the number of corresponding articles, the result is determined to be acceptable, and if not, the excess or deficiency is calculated.

[0030] The article color information registered in the article color information storage means is based on image information obtained by photographing various articles to be inspected for each article in advance. This can be realized by learning and registering means for article color information, which finds an area of an image that matches within a certain allowable range and obtains a characteristic color and an area of the characteristic color for each type of article.

Further, in order to achieve the second object of the present invention, an article identification method according to the present invention for specifying the type of an article based on image information obtained by photographing one article to be identified, comprises: A color representing each characteristic is set for each type, and identification color information including the set color and the area of the color is set in advance, and a color matching the color of the identification color information and The area is determined, and the type of the article corresponding to the identification color information whose area matches the determined color within a certain allowable range is specified as the type of the article.

Further, in order to achieve the third object of the present invention, the article counting method of the present invention for obtaining the quantity of the articles based on image information obtained by photographing a plurality of the same articles to be counted, comprises: Color information consisting of a color representing the characteristic of the color information and an area of the color is set, an area of a color corresponding to the color of the color information is obtained based on the image information, and the obtained area is compared with the area of the color information The quantity of the article is obtained based on

[0033]

With the above-described means, the above object can be achieved by the following functions.

First, the principle of the present invention will be described. The conventional pattern matching method has extremely high matching or discrimination accuracy under ideal conditions, but it has been found that the problem to be solved by the present invention is excessive in information and therefore has the above-mentioned problem.

That is, considering the method of identifying the type of an article by visual observation, small letters and complicated shapes are auxiliary means, and accurate recognition can be performed even if the focus is shifted or the attitude of the article is random. It is considered that the reason for this is that the judgment is made based on macro information such as large characters, numbers, characteristic color arrangements, and areas of the colors. In other words, when analyzing the visual observation of the identification of the article type, by focusing on the schematic shape, characteristic figure, and color of the target article,
It turned out to be making fast and accurate decisions. In particular, the same applies to the case where the article is packed, and it has been found that characters and small figures on the surface of the packing are often referred to only in an auxiliary manner.

Therefore, the present invention focuses not only on the characteristic pattern of the article or the packaging surface of the article, but also on the difference between the color representing the characteristic of the article or the packaging surface of the article and the area occupied by the color. They tried to identify the articles by using them as article color information unique to each article. In particular, when an article is packed, it is effective because labels printed with various colors are attached to the packing. That is, with the development of printing technology, the reproducibility of the same color and the accuracy of the printing position have become extremely high, and the color pattern of the label has also become colorful in order to enhance the discriminability of the product (article).

As is well known, the color has three colors, and all the light C can be represented by the following equation: C = r.R + g.G + b.B. Here, R is red, G
Indicates the original stimuli of mutually independent colors of green and B, and r, g and b indicate the amounts of brightness (brightness) of the respective colors. R, G, and B can be replaced with other color sets as long as they are independent, but in the present description, R, G, and B will be described below. Even if the brightness changes, the ratio of r, g, b does not change. Therefore, by normalizing so that the sum of r, g, b always becomes 1.0, it can be adopted as color feature information. There are several other methods of expressing colors, but since they can be handled by an equivalent set, they will be described below in the RGB color system.

Assuming that the brightness gradation of each RGB of each pixel is 256, there are at least 256 cubes, that is, about 16.7 million colors.

Therefore, by assigning a characteristic color that does not systematically become the same color to each article or a label attached to the article, it becomes possible to identify the type of the article. Here, in the case where the content of the inspection is to identify the type of one article, or in the case of identifying a small number of articles, only one characteristic color is often required. The same applies to the case where the quantity of the same article is counted. However, when the types of articles that need to be identified at a time increase, various sizes, shapes, and colors are used depending on the labels of the articles. There is a high possibility that the same or similar article will occur, and high identification accuracy cannot be obtained if only one characteristic color is used.

Therefore, not only the color of the article, such as a label attached to the article, but also the difference in the area occupied by the color and the article color information obtained by combining the colors are considered. It is used to represent the characteristics of.

