CN116997769A - Inspection device, inspection method, glass plate manufacturing method, and inspection program - Google Patents

Abstract

The application reduces the labor cost in the size detection of defects generated in the glass plate and maintains the detection precision. The inspection device (1) comprises: an image acquisition unit (101) that acquires a captured image obtained by capturing a glass plate; and a size detection unit (103) that detects, as the size of the defect, the size of the range in the estimation result obtained by inputting the captured image to a size detection model (112), the size detection model (112) being a model that learns the position and range of the defect generated in the glass sheet so as to estimate the position and range of the defect generated in the glass sheet.

Description

Inspection device, inspection method, glass plate manufacturing method, and inspection program

Technical Field

The present application relates to an inspection apparatus and the like for inspecting defects of a glass sheet based on capturing an image of the glass sheet.

Background

Conventionally, a defect inspection of a glass sheet at the time of manufacturing the glass sheet is performed by visual inspection, and particularly, a size detection of a defect generated in the glass sheet is unavoidable. To solve this problem, patent document 1 discloses a technique for detecting the size of a defect by performing image processing on an image of an end surface of a glass plate.

Prior art literature

Patent literature

Patent document 1: international publication No. 2004/079352

Disclosure of Invention

Problems to be solved by the application

The inventors of the present application have found that the conventional technique described above may not detect the size of the defect with sufficient accuracy. One of the reasons for this is considered to be that the brightness of the periphery of the defect is unclear in the photographed image of the glass plate.

The purpose of the present application is to provide an inspection device or the like that can reduce labor costs and maintain inspection accuracy when detecting the size of defects that occur in a glass sheet.

Technical scheme for solving problems

In order to solve the above problems, an inspection apparatus according to an aspect of the present application includes: an image acquisition unit that acquires a captured image obtained by capturing a glass plate; and a size detection unit that detects, as the size of the defect generated in the glass sheet that has been photographed, the size of the range in the result of the inference obtained by inputting the photographed image to a trained model that learns the position and range of the defect generated in the glass sheet so that the position and range of the defect generated in the glass sheet are inferred.

In order to solve the above-described problems, an inspection method according to an aspect of the present application is an inspection method performed by an inspection apparatus, including: an image acquisition step of acquiring a photographed image obtained by photographing the glass plate; and a size detection step of detecting, as the size of the defect generated in the glass sheet that is photographed, the size of the range in the inferred result obtained by inputting the photographed image to a trained model that learns the position and range of the defect generated in the glass sheet so that the position and range of the defect generated in the glass sheet is inferred.

In order to solve the above-described problems, a method for manufacturing a glass sheet according to an aspect of the present application includes a step of molding a glass raw sheet into a glass sheet having a predetermined size, and an inspection step of the glass sheet performed by an inspection apparatus, wherein the inspection step includes: an image acquisition step of acquiring a photographed image obtained by photographing the glass plate; and a size detection step of detecting, as the size of the defect generated in the glass plate, the size of the range in the inferred result obtained by inputting the captured image to a trained model that learns the position and range of the defect generated in the glass plate so that the position and range of the defect generated in the glass plate are inferred.

Effects of the application

According to one aspect of the present application, in size detection of defects generated in a glass sheet, it is possible to reduce labor costs and to maintain detection accuracy.

Drawings

Fig. 1 is a block diagram showing an outline of an inspection system according to embodiment 1 of the present application and an example of a main configuration of an inspection apparatus included in the inspection system.

Fig. 2 is a diagram showing an example of the position and the range of the predicted defect based on the size detection model.

Fig. 3 is a flowchart showing an example of a process for manufacturing a glass sheet in association with inspection using the inspection apparatus.

Fig. 4 is a flowchart showing an example of the inspection process included in the manufacturing process.

Fig. 5 is a diagram showing an example of size detection of a defect based on reliability.

Fig. 6 is a flowchart showing an example of the inspection process according to embodiment 2 of the present application.

Fig. 7 is a diagram showing an outline of the defect size detection performed by the size detection unit according to embodiment 3 of the present application.

Fig. 8 is a flowchart showing an example of the inspection process according to embodiment 3 of the present application.

