JP2015201193A - Image recognition device and image recognition method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image recognition device and an image recognition method with which a user can learn the shape and size of an article by taking an image of the article.SOLUTION: An image recognition device photographs a first image capturing an article (S402), calculates the length of a first diagonal line and the length of a second diagonal line of the article on the basis of the first image (S404 to S410), and when the value of a ratio of the length of the first diagonal line to the length of the second diagonal line falls within a first range (S414: YES), outputs a result of the recognition (S416).

Description

  The present invention relates to an image recognition apparatus and an image recognition method, and more particularly to an image recognition apparatus and an image recognition method for recognizing the shape and size of an article in an image.

  When assembling two articles in an assembly production line, measuring the article size and recognizing the article shape are essential to the assembly process. For example, when assembling glass into an aluminum frame, if it is pressed more than necessary, its joint surface will protrude sharply, or a sharp opening will be formed because it is not completely in close contact, so packaging workers Alternatively, there is a risk that the worker of the assembly work at the installation site at the final stage may be injured. In another example, if the shape of the module changes, when it is assembled to the rail at the installation site in the final stage, even if it is assembled, stress will remain at the locking point, and it will be deformed and function over time. There is a possibility that it may be lost. In more serious cases, even locking is impossible, leading to loss of clients and manufacturers.

  However, the current method still has many problems because a person measures the size of the article or visually recognizes the shape of the article. For example, when the measurer reads the scale of the tape by mistake when measuring the article, or when the shape of the article itself is difficult to measure (such as when the edge of the article is an arc), The article to be recognized becomes unable to be assembled due to a measurement error. And in an assembly production line, if a person measures as described above for each article, it becomes difficult to improve the efficiency of the assembly production line.

  An object of the present invention is to provide an image recognition device and an image recognition method in view of the above problems.

  The image recognition apparatus and the image recognition method according to the present invention determine the shape of an article by comparing the number of pixels of the article in the captured image, and measure the actual size of the article and the number of pixels of the article in the image. By acquiring the size of the article, the user can know the shape and size of the article simply by taking an image of the article.

  The image recognition method disclosed by the present invention includes the following steps. First, a first image showing an article is captured using an image recognition device. Next, based on the first image, the length of the first diagonal line and the length of the second diagonal line of the article are calculated. Furthermore, when the value of the ratio between the length of the first diagonal line and the length of the second diagonal line is within the first range, a recognition result is output.

  The image recognition apparatus disclosed by the present invention includes an imaging module and an image processing module. The imaging module is used to capture an image including an article. The image processing module is communicably connected to the imaging module, and executes an edge line detection program to obtain the positions of a plurality of edge lines in the article from the image, and based on the positions of the plurality of edge lines in the article, The positions of the first diagonal line and the second diagonal line among the plurality of diagonal lines of the article in the image are acquired, the number of pixels at the position of the first diagonal line and the second diagonal line is calculated, and the pixels of the first diagonal line and the second diagonal line It is determined whether the value of the number ratio is within the first range. When the value of the ratio of the number of pixels between the first diagonal line and the second diagonal line is within the first range, a recognition result is output.

As described above, the apparatus and method for recognizing an article in an image according to the present invention obtains a plurality of edge lines of the article in the image by executing an edge line detection program, and the article from the plurality of edge lines of the article. Multiple diagonals, and finally comparing the number of pixels at each of the diagonal positions of the article in the image to determine the shape of the article, so that the user can even take an image of the article You can know the shape of the article.
The above description relating to the contents of the present invention and the description relating to the mode for carrying out the invention described below are intended to illustrate and interpret the gist and principle of the present invention and further interpret the scope of the claims of the present invention. .

It is a functional block diagram of the image recognition apparatus by the Example which concerns on this invention. It is a side view at the time of imaging the image of the articles | goods by the Example which concerns on this invention. 1 is a schematic illustration of an image including an article according to an embodiment of the present invention. 1 is a schematic illustration of an image including a portion of an article according to an embodiment of the present invention. It is a side view at the time of imaging the image of the reference sample by the Example which concerns on this invention. 3 is a flowchart of an image recognition method according to an embodiment of the present invention. 3 is a flowchart of a method for correcting an image recognition apparatus according to an embodiment of the present invention.