Specifically, as described above, two or three types of characteristic colors are set in advance for one type of article, and the normalized lightness r, g, and color components R, G, and B of each characteristic color are set. It is desirable that b and its area are analyzed and stored in advance, and the type of the article is identified by a combination of the characteristic colors. That is, at the time of actual inspection, since a plurality of types of articles are present on the same screen, a common characteristic color exists between different types of articles regardless of the uniqueness of the set characteristic colors. Things can happen. In this case, the recognition accuracy can be maintained at a high level by excluding both corresponding characteristic colors from the inspection information and identifying them with the remaining characteristic colors.

In this manner, the characteristic color of an image and its actual color area are determined based on the image information obtained by photographing a plurality of articles to be inspected, and the actual color area is calculated as the area per article of the corresponding characteristic color. (Here, the number of pixels) is divided, and the result is estimated as the number of existing articles.

By comparing the estimated value obtained in this way with a predetermined number of members, it is possible to determine whether the type of the article and the number of members match or not.

Note that the image information obtained by photographing the target article has uncertain factors such as, for example, variations in illumination, variations in the storage state and posture of the articles, and deviations in focus. Although such an uncertain element has a fatal effect on the recognition accuracy according to the conventional pattern matching method, in the present invention, a value of a matching degree (hereinafter referred to as a threshold) relating to the pass / fail judgment of each information is given. Can be avoided by adjusting the value.

The determination of the boundary region in the comparison of each characteristic color and area can be avoided by applying fuzzy inference.

Further, according to the present invention, the inspection can be completed in an extremely short time because the processing only needs to be performed once for each pixel in the effective area.

In the above, an example has been described in which the type and the number of members of a plurality of types of articles are inspected at the same time. However, if the principle of identifying the articles based on the color information of the present invention is used, it is possible to identify the type of one article. It is needless to say that what can be done, and counting the number of articles of the same type can likewise be understood.

[0048]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a block diagram showing a functional configuration of an embodiment of an article inspection apparatus according to the present invention. This embodiment is based on image information obtained by photographing an article 13 stored in a bucket 12 being conveyed on a conveyor 11 in a vertically downward direction with a television camera 14 installed above. This is an example in which it is checked whether the type and quantity are stored in a predetermined number of members, and the result is output and displayed on the display device 15.

The article inspection apparatus according to the present embodiment includes an effective area cutout unit 2, an inspection reference color information generation unit 3, an image color information classification unit 4, an inspection determination unit 5, a result display processing unit 6, A memory 7, an article color information learning registration unit 8,
And an inspection information generating means 9.

The memory 7 stores an article color information table in which a characteristic color of the article surface and an area occupied by the characteristic color are registered for each of various articles to be inspected. In addition, the memory 7 stores an inspection information table that defines the types of articles serving as inspection standards and the number of articles.

The effective area extracting means 2 receives image information from the television camera 14 as input, and extracts an effective area corresponding to the inside of the bucket 12 from the photographed image.
The cut-out image information is output to the inspection reference color information generating means 3.

The inspection reference color information generating means 3 uses the contents of the inspection information table and the article color information table stored in the memory 7 to register the registered characteristic colors of the images of all the articles to be inspected, An inspection reference color information table is created by calculating the reference color area of the registered characteristic color, and this is stored in the memory 7.

The image color information classifying means 4 performs color analysis of all the pixels present in the effective area, and if any of the analyzed colors corresponds to a registered characteristic color in the inspection reference color information table, the registered color is registered. The integration is performed as the measured color area of the characteristic color.

The inspection judging means 5 comprises the image color information classifying means 4
Then, the measured color area of each registered characteristic color obtained by the above is compared with the corresponding reference color area, and the reliability of the system including the calculation processing of the measured color area is estimated. If the estimated reliability is better than a predetermined reference value, the actually measured color area for each registered feature color is divided by the reference color area, and the quotient obtained by this is applied to each registered feature color. Estimate the number of items present. Next, the estimated number of existing persons is compared with the number of scheduled members, and when the number of members matches the number of planned members, the result is judged to be acceptable.

The result display processing means 6 is adapted to output and display the inspection result output from the inspection determining means 5 on the display device 15.