Fig. 9 is a flowchart showing an example of the inspection process according to embodiment 4 of the present application.

Detailed Description

Embodiment 1

(inspection System 100)

An embodiment of the present application will be described in detail below. Fig. 1 is a block diagram showing an outline of an inspection system 100 according to the present embodiment and an example of a main configuration of an inspection apparatus 1 included in the inspection system 100.

The inspection system 100 is a system that detects the size of defects generated in the end face of a glass sheet. In the present embodiment, a case where the glass plate is rectangular is described, but the shape of the glass plate is not limited to this example. The defects include, for example, cracks, flaws, and the like generated at the end face of the glass plate. The inspection system 100 includes an inspection device 1, a photographing device 2, an image storage device 3, and a detection result storage device 4.

The inspection apparatus 1 is an apparatus for detecting the size of a defect generated in an end face of a glass plate. Conventionally, size detection of defects on an end surface of a glass plate is performed visually, and in size detection using image processing, there is a case where it is difficult to perform accurate size detection because brightness around defects in a captured image is unclear. The inspection apparatus 1 detects, as the size of the defect, the size of a range in the result of estimating the defect obtained by inputting the captured image to the trained model that has been learned to estimate the position and the range of the defect generated in the end face, as will be described later. That is, since no human labor is involved in the dimension detection of the defect of the end face, the human labor cost associated with the detection can be reduced. In addition, in the learning of the trained model, the image of the end face with unclear brightness around the defect is learned as training data a plurality of times, whereby the defect can be detected with high accuracy even from the end face image with unclear brightness around the defect. Therefore, in detecting the size of the defect generated in the end face of the glass plate, the labor cost can be reduced and the detection accuracy can be maintained.

The imaging device 2 is a device that images an end surface of a glass plate. Here, the end face means a side face portion when the two widest surfaces of the glass plate are respectively the upper surface and the lower surface, and may be referred to as an outer peripheral portion. Hereinafter, the four end surfaces of the glass plate may be referred to as an X1 end surface, an X2 end surface, a Y1 end surface, and a Y2 end surface, respectively. The end faces X1 and X2 are parallel to each other, and the end faces Y1 and Y2 are parallel to each other.

As an example, the imaging device 2 is disposed at a position along a conveyance path of a conveyance device (not shown) that conveys a glass sheet, and continuously images one of end surfaces of the glass sheet being conveyed by the conveyance device at predetermined time intervals. In this way, by imaging one end surface a plurality of times, a plurality of imaged images in which one end surface is divided a plurality of times can be obtained, and the resolution of the end surface of each imaged image can be sufficiently increased to the degree necessary for inspection. Of course, if a shot image of sufficiently high resolution is obtained by shooting the entire one end face by one shot, the number of shots may be one.

Fig. 1 illustrates only one photographing device 2, but the photographing device 2 may be provided for the end face of each glass plate. For example, the two photographing devices 2 may be disposed to face each other across the conveying path, thereby photographing end faces (for example, X1 end face and X2 end face or Y1 end face and Y2 end face) parallel to each other at the same time. In the case where the conveyance path branches, the imaging device 2 may be provided for each branch.

Although not shown, the imaging device 2 includes an illumination device and an information processing device, and thus the end face of the glass plate can be imaged by the imaging device 2 while light is irradiated to the end face by the illumination device.

The information processing device is a device that adds additional information to the captured image captured by the imaging device 2 and stores the additional information in the image storage device 3. The additional information includes glass identification information indicating a glass plate in the captured image and glass position information indicating which part of the glass plate the captured image is.

The glass identification information may be, for example, an identification number provided on each glass plate. The glass position information may be information indicating which part of the glass plate. For example, in the case where the conveyance speed of the glass plate is constant and the number of shots per one end face is also constant, information indicating a shot image obtained by shooting several times may be used as the glass position information.