  Detailed features and merits of the present invention will be specifically described in the following embodiments for carrying out the invention. The contents are sufficient for those skilled in the art to fully understand and implement the technical contents of the present invention, and based on the contents disclosed in the present specification, the claims and the drawings. Those skilled in the art can easily understand the related objects and advantages of the present invention. The following examples illustrate the invention in more detail, but are not intended to limit the scope of the invention.

  FIG. 1 is a functional block diagram of an image recognition apparatus according to an embodiment of the present invention. As shown in FIG. 1, the apparatus 10 for recognizing an article in an image includes an imaging module 102 and an image processing module 104. The image processing module 104 is communicably connected to the imaging module 102. The apparatus 10 further includes a storage module 106 that stores various types of information. The storage unit 106 is configured by, for example, a ROM, a hard disk, or the like. The storage module 106 is connected to the image processing module 104.

  FIG. 2 is a side view when an image of an article according to the embodiment of the present invention is taken, and FIG. 3 is a schematic view of an image including the article according to the embodiment of the present invention. As shown in the figure, the imaging module 102 is used to capture an image 14 including an article 12 (an image 14 showing the article 12). More specifically, when the user tries to recognize the article 12, the article 12 is placed in parallel on the stage 16 positioned below the apparatus 10 that recognizes the article in the image, and then the article 12 is The imaging module 102 is controlled so as to capture the image 14 including it. The imaging module 102 may be, but is not limited to, a camera lens, a video camera lens, a network imaging device lens, or other appropriate imaging device that can capture the image 14.

  Subsequently, the image processing module 104 described above acquires the positions of the plurality of edge lines of the article 12 from the image 14 by executing the edge line detection program, and also the plurality of edge lines L1, L2, L3, and L4 of the article 12. The positions of the diagonal line DI1 and the diagonal line DI2 among the plurality of diagonal lines of the article 12 in the image 14 are acquired based on the positions of

  For example, when the imaging module 102 captures an image 14 including an article 12 (for example, a rectangle), the imaging module 102 transmits the image 14 to the image processing module 104 and executes an edge line detection program. The operation mode of the edge line detection program will be described in detail later. When the image processing module 104 detects the positions of the edge lines L1, L2, L3, and L4 of the article 12 in the image 14 by the edge line detection program, the intersection position of the edge line L1, the edge line L2, and the edge line L4. In addition, the positions of the diagonal line DI1 and the diagonal line DI2 in the image 14 are detected from the intersection positions of the edge line L3, the edge line L2, and the edge line L4. The image processing module 104 may be a microprocessor, a graphics processing unit, a central process unit, or other elements suitable for arithmetic processing, but the present invention is not limited thereto. Absent.

  The image processing module 104 calculates the number of pixels at the positions of the diagonal line DI1 and the diagonal line DI2, and determines whether the value of the ratio of the number of pixels between the diagonal line DI1 and the diagonal line DI2 is within the first range. When the image processing module 104 determines that the value of the ratio of the number of pixels of the diagonal line DI1 and the diagonal line DI2 is within the first range, the image processing module 104 displays the recognition result, for example, a display not shown or storage Output to module 106.

  More specifically, when acquiring the positions of the diagonal line DI1 and the diagonal line DI2, the image processing module 104 calculates the number of pixels of the diagonal line DI1 and the number of pixels of the diagonal line DI2 along the diagonal line DI1 and the diagonal line DI2, respectively.

  Next, the image processing module 104 calculates the ratio of the number of pixels between the diagonal line DI1 and the diagonal line DI2. That is, the image processing module 104 divides the number of pixels of the diagonal line DI1 by the number of pixels of the diagonal line DI2 to obtain a ratio value. When the obtained ratio value is within the first range (for example, a range from 0.9 to 1.1), the number of pixels of the diagonal line DI1 and the diagonal line DI2 is equal or the difference in the number of pixels is within the allowable range. , Means that the actual lengths of the diagonal line DI1 and the diagonal line DI2 are equal or the difference between the actual lengths is within an allowable range. At this time, the image processing module 104 displays the recognition result, for example, a display not shown, The data is output to the storage module 106. The recognition result is a recognition result that the recognized article 12 is determined to be rectangular, or a recognition result that the recognized article 12 is determined to be pentagonal, hexagonal, or any polygonal shape. However, the present invention is not limited to this.