The article color information learning and registering means 8 creates article color information for each article stored in the memory 7 prior to each of the above-described processes. That is, the sample 16 of the inspection target article is photographed in advance by the same television camera 17 as the television camera 14, the specified characteristic color is extracted from the image information, the area occupied by the characteristic color is obtained, and the article color information is created. Then, the result is stored in the article color information table in the memory 7.

The inspection information generating means 9 creates inspection information for each inspection target bucket 12 based on information given from the upper level production management computer 10 and stores the inspection information in the memory 7 in the table. ing.

FIG. 2 shows an example of a storage state in the bucket 12 in which a plurality of articles 13 to be inspected are stored. FIG. 3A is a view from the direction of the television camera 14, and FIG. 3B is a side sectional view. In the illustrated example, each article 13 has a long rectangular shape and is stored in the longitudinal direction. Further, a label indicating the content of each article 13 is provided on the upper surface in the longitudinal direction, and the name of the article is indicated by capital letters. As shown in the figure, the bucket 12 has a brand “AAA”
5 pieces, "BBB" 2 pieces, "CCC" 2 pieces, "D"
Is one, "EEE" is two, "F" is two, and "G" is two
Are stored in any state. Also, as shown in FIG. 3B, the height of each article is low for "AAA", slightly high for "BBB", "CCC", and "D",
It can be seen that "G" is quite high. Therefore, the focal point of the television camera 14 facing downward vertically is, for example, “BB
If it is adjusted to be on the upper surface of “B”, “AAA” and “G” will not be in focus, and the edge of the detailed image will be blurred.

Hereinafter, the detailed configuration of each part of the present apparatus and its operation will be described with reference to FIGS. FIG. 3 is a flowchart showing the processing procedure of the effective area extracting means 2. First, in step 2A, the edge of the bucket 12 is detected from the captured image 21, and the effective area 22 is cut out. This is because there are various machines and equipment around the bucket 12, and if any of these colors matches the color of the article, disturbance will occur. The image information only in the area 12 is extracted. In order to facilitate this cutout process, it is desirable to apply a specific color to the bucket 12 or to perform accurate illumination.

Next, in step 2B, inspection target pixel designation information CK (v, h) including check bits to which the following bit data is added so that only the pixels of the image corresponding to the extracted bucket range are to be inspected. Create

1: Inspection pixel 0: Non-inspection pixel CK indicates a work area, v indicates a vertical direction, and h indicates a pixel address in a horizontal direction.

Next, in step 2C, noise removal is performed. This is based on the assumption that random noise is scattered due to light rays, shadows, and the like.Specifically, pixels that do not continuously reach a certain area are excluded from the inspection target even in the bucket. Achieved.

FIG. 4 shows a processing procedure of the article color information learning registration means 8. In this process, before conducting the inspection online, the article color information including the color representing the characteristic of each article and the area occupied by the color is analyzed by using the sample 16 of the target article, and this is obtained for each article. Is registered in the memory 7 in advance as an article color information table. Hereinafter, each step will be described.

(Step 8A): A plurality of registered characteristic colors specified for the same article are scheduled, and a loop counter rc for repeating the analysis process by the number is determined. In this example, up to three colors are registered.

(Step 8B): By pointing an arbitrary area of the article image 24 with an input means such as the mouse 23, image information of the designated characteristic color is fetched.

(Step 8C): The color of the captured image information is analyzed and stored.

(Step 8D): In order to confirm that the analyzed color is not different from the designated color, only the image portion of the color is taken out, output and displayed on the display device, and the availability of the operator is determined.

(Step 8E): The availability of the operator is determined
If "K", the process proceeds to the next color process. If "NG", the process returns to step 8B and is executed again.

In this way, the processing of steps 8B to 8E is repeated for the designated number of characteristic colors. FIG. 5 is a diagram for explaining a state of the color analysis processing that is repeatedly performed. As shown in the figure, for example, a feature of a pattern "AAA" is designated, and the color analysis of one pixel is performed.
It is assumed that the brightness levels of G and B are 178, 25, and 50 at 0 to 255 gradations, respectively. This brightness is 0.0
When normalized in the range of ~ 1.0, the normalized brightness r,
g and b are respectively r = 0.7 g = 0.1 b = 0.2.