The image storage device 3 is a storage device that stores the captured image captured by the imaging device 2. The detection result storage device 4 is a storage device that stores information indicating the size of the defect, which is the detection result detected by the inspection device 1. The image storage device 3 and the detection result storage device 4 may be provided in plural numbers, respectively. For example, the image storage device 3 and the detection result storage device 4 may be provided for the end face of each glass plate. In addition, the image storage device 3 and the detection result storage device 4 may be omitted. In this case, the imaging device 2 may be configured to transmit the captured image to the inspection device 1, and the inspection device 1 may store the detection result in the storage unit 11.

(inspection device 1)

As shown in fig. 1, the inspection apparatus 1 includes a control unit 10, a storage unit 11, and a communication unit 12. The control unit 10 controls the respective units of the inspection apparatus 1 as a whole. The storage unit 11 stores various data used by the inspection apparatus 1. The communication unit 12 is used to communicate with the inspection apparatus 1 and other apparatuses. As typical examples of the other devices, there are an image storage device 3 and a detection result storage device 4.

As shown in fig. 1, the control section 10 includes an image acquisition section 101, a defect determination section 102, and a size detection section 103. The storage section 11 stores a defect judgment model 111 and a size detection model 112.

The image acquisition unit 101 acquires a captured image in which a glass plate is captured. As described above, in the present embodiment, the captured image acquired by the image acquisition unit 101 is a captured image of the end surface of the glass plate captured by the imaging device 2. As an example, the image acquisition unit 101 receives a captured image from the image storage device 3 via the communication unit 12.

The defect determination unit 102 determines whether or not there is a defect in the end face of the captured image. Specifically, the defect determination unit 102 inputs the captured image acquired from the image acquisition unit 101 to the defect determination model 111, and determines the presence or absence of a defect based on the estimation result output from the defect determination model 111.

Here, the defect determination model 111 will be described. The defect determination model 111 is a trained model that is learned to infer whether or not there is a defect in a captured image in which the end face of the glass plate is captured. Such a defect judgment model 111 can be constructed by machine learning using a plurality of captured images for which whether or not a defect is known as training data. The algorithm of the machine learning is not particularly limited as long as the defect determination model 111 classifying the captured image into two types of defects and non-defects can be generated. For example, a deep learning convolutional neural network or the like with high classification accuracy of an image is preferable, but this is not a limitation.

In addition, the defect determination unit 102 may determine the type of defect. In the case where the type of defect is also determined, training data may be prepared for each type of defect, and machine learning may be performed using these as training data. Examples of the type of defect include the fracture and the defect.

The size detection unit 103 detects the size of the range of defects in the estimation result obtained by inputting the captured image to the size detection model 112 as the size of defects generated in the captured glass sheet.

Here, the size detection model 112 will be described. The dimension detection model 112 is a trained model that learns the location and extent of defects generated in a glass sheet to infer the location and extent of defects generated in the glass sheet. Such a size detection model 112 may be constructed by machine learning using training data that correlates the position and extent of the defect as correct data to the captured image of the defect. The location and extent of a defect may be represented, for example, by a rectangle surrounding the defect. In this case, the position of the rectangle indicates the position of the defect, and the width and height of the rectangle indicate the range, i.e., the size, of the defect.

As such training data, a plurality of captured images in which the brightness around the defect is unclear are preferably used. Thus, the size of the defect can be detected with high accuracy from the photographed image in which the brightness around the defect is unclear. Similarly to the defect determination model 111, the algorithm of the machine learning is not particularly limited.

In addition, in the case where the defect determination unit 102 determines the type of defect, the size detection model 112 may be prepared in advance for each type of defect, and the size detection unit 103 may perform size detection using the size detection model 112 corresponding to the type determined by the defect determination unit 102. This can improve the accuracy of detecting the size. For example, the size detection model 112 for the broken defect may be constructed using a shot image of the end face of the glass plate in which the broken is generated as training data, and the size detection model 112 for the broken defect may be constructed using a shot image of the end face of the glass plate in which the broken is generated as training data. In this case, the size detection of the captured image in which the defect determination unit 102 determines that the broken defect is generated may be performed by using the size detection model 112 for the broken defect. On the other hand, the size detection model 112 for the defect may be used to detect the size of the captured image in which the defect determination unit 102 determines that the defect is generated.