  FIG. 4 is a schematic view of an image including a part of an article according to an embodiment of the present invention. As shown in FIG. 4, the above-described edge line detection program determines whether or not the gray tone value difference between the pixel 1420 of the image 14 and the other pixels 1412, 1414, 1416,. .., 1428 and the other pixels 1412, 1414, 1416,..., 1428 are higher than the threshold value, the position of the pixel 1420 in the image 14 is a plurality of positions in the article 12. This is the pixel position of the edge lines L1, L2, L3, and L4. However, the pixel 1420 is adjacent to the pixels 1412, 1414, 1416,. In other words, when the user places the article 12 below the imaging module 102 in parallel, if a light source is provided below the article 12, the pixel position where the article 12 is located is partially blocked by the article 12 itself. Therefore, the brightness is lower than the brightness of other pixel positions where the article 12 is not located. Further, since the luminance of the pixel at the edge line position of the article 12 is greatly different from the luminance of the other pixels adjacent to the edge line, the gray gradation value difference between the adjacent pixels is different for each pixel in the image 14. Compare.

  For example, the imaging module 102 captures an image 14 of an article 12 (for example, a pentagon), and then transmits the image 14 including the article 12 to the image processing module 104. The image processing module 104 calculates the gray gradation value of each pixel in the image 14 and also gray gradation between the other pixels 1412, 1414, 1416,. It is calculated whether or not the value difference is higher than a threshold (for example, whether or not it is higher than 125). For example, the gray gradation value of the pixel 1420 is 255, the gray gradation values of the pixels 1412, 1418, and 1424 adjacent to the left of the pixel 1420 are 10, 20, and 50, respectively, and the pixel 1416 that is adjacent to the right of the pixel 1420 , 1422, and 1428 have gray gradation values of 168, 175, and 192, respectively, and the pixel 1414 that is adjacent above the pixel 1420 and the pixel 1426 that is adjacent below the pixel 1420 have gray gradation values of 245 and 252 respectively. . When the gray gradation value of the pixel 1420 and the gray gradation values of the pixels 1412, 1418, and 1424 are subtracted, the difference is higher than the threshold value. When the gray gradation value of the pixel 1420 is compared with the gray gradation values of the pixels 1416, 1422, and 1428, and the gray gradation values of the pixels 1414 and 1426, both are lower than the threshold value. In this case, the image processing module 104 can determine that the pixel 1420 is the left edge line of the article 12.

  In addition, the present invention has a plurality of light sources so that the difference between the gray gradation value of the pixel in the edge line of the present invention and the gray gradation value of the pixel adjacent to the edge line becomes clearer. In addition, the positions of these light sources may be not only below the article 12 but also at a plurality of different positions according to the user's request. In this way, the edge line of the article 12 of the present invention is obtained more accurately.

  In addition, the image processing module 104 further receives a plurality of measurement values corresponding to any of the plurality of edge lines L1, L2, L3, and L4 or the actual lengths of the plurality of diagonal lines DI1 and DI2. In addition, the image processing module 104 acquires the positions of the plurality of edge lines L1, L2, L3, and L4 of the article 12 from the image by executing an edge line detection program. Then, the image processing module 104 determines the pixel size based on the relational expression (Formula 1) between the received measurement value and the number of pixels at the positions of the corresponding edge lines L1, L2, L3, L4 or diagonal lines DI1, DI2. Is calculated. In one embodiment, the user recognizes the article in the image by measuring the actual lengths of the edge lines L1, L2, L3, and L4 and the two diagonal lines DI1 and DI2 of the article 12 in advance. The above measured value may be input to. The image processing module 104 can calculate the pixel size of the image from the actual lengths of the edge lines L1, L2, L3, L4 and the two diagonal lines DI1, DI2, and the number of pixels at that position.