Next, the number of pixels having the same r, g, b ratio is extracted as the area of the characteristic color. An example of the article color information table thus obtained is shown in FIG.
Shown in As shown in the figure, 3
Lightness r, g,
b and the occupied area per article of the characteristic color are registered in association with each other.

After one piece of article color information 1 to 3 is registered in this way, it is checked whether or not the registered characteristic colors between the articles are similar.

(Step 8F): After the color information for the predetermined number of characteristic colors is registered, the similarity to other articles already registered is checked. In other words, this is to avoid difficulties in distinguishing between different brands having the same color configuration during the actual inspection. Therefore, if any of the registered brands has the same article color information (combination of color information 1 to 3) as the brand to be newly registered, search for the same brand to disable registration. If it exists, a message is output to the operator to register another color as a characteristic color, and the process returns to step 8A.

FIG. 7 shows an example of an inspection information table showing the brands of articles to be in the inspection target bucket 12 and the number of the articles. The inspection information is obtained from another system 10 in advance.
Alternatively, it is manually input and stored in the memory 7 by the test information generating means 9.

FIG. 8 shows the configuration of the color information of the entire image obtained by photographing a plurality of articles in the inspection target bucket 12 (that is, the inspection reference color information). The following describes a procedure for creating a color based on color information.

(Step 3A): Inspection information and article color information for each brand are fetched from the memory 7. (Step 3B): Brand No. specified in the inspection information. m
According to (m = 1 to M), the repetition counter of the brand m is set.

(Step 3C): Number of registrations of color information of the brand specified in Step 3B 1 to RC (in this example,
According to 3), a repetition counter for the number of registered colors rc is set.

(Step 3E): The number of articles corresponding to the brand of the inspection information shown in FIG. 7, the registered characteristic color of the article color information shown in FIG. 6, and the area per article of the registered characteristic color Is used to calculate the area of the registered characteristic color to be present in the screen to obtain inspection reference color information. Then, such a process is performed for each article included in the inspection target.

FIG. 9 shows an example of the inspection reference color information table MEN (m, rc) obtained by the processing of FIG.
For each brand, color area 1 corresponding to each feature color information 1-3
Reference color areas (reference values) 1 to 3 are registered in association with each other.

FIG. 10 shows a processing procedure of the image color information classifying means 4. This process is performed on the pixels in the screen for which the instruction information CK (v, h) of the inspection target pixel created by the effective area extracting unit 2 is “1”.

(Step 4A): A counter v for repeating processing for the number of pixels in the vertical direction is sequentially set to 1 to V.

(Step 4B): A counter h for repeating the processing for the number of pixels in the horizontal direction is sequentially set to 1 to H.

(Step 4C): CK (v, h) is confirmed. If the pixel is the inspection target, the process proceeds to Step 4D. If not, the process returns to Step 4B or 4C.

(Steps 4D) to (Step 4E): As shown in FIG. 11, in Step 4D, the lightness (shade / tone) of each of the three elements R, G, and B of a pixel is set to 0 to 255.
Analyze at the stage. Next, in step 4E, the brightness is normalized to obtain r, g, and b. In the illustrated example, r = 0.70 g = 0.10 b = 0.20 is obtained.

(Step 4F): This is a step of performing an actual color measurement area integration process, and FIG. 12 shows a detailed processing procedure thereof. This process is performed for all brands included in the inspection information and for all registered characteristic colors (steps 4Fb and 4Fc).

(Step 4Fa) to (Step 4F)
c): Initialize the area update counter, set the repetition counter for the number of brands m, and set the repetition counter for the number of registered colors rc.

(Step 4Fd): One registered color information is taken out and it is determined whether or not the registered color information is valid. If invalid, the process proceeds to the process for the next pixel without performing the process for that pixel.

(Step 4Fe): If the registered color information is valid, the pixel GASO (v,
h) color information C (GASO (v,
Regarding h)), it is determined whether or not the color is the same, depending on whether the color matches the registered color or within a certain allowable range.