(example of inference)

Fig. 2 is a diagram showing an example of the position and the range of the inferred defect by the size detection model 112. In this example, the image of the end face of the glass plate captured by the imaging device 2 is a captured image 21. The black portion extending laterally in the center portion of the captured image 21 is an end face of the glass plate, and a portion in the vicinity of the center portion of the end face of the glass plate, which is photographed white, is a defect. As described above, the captured image 21 is stored in the image storage device 3.

The image acquisition unit 101 of the inspection apparatus 1 acquires the captured image 21 from the image storage device 3. When the captured image 21 is acquired, first, the defect determination section 102 inputs the captured image 21 to the defect determination model 111, and determines whether or not there is a defect based on the estimation result output from the defect determination model 111. Since the defect described above is shot in the shot image 21, the defect determination section 102 determines that the defect is present. As described above, the defect determination unit 102 may determine the type of defect, but only determine whether the defect exists or not.

The size detection by the size detection unit 103 is performed on the captured image 21 determined to be defective by the defect determination unit 102. That is, the size detection unit 103 inputs the captured image 21 to the size detection model 112, and obtains the estimation result output by the size detection model 112. The result of the inference indicates the position and the range of the defect in the captured image 21.

The size detection unit 103 stores the range indicated by the estimation result in the detection result storage device 4 as size information indicating the size of the defect. For example, in the case where the above-described range is a rectangle, the size detection unit 103 may store representative coordinates (for example, coordinates of the upper left corner of the rectangle) indicating the position of the rectangle in the captured image 21 and size information indicating the width and height of the rectangle.

In the image 31 shown in fig. 2, the result of the inference of the size detection model 112 is shown as a rectangle 41. As shown, the width and height of the rectangle 41 are equal to those of the defect, and it is known that the correct estimation is performed.

Such a rectangle may be referred to as an annotation. By displaying the comment, the user of the inspection apparatus 1 can visually confirm the estimation result. As described above, the size detection model 112 constructed by machine learning with the captured image of the shot defect as training data is used, so that the defective portion can be accurately detected.

(inspection in the production Process of glass sheet)

The inspection by the inspection device 1 may be performed as a loop in the process of manufacturing the glass sheet. An example of a process for manufacturing a glass plate according to the inspection by the inspection apparatus 1 will be described with reference to fig. 3. Fig. 3 is a flowchart showing an example of a process for manufacturing a glass sheet according to the inspection by the inspection apparatus 1.

In S101, a glass raw plate is formed into a glass plate having a predetermined size. The glass raw plate is a glass plate having a larger size than the glass plate to be a product, and is manufactured by a manufacturing apparatus for the glass raw plate. In S101, the glass raw plate is adjusted to a glass plate of a predetermined size by, for example, a cutting device for cutting the glass raw plate to a predetermined size and a processing device for processing an end surface of the glass plate after cutting.

In S102, the inspection apparatus 1 performs an inspection process. The glass plate judged to be good in the inspection step becomes a product. The details of the inspection process are described below with reference to fig. 4.

(flow of inspection Process)

Fig. 4 is a flowchart showing an example of the inspection process shown in fig. 3. Before the inspection process is started, the imaging device 2 images the end face of the glass plate to be inspected and stores the imaged image in the image storage device 3. The time point at which the inspection process shown in fig. 4 is performed is not particularly limited. For example, the imaging may be performed each time a new image is captured and stored in the image storage device 3, may be performed each time the imaging of one end face of one glass plate is completed, or may be performed after the imaging of all end faces of all glass plates to be inspected is completed.

In S1 (image acquisition step), the image acquisition section 101 acquires a captured image from the image storage device 3. In the case where a plurality of captured images are stored in the image storage device 3, it is only necessary to acquire captured images that have not yet been supplied to the inspection process. Then, the image acquisition unit 101 outputs the acquired captured image to the defect determination unit 102.

In S2, the defect determination unit 102 inputs the captured image acquired in S1 to the defect determination model 111, and determines whether or not there is a defect based on the estimation result output from the defect determination model 111. Here, the set defect determination unit 102 also determines the type of defect.