  In another embodiment, when the article 12 is a pentagon, the user measures the actual lengths of the five sides of the pentagon and the actual lengths of the three diagonals before taking the pentagonal image. You may input the above-mentioned measured value into the image recognition apparatus 10 which recognizes the articles | goods in it. Subsequently, the image processing module 104 acquires the positions of the five sides of the pentagon in the image using the above-described edge line detection program, and then acquires the positions of the three diagonal lines of the pentagon in the image based on the positions of the five sides. . Finally, the image processing module 104 can calculate the pixel size of the image 14 from the input actual lengths of the sides and diagonals and the number of pixels at the positions of these sides and diagonals in the image.

  The aforementioned relational expression (Formula 1) is as follows.

  The total ratio value is a ratio value between each of the measured values and the corresponding number of pixels of the edge lines L1, L2, L3, and L4 or the corresponding diagonal lines DI1 and DI2. The measured number is the number of a plurality of received measurement values. For example, when the actual lengths of the edge lines L1, L2, L3, and L4 and the two diagonal lines DI1 and DI2 are measured, the image processing module 104 calculates the actual length of the edge line L1 by the number of pixels in the edge line L1. Divide to obtain the value of the ratio of the edge line L1. Further, the image processing module 104 divides the actual length of the edge line L2 by the number of pixels in the edge line L2 to obtain the value of the ratio of the edge line L2. The same calculation is performed to obtain the value of the ratio of the four edge lines L1, L2, L3, L4 and the two diagonal lines DI1, DI2. The image processing module 104 obtains the pixel size of the image by calculating the arithmetic average of the six ratio values (that is, by adding the six ratio values and dividing by six).

  The relational expression may be a relational expression that calculates the size of a pixel by a weighted average in addition to the above-described mathematical expression 1, and a mathematical expression 2 may be used as such a relational expression.

And
The total length is obtained by selectively weighting and integrating the measured values, and the total number of pixels is obtained by integrating the number of pixels on the edge line or the corresponding diagonal line corresponding to the measured values. Specifically, the image processing module 104 may give a large weight to a part of the measurement value so that the size of the obtained pixel becomes more accurate. For example, the lengths of two diagonal lines DI1, DI2 are longer than those of the other edge lines L1, L2, L3, L4, and the number of pixels of the two diagonal lines DI1, DI2 is the same as that of the other edge lines L1, L2, L3, L4. When the number of pixels is larger, if a large weight (for example, 1.4) is given to the lengths of two diagonal lines, the error of the calculated single pixel becomes the error of the single pixel calculated without giving the weight. And the calculated single pixel size becomes more accurate.

  In addition to the above contents, the present invention is not limited to the fact that the pixel size of the image 14 cannot be calculated unless all the actual lengths of the edge lines L1, L2, L3, L4 and the two diagonal lines DI1, DI2 are measured. Even if only the line L1, the edge line L3, and the diagonal line DI1 are measured, the pixel size of the image 14 can be obtained from the above two relational expressions.

  In addition, the image recognition apparatus 10 for recognizing an article in an image has a gray gradation value for each pixel in the image, the positions of the edge lines L1, L2, L3, and L4 of the article 12 in the image 14 and the diagonal lines DI1 and DI2. A calculation that stores the number of pixels and the measured values of each of the received edge lines L1, L2, L3, and L4 and diagonal lines DI1 and DI2, and is used to take an image of the next article and calculate the size of the article. A storage module 106 connected to the image module 104 for storing the obtained pixel size may be further included.

  In addition to the above contents, in the present invention, a correction program for correcting the image recognition apparatus 10 is further executed (see FIGS. 1 and 5). FIG. 5 is a side view when an image of a reference sample is taken according to an embodiment of the present invention. As shown in the figure, the imaging module 102 captures a second image (not shown) in which the reference sample 18 is captured. Subsequently, the image processing module 104 recognizes the positions of a plurality of edge lines (not shown) of the reference sample 18 from the second image, but the recognition mode is the same as described above, and will not be described here again. The image processing module 104 reads the number of pixels at a plurality of edge line positions of the reference sample 18 and then collates the actual lengths of the plurality of edge lines of the reference sample 18 measured by the received person with the corresponding plurality of pixels. Then, the actual length corresponding to one pixel is calculated (not shown).