This processing will be described with reference to FIG. As shown in the figure, the color information of the pixel GASO (v, h) is as follows: C (GASO (v, h)) = (r, g, b) = (0.72, 0.10, 0.18) Suppose. Here, the color information 1 of the brand “AAA” is (r, g, b) = (0.70, 0.1) as shown in FIG.
0, 0.20), so that the allowable range for determining the same is ± 0.03, as shown in FIG. 13, the allowable ranges P (r), P (g), and P (b) of each color element. Is in the range indicated by hatching. As a result, the color information of GASO (v, h) is included in the allowable range of the color information 1 of the brand “AAA”, and is determined to be the same as the color of the color information 1.

Note that the allowable range is not limited to ± 0.03. For example, when more flexible judgment is performed, the difference between the registered color and (r, g, b) is evaluated as a fuzzy amount. What is necessary is just to comprise.

(Step 4Ff) to (Step 4F
h): If it is determined that the color is the same as the registered color, as shown in FIG. 14, the area of one pixel is added to the value in the column of the actually measured value (actually measured color area) of the color information 1 of the brand “AAA”. Update.

(Step 4Fi): 2 for the same pixel
It is determined whether or not the actually measured color areas of the registered colors at more than three locations have been updated.

(Step 4Fj): If double registration is to be performed, the columns of the actually measured values of both corresponding registered colors are set in FIG.
As shown in FIG. 5, "-1" is stored and invalidated. That is, in FIG. 15, “−1” is set to the color area 1 of the brand “AAA” and the color area 3 of the brand “E”.

FIG. 16 shows an image obtained at the time when the processing of the image color information classifying means 4 for all effective pixels is completed after the processing including the invalid measured color area as in the example of FIG. This is a color information classification table MEN (m, rc).

FIG. 17 shows a detailed processing procedure of the inspection judging means 5. The inspection determination means 5 determines the image color information classification table MEN (m, rc) obtained as shown in FIG.
, The quantity for each type of article is determined, and it is determined whether the quantity matches the inspection information. A judgment is made for each registered color of each brand (Steps 5A to 5I), and the result is used to determine whether the number of members for each brand is acceptable (Steps 5J to 5K). Hereinafter, the processing for each step will be described.

(Step 5A): In order to repeat the process for each brand, the repetition pointer m is set from 1 to the maximum brand No.
M is set sequentially.

(Step 5B): As a work area for determination, the maximum value RLMAX of the reliability of the actually measured color area is set to zero and initialized.

(Step 5C): The number of repetitions rc for the registered colors is sequentially set to 1 to RC.

(Step 5D): It is determined whether or not the registered color rc is valid based on whether or not “−1” indicating invalid is stored in the measured color area field of the registered color set in Step 5C.

(Step 5E): If the measured color area of the registered color is not invalid, the reliability RL of the data is calculated as follows.

RL = [{MOD (WCR / WCS)} /
WCS] × 100 (%) where WCS = area per article of registered color rc, WCR = actually measured color area of registered color rc, and MOD (A / B) indicates an integer remainder of A / B.

(Step 5F): Validity / invalidity of the reliability is determined based on whether or not the obtained reliability RL is within the reference range. That is, when RL ≦ RL1, or RL ≧ RL2, the reliability is determined to be valid, and the measured color area of the registered color is also determined to be valid. For example, RL
Assuming that 1 is 5.0 (%) and RL2 is 95.0 (%),
If the RL is 5.0 (%) or less, or 95.0 (%) or more, it can be determined that reliable measurement has been performed. This is because even if the number of articles is an integer multiple due to various noises, the actual color area is not always an integer multiple of one color area. Here, RL1 and RL2 are
It may be determined from the state of the label of the target article, the resolution of the camera, or an experiment. In general, a method of performing fuzzy inference by treating RL as a fuzzy amount is conceivable.

If RL does not satisfy the above inequality, it can be determined that the measurement of the corresponding color has not been performed properly.
It is determined that the measurement of the relevant brand is defective, the symbol “*” is set, and the process proceeds to the next processing (step 5I). For example, the brand “G” in FIG. 18 corresponds to this.

(Step 5G): If the reliability RL is determined to be valid, it is determined whether or not the reliability RL is the maximum of the reliability of each registered color of the article.
That is, it is determined whether or not RL> RLMAX is satisfied by comparing with the already obtained maximum value of reliability RLMAX. If this is not satisfied, the process proceeds to the next registered color, and if it is satisfied, the process proceeds to step 5H.

(Step 5H): First, the criterion for determining the maximum reliability is set to RLMAX = RL, and then the article excess / deficiency value KF of the brand is calculated as follows.

KF = INT ((measured color area−reference color area)
/ (Color area per color) +0.5) where INT is an integer conversion function.

For example, in FIG. 16, the brand “AAA”
Looking at the color area 2 of, the measured color area = 9130, the reference color area = 9145, and the color area per piece = 1829 (see Fig. 6). KF = INT ((9130-9145) / 1829 +
0.5) = INT (0.508) = 0, and it can be determined that the planned number of members exists.

On the other hand, looking at the color area 3 of the brand “BBB” in FIG. 16, the measured color area = 1200, the reference color area = 2422, and the area per piece = 11211 (see FIG. 6). = INT ((1200-2422) / 1211 +
0.5) = INT (−0.51) = − 1, which is estimated to be one less than the expected number of members.

In this way, the reliability of the registered colors is checked for all the brands, and the excess or deficiency of the planned number of members of the brand is estimated using the actually measured color area of the registered color showing the maximum reliability. Obtain the test results as shown in

(Steps 5J) to (Step 5L): If the measured color areas of all brands are reliable (valid) and the numbers match the plan, set ○ as the overall judgment result, otherwise set x. I do. These pieces of information are output and displayed on the display device 15 by the result display processing unit 6, and are notified to the operator.

That is, in the example of FIG. 18, the brand “BB”
"B" indicates one shortage, brand "EEE" is one more, and brand "G" indicates measurement failure. With this information, the worker can quickly review the bucket and correct the quantity of each article to a predetermined number.

FIG. 19 shows the image color information classification result by the inspection after the correction by such an operator. As shown in FIG. 20, the results of the inspection determination process based on this can all match the planned number of members.

In the above description, the reliability of the measured value,
The number of members has been explained using quantitative thresholds and mathematical expressions.However, when imaging conditions are severe, it is needless to say that each state quantity is treated as a fuzzy quantity and a flexible judgment can be made by performing fuzzy inference. No.

As described above, according to the present embodiment,
The characteristic color and the area occupied by the characteristic color are defined as color information unique to the article, the color information of an image obtained by photographing a plurality of types of articles is analyzed, the characteristic color is classified for each screen, and the area is measured. Since the quantity of each type of article is counted based on the actual measurement value, compared to the case of pattern matching, the planned number of members for many types of articles is stored in the bucket at the same time in a short time. In addition to determining whether or not there is an excess or deficiency, it can be determined. Thus, the present invention can be widely applied to a shipping inspection of a picking system, an inspection between manufacturing lines, and the like.

FIG. 21 shows a comparison of the processing time with respect to the number of target articles when the inspection is performed by the conventional pattern matching method and the method according to the present invention. It can be seen that in the conventional method, as the number of articles increases, basic and many processing steps such as enlargement, reduction, and rotation are required, and the processing time is rapidly increasing. In this example, it can be seen that the number of articles is about 30 in terms of processing time, which is the maximum number that can be inspected. However, in practice, it goes without saying that the resolution is limited.

On the other hand, in the method according to the present invention, since the increase in the number of articles hardly affects the increase in the number of image processing procedures, the inspection is completed in 3 seconds or less even when the number of articles reaches 40. In addition, an experimental result has been obtained that since precise shape recognition is not required, deterioration of inspection accuracy is small.

In the above-described embodiment, an example in which the quantity of a plurality of types of articles is counted and inspected has been described. However, the color information of one article is analyzed by utilizing the principle of recognizing articles based on color information of the present invention. Then, by comparing with the registered article color information, it is possible to identify the type of article at high speed and with high accuracy by simple processing.