  After obtaining the actual length corresponding to each pixel, the image processing module 104 transmits a command signal to the imaging module 102 so that the imaging module 102 captures a third image (not shown). When the imaging module 102 captures the third image, the imaging module 102 transmits the third image to the image processing module 104 and performs image processing, thereby acquiring the number of pixels at a plurality of edge line positions of the reference sample 18 in the third image. To do. The image processing module 104 calculates the plurality of edge lines of the reference sample 18 from the actual length corresponding to each pixel obtained by the above calculation and the number of pixels at the plurality of edge line positions of the reference sample 18 obtained from the third image. Calculate the length.

  The image processing module 104 calculates the lengths of the plurality of edge lines obtained by calculating from the actual lengths of the plurality of edge lines of the reference sample 18 measured by a person as described above and the actual lengths corresponding to the third image and each pixel. From the above, it is determined whether the ratio value of the two lengths is within a second range (for example, 0.9 to 1.1), and the ratio value is determined to be within the second range. If so, stop the correction program. When the image processing module 104 determines that the ratio value is not within the second range, the imaging module 102 captures a fourth image (not shown) and corrects the recognized image. A command signal is transmitted to the imaging module 102 so as to execute the correction program again.

  In order to make those skilled in the art understand the image recognition apparatus for recognizing an article in an image described in the present invention, further description will be given in connection with the image recognition method for recognizing an article in an image according to the present invention (FIGS. 1 and 2 and FIG. 6). FIG. 6 is a flowchart of an image recognition method according to an embodiment of the present invention. As shown in FIG. 6, in step S400, the image processing module 104 receives a plurality of measurement values. Each of these measured values corresponds to one of a plurality of edge lines or actual lengths of a plurality of diagonal lines in the article 12. In step S <b> 402, the imaging module 102 captures an image 14 including the article 12 (an image 14 in which the article 12 is captured). In step S404, the image processing module 104 executes the edge line detection program to acquire the positions of the plurality of edge lines L1, L2, L3, and L4 of the article 12 from the image 14.

  In step S406, the image processing module 104 acquires the positions of the diagonal lines DI1 and DI2 among the plurality of diagonal lines of the article 12 in the image 14 based on the positions of the plurality of edge lines L1, L2, L3, and L4 of the article 12. To do. In step S408, the image processing module 104 calculates the number of pixels at the positions of the diagonal line DI1 and the diagonal line DI2 of the article in the image 14. In step S410, the image processing module 104 calculates the number of pixels at the positions of the edge lines L1, L2, L3, and L4 of the article 12 in the image 14. In step S412, the image processing module 104 calculates the pixel size of the image 14 from the number of pixels at the positions of the edge lines L1, L2, L3, and L4 and the diagonal lines DI1 and DI2 of the article 12. In step S414, the image processing module 104 determines whether the value of the ratio of the number of pixels of the diagonal line DI1 and the diagonal line DI2 is within the first range. When the image processing module 104 determines that the value of the ratio of the number of pixels of the diagonal line DI1 and the diagonal line DI2 is within the first range, the image processing module 104 outputs a recognition result indicating that the article 12 is rectangular (step S416). When the image processing module 104 determines that the value of the ratio of the number of pixels of the diagonal line DI1 and the diagonal line DI2 is not within the first range, the image processing module 104 outputs another recognition result indicating that the article 12 is not rectangular (step S418).

  The order of step S400, step S402, step S404, and step S406 may be changed. For example, in the present invention, step S402, step S404, and step S406 may be performed first, and then the operation of step S400 may be performed. In addition, the order of step S408 and step S410 can be interchanged. That is, first, the number of pixels at the positions of the edge lines L1, L2, L3, and L4 of the article 12 in the image 14 is calculated, and then the number of pixels at the positions of the diagonal lines DI1 and DI2 of the article 12 in the image 14 is calculated. May be. This may be determined according to the user's request, but the present invention is not limited thereto.

  Next, description will be made with reference to FIGS. FIG. 7 is a flowchart of a method for correcting an image recognition apparatus according to an embodiment of the present invention. As shown in the figure, in step S500, the image processing module 104 receives the actual lengths of a plurality of edge lines of the reference sample 18 measured by a person. In step S502, the imaging module 102 captures a second image (not shown) in which the reference sample 18 is captured. In step S <b> 504, the image processing module 104 reads a plurality of pixel numbers at positions of a plurality of edge lines (not shown) of the reference sample 18. In step S <b> 506, the image processing module 104 compares the actual lengths of the plurality of edge lines of the reference sample 18 with the corresponding number of pixels to calculate the actual length corresponding to one pixel (not shown). In step S508, the imaging module 102 captures a third image (not shown) in which the reference sample 18 is captured.

  In step S510, the image processing module 104 calculates the lengths of the plurality of edge lines of the reference sample 18 based on the third image and the actual length corresponding to each pixel. In step S512, the image processing module 104 determines whether the ratio value between the actual lengths of the plurality of edge lines of the reference sample 18 and the length calculated from the third image is within the second range. If the image processing module 104 determines that the ratio of the actual length of the plurality of edge lines of the reference sample 18 and the length calculated from the third image is not within the second range, the step in the image recognition method according to the present invention The method for correcting the image recognition apparatus after S500 is executed again. On the other hand, when the image processing module 104 determines that the ratio value between the actual lengths of the plurality of edge lines of the reference sample 18 and the length calculated from the third image is within the second range, the correction program ends. (Step S514). Thereafter, the image recognition method according to the present invention is started (step S400).

  As described above, the image recognition apparatus for recognizing an article in an image according to the present invention acquires a plurality of edge lines of an article in an image by executing an edge line detection program, and a plurality of articles from the plurality of edge lines of the article. Can be obtained. Then, by comparing the number of pixels at the positions of a plurality of diagonal lines of the article in the image, the user can know the shape of the article by taking an image of the article. Furthermore, the present invention obtains the pixel size in the image by measuring the actual lengths of the edge and diagonal lines of the article and the number of pixels on the edge and diagonal lines of the article in the image. This eliminates the need for the user to measure the size of all the articles, and if only the actual length of the first article is measured, the image of the article is taken for the other subsequent articles, and the image The actual lengths of the edge lines of other articles can be directly acquired from the pixel size.

In addition, the present invention supports pixels from the number of pixels at the positions of the plurality of edge lines of the reference sample in the captured second image and the actual length of the plurality of edge lines of the reference sample measured by a person. And a correction program for correcting the image recognition device for calculating the actual length to be calculated. Further, the lengths of the plurality of edge lines of the reference sample are calculated from the actual length corresponding to the pixel and the captured third image. Finally, the image processing module determines whether the value of the ratio between the actual lengths of the plurality of edge lines measured by a person and the calculated lengths of the plurality of edge lines is within the second range. Here, not the manual correction by the user but the automatic correction is performed by the image recognition device, and the accuracy of the article measurement by the assembly production line is increased.

Claims (10)

An image recognition method executed by an image recognition apparatus,
Capturing a first image of an article;
Calculating a length of a first diagonal line and a length of a second diagonal line of the article based on the first image;
Outputting a recognition result when a value of a ratio between the length of the first diagonal line and the length of the second diagonal line is within a first range;
An image recognition method including: Calculating the length of the first diagonal and the length of the second diagonal of the article,
Acquiring the positions of a plurality of edge lines in the article from the first image by executing an edge line detection program;
Based on the positions of the plurality of edge lines in the article, the positions of the first diagonal line and the second diagonal line of the article in the first image are acquired, and pixels at the positions of the first diagonal line and the second diagonal line are obtained. Reading a number to calculate the length of the first diagonal and the length of the second diagonal of the article;
including,
The image recognition method according to claim 1. The edge line detection program is for comparing whether or not a gray gradation value difference between a first pixel in the first image and a second pixel adjacent to the first pixel is higher than a threshold value. When the gray gradation value difference between the first pixel and the second pixel in the first image is higher than a threshold value, the position of the first pixel in the first image is a plurality of pixels of the edge line in the article It is determined that the position is
The image recognition method according to claim 2. Receiving a plurality of measurement values respectively corresponding to actual lengths of a plurality of diagonal lines including the plurality of edge lines and the first diagonal line and the second diagonal line;
After the step of obtaining the positions of the plurality of edge lines in the article from the first image by executing the edge line detection program, the plurality of measurement values, the plurality of edge lines, and the plurality of diagonal lines Obtaining and outputting a size of a pixel in the first image based on a relational expression with any of the positions;
The image recognition method according to claim 3. The relational expression is