Similarly, the present invention can be applied to a method of measuring the quantity of articles when a large number of identical articles are stored in a bucket and transported.

[0119]

According to the present invention, there is an effect that the kind and quantity of various kinds of articles can be accurately inspected by a simple apparatus within a very short time.

[Brief description of the drawings]

FIG. 1 is a diagram showing an overall configuration according to an article quantity inspection apparatus according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating a state of an article to be inspected.

FIG. 3 is a flowchart showing a procedure of an effective area extracting method and an explanatory diagram thereof.

FIG. 4 is a flowchart showing a processing procedure of an article color information learning registration unit and an explanatory diagram thereof.

FIG. 5 is a diagram for explaining designation of characteristic color information in a processing procedure of an article color information learning registration unit and an analysis method of the color;

FIG. 6 is an example of an article color information table created by an article color information learning registration unit.

FIG. 7 is an example of an inspection information table in which types and quantities of desired articles serving as inspection standards are defined.

FIG. 8 is a flowchart illustrating a processing procedure of an inspection reference color information generation unit.

FIG. 9 is an example of an inspection reference color information table created by an inspection reference color information generation unit.

FIG. 10 is a flowchart illustrating a processing procedure of an image information classification unit.

FIG. 11 is a diagram for explaining a processing procedure when analyzing the color of a pixel in the processing of the image information classification unit.

FIG. 12 is a flowchart showing a detailed procedure of a main part of a processing procedure of an image information classification unit.

FIG. 13 is a diagram for explaining an allowable range when judging a match between an analysis color and a registered characteristic color in a process of an image information classification unit.

FIG. 14 is a diagram showing an integrated state of actually measured color areas in the image color information classification processing.

FIG. 15 is a diagram illustrating an example in which an actually measured color area is invalidated in image color information classification processing.

FIG. 16 is a diagram illustrating an example of the contents of the processed image color information classification result including a case where the actually measured color area is invalidated in the image color information classification process.

FIG. 17 is a flowchart illustrating a processing procedure of an inspection determining unit.

FIG. 18 is a diagram illustrating an inspection result that is a processing result of an inspection determining unit.

FIG. 19 is a diagram illustrating the contents of the image color information classification result when the type and number of articles are normal.

FIG. 20 is a diagram showing a normal inspection result corresponding to FIG. 19;

FIG. 21 is a diagram showing a comparison between the time required for the inspection of the present invention and the time required for the inspection of the conventional example.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 product quantity inspection device 2 effective area detection means 3 inspection reference color information generation means 4 image color information classification means 5 inspection determination means 6 result display processing means 7 memory 8 article color information learning registration means 9 inspection information generation means 12 bucket 13 Article 14 TV camera 15 Display device 16 Sample 17 TV camera

──────────────────────────────────────────────────続 き Continuing from the front page (72) Inventor Tetsuo Ozawa 3-2-1, Sachimachi, Hitachi City, Ibaraki Prefecture Within Hitachi Engineering Co., Ltd. (56) References JP-A-56-71185 (JP, A) JP-A Sho JP-A-63-263594 (JP, A) JP-A-62-52694 (JP, A) JP-A-3-290781 (JP, A) JP-A-3-179205 (JP, A) A) (58) Field surveyed (Int. Cl. ⁶ , DB name) G06M 7/04 G06M 11/00 B65G 1/137 G06F 15/00

Claims (14)