And
The total ratio value is the sum of the ratio values of the plurality of measurement values and the corresponding number of pixels of the edge line and the plurality of diagonal lines, and the measurement number is the received measurement value The number of
The image recognition method according to claim 4. The relational expression is

And
The total length is obtained by selectively weighting and integrating a plurality of measurement values, and the total pixel number is obtained by integrating the corresponding number of pixels on the edge line or the corresponding plurality of diagonal lines. ,
The image recognition method according to claim 4. Before the step of capturing the first image showing the article,
Including a step of executing a correction program;
The correction program is
(A) receiving actual lengths of a plurality of edge lines of a reference sample measured by a person;
(B) capturing a second image of the reference sample by the image recognition device;
(C) reading a plurality of pixel numbers at a plurality of positions of the edge lines of the reference sample based on the second image;
(D) collating the actual length of the plurality of edge lines of the reference sample with the corresponding number of pixels to calculate the corresponding actual length for each pixel;
(E) re-imaging the third image showing the reference sample by the image recognition device;
(F) calculating the lengths of the plurality of edge lines of the reference sample based on the third image and the actual length corresponding to each pixel;
(G) determining whether the value of the ratio between the actual length of the plurality of edge lines of the reference sample and the length calculated from the third image is within a second range;
Including
When it is determined that the ratio value between the actual lengths of the plurality of edge lines of the reference sample and the length calculated from the third image is within the second range, the execution of the correction program is terminated. When it is determined that the ratio value between the actual length of the plurality of edge lines of the reference sample and the length calculated from the third image is not within the second range, the steps (a) to ( g) repeat,
The image recognition method according to claim 1. An imaging module used to capture an image including the article;
An image processing module communicably connected to the imaging module;
With
The image processing module executes the edge line detection program to obtain the positions of the plurality of edge lines in the article from the image, and based on the positions of the plurality of edge lines in the article, the image processing module Obtaining the positions of the first diagonal line and the second diagonal line of the plurality of diagonal lines of the article, calculating the number of pixels at the position of the first diagonal line and the second diagonal line, the first diagonal line and the second diagonal line, It is determined whether or not the ratio value of the number of pixels is within the first range, and the ratio value of the number of pixels between the first diagonal line and the second diagonal line is within the first range. Output the result,
Image recognition device. The edge line detection program is for comparing whether or not a difference in gray gradation value between a first pixel in the image and a second pixel adjacent to the first pixel is higher than a threshold value. When the difference in gray gradation value between the first pixel and the second pixel is higher than the threshold, the position of the first pixel in the image is acquired as the position of a plurality of edge line pixels in the article. And
The image processing module receives a plurality of measurement values respectively corresponding to actual lengths of the plurality of edge lines or the plurality of diagonal lines, and further executes the edge line detection program , After obtaining the positions of the plurality of edge lines in the article from the image, the image based on a relational expression between the plurality of measured values and the positions of the plurality of edge lines and the plurality of diagonal lines Get and output the pixel size at
The relational expression is

And
The total ratio value is obtained by integrating the ratio values of the plurality of measurement values and the corresponding number of pixels of the edge line and the plurality of diagonal lines, and the measurement number is a plurality of received values. The number of measurements.
The image recognition apparatus according to claim 8. The image processing module receives a plurality of measurement values respectively corresponding to actual lengths of the plurality of edge lines or the plurality of diagonal lines, and further executes the edge line detection program , After obtaining the positions of the plurality of edge lines in the article from the image, the image based on a relational expression between the plurality of measured values and the positions of the plurality of edge lines and the plurality of diagonal lines Get and output the pixel size at
The relational expression is

And
The total length is obtained by selectively weighting and integrating a plurality of measurement values, and the total pixel number is obtained by integrating the corresponding number of pixels on the edge line or the corresponding plurality of diagonal lines. ,
The image recognition apparatus according to claim 8.

Images