(57) [Claims] 1. An article quantity inspection method for inspecting whether or not a quantity for each type of article has a predetermined number of members based on image information obtained by photographing a plurality of types of articles to be inspected, In advance, article color information including the characteristic color of the article surface and the area of the characteristic color is registered for each type of article to be inspected, and a plurality of types of articles to be inspected and the number of members for each type are given. Then, based on the registered article color information, the reference color area of the registered feature color is obtained by multiplying the area of the registered feature color of the article by the number of members for each type of article based on the registered article color information. Based on image information obtained by photographing a plurality of articles, a color of each unit area of the image is analyzed, and an actually measured color area of each registered feature color that matches the analyzed color within a certain allowable range is integrated. Integration of measured color area for each registered feature color And on the basis of the degree of coincidence between the reference color area, quantity inspection method of an article, which comprises determining whether there membered appointment quantities for each type of article. 2. The article according to claim 1, wherein the article color information is:
An article quantity inspection method, wherein at least two or more items are set for each article. 3. The quantity inspection of an article according to claim 1, wherein the characteristic color is represented by a combination of three color elements of light indicating original stimuli of mutually independent colors and their lightness. Method. 4. The article quantity inspection method according to claim 1, wherein the degree of coincidence between the integrated value of the actually measured color area and the reference color area is ± 5%. 5. The apparatus according to claim 1, wherein, when there are two or more registered characteristic colors that match the analysis color for each unit area of the image within a predetermined allowable range, the unit area relating to the analysis color is determined by the measured color area. A method for inspecting the quantity of articles, wherein the quantity is not added. 6. The article color information according to claim 1, wherein:
When two or more are set for each article, whether or not the quantity for each type of article has the expected number of members based on the highest degree of coincidence between the integrated value of the actually measured color area and the reference color area A method for inspecting the quantity of articles. 7. A television camera for photographing a plurality of types of articles to be inspected from a predetermined direction, an effective area extracting means for inputting the photographed image information and extracting an effective area where the articles are present, Storage means for storing a plurality of types of articles to be present in the effective area and inspection information of the number of members for each type; a characteristic color of the article surface and an area of the characteristic color for each type of article to be inspected in advance; An article color information storage unit in which article color information is registered, the registration being performed by multiplying the area of the registered characteristic color of the article by the number of members for each type of article based on the article color information and the inspection information. An inspection reference color information generating means for obtaining a reference color area of the characteristic color; and analyzing the color of each unit area of the image existing in the effective area, and for each of the registered characteristic colors matching the analysis color within a certain allowable range. Actual color An image color information classifying means for accumulating the product; and, based on a degree of coincidence between the integrated value of the actually measured color area for each of the registered characteristic colors and the reference color area, whether or not the quantity of each type of article has a predetermined number of members. And an inspection determining means for determining whether or not the number of articles is high. 8. The method according to claim 7, wherein, when the degree of coincidence is higher than a predetermined value, the inspection determining means calculates an integrated value of the actually measured color area as an area per article of the corresponding registered characteristic color. And calculate the quantity of the corresponding article.If the calculated quantity matches the number of corresponding articles, the result is accepted.If not, the excess or deficiency is calculated. An article quantity inspection device comprising a display means for displaying and outputting an inspection result. 9. An image according to claim 7, wherein a plurality of types of articles to be inspected are imaged in accordance with image information obtained for each of the articles, and which coincides with a characteristic color of a designated article surface within a certain allowable range. An article color information learning and registering means for registering article color information comprising a characteristic color and an area of the characteristic color for each type of article in the article color information storage means for each type of article. Quantity inspection device. 10. The article quantity inspection device according to claim 7, wherein at least two or more of the article color information are set for each of the articles. 11. The method according to claim 7, wherein the characteristic color is represented by a combination of three color components of light indicating a primary stimulus of mutually independent colors and their lightness. Characteristic quantity inspection device for goods. 12. The apparatus according to claim 7, wherein the article color information classifying unit determines that the number of the registered characteristic colors that match the analysis color for each unit area of the image within a predetermined allowable range is two or more. A unit area for the analysis color is not added to the actually measured color area. 13. The inspection determining device according to claim 7, wherein the article color information is 2 for each article.
When the above is set, it is determined whether or not the quantity for each type of article has the expected number of members based on the highest degree of coincidence between the integrated value of the actually measured color area and the reference color area. An article quantity inspection device characterized by the following. 14. An article identification method for identifying the type of an article based on image information obtained by photographing one article to be identified, wherein a color representing each feature is set for each type of the article. An article color information including a set color and an area of the color is set, and a color and an area corresponding to the color of the article color information are determined based on the image information. A method for identifying an article, characterized in that a type of an article corresponding to the article color information that matches within a certain allowable range is specified as the type of the article.