In S3, the size detection unit 103 determines, as a size detection model used in the size detection, the size detection model 112 corresponding to the defect type determined in S2, among the plurality of size detection models stored in the storage unit 11 in advance. If it is determined in S2 that the image is not defective, the processing after S3 is not performed, and the inspection process for the captured image acquired in S1 is completed.

In S4 (size detection step), the size detection unit 103 inputs the captured image acquired in S1 to the size detection model 112 specified in S3. Then, the size detection unit 103 detects the size of the range (in other words, the size of the range indicated by the estimation result) in the estimation result outputted from the size detection model 112 as the size of the defect. For example, in the case where the above-described range is rectangular, the size detection section 103 detects the width and the height of the rectangle as the width and the height of the defect. The size detection unit 103 may convert the detected width and height into actual sizes.

In S5, the size detection unit 103 stores the size detected in S4 in the detection result storage device 4. Specifically, the size detection unit 103 transmits and stores size information indicating the width and height of the defect to the detection result storage device 4. This ends the inspection process for the captured image acquired in S1.

Although not shown in fig. 4, the inspection process of fig. 4 is repeated until the inspection of the captured image captured on all the end surfaces of at least one glass plate to be inspected is completed, and thereafter, the process returns to the manufacturing process of fig. 3.

The inspection step of fig. 4 may further include a step of judging whether the glass plate has good or defective products with respect to all the judged end faces. In this step, the glass plate judged to be good becomes a product. The judgment criteria for good products and defective products may be appropriately defined. For example, a glass plate in which a defect of a predetermined size or more is detected may be set as a defective product, and a glass plate in which a defect is not detected or a detected defect is smaller than a predetermined size may be set as a good product. This determination may be performed by the inspection apparatus 1 or by another information processing apparatus different from the inspection apparatus 1.

(action and Effect)

As described above, the inspection apparatus 1 of the present embodiment includes the image acquisition unit 101, and the image acquisition unit 101 acquires a captured image obtained by capturing an image of an end surface of a glass plate. Further, the inspection apparatus 1 includes a size detection unit 103, and the size detection unit 103 detects the size of the corresponding range in the estimation result obtained by inputting the captured image to the size detection model 112 as the size of the defect generated in the captured glass sheet. Here, the size detection model 112 is a trained model that learns the position and the range of defects generated in the end face to infer the position and the range of defects generated in the end face.

As described above, the inspection method according to the present embodiment includes: an image acquisition step (S1) of acquiring a photographed image obtained by photographing the glass plate; and a size detection step (S4) of detecting, as the size of the defect generated in the glass sheet that is photographed, the size of the range in the inferred result obtained by inputting the photographed image to a size detection model 112, the size detection model 112 being a model that learns the position and range of the defect generated in the glass sheet to infer the position and range of the defect generated in the glass sheet.

As described above, the method for manufacturing a glass sheet according to the present embodiment includes: a step (S101) of forming a glass raw plate into a glass plate of a predetermined size; and an inspection step (S102) of the glass sheet performed by the inspection apparatus 1. The inspection step includes: an image acquisition step (S1) of acquiring a photographed image obtained by photographing the glass plate; and a defect detection step (S4) of detecting, as the size of the defect generated in the glass sheet, the size of the range in the estimation result obtained by inputting the captured image to a size detection model 112, the size detection model 112 being a trained model that learns the position and range of the defect generated in the glass sheet to estimate the position and range of the defect generated in the glass sheet.

According to this configuration, since no human effort is required for detecting the size of the defect, the labor cost associated with the detection can be reduced. In addition, in the learning of the size detection model 112, a plurality of images in which the brightness is unclear around the defect are learned as training data, and thus the defect can be detected with high accuracy from the image in which the brightness is unclear around the defect. Therefore, in the size detection of defects generated in the glass plate, the detection accuracy can be maintained while the labor cost can be reduced.

[ embodiment 2 ]

Hereinafter, other embodiments of the present application will be described. The same reference numerals are given to members having the same functions as those described in the above embodiments, and repeated description thereof is omitted for convenience of description. The same applies to embodiment 3 and embodiment 4.

In this embodiment, an example will be described in which the size of a defect is detected based on the reliability of the estimation result output together with the estimation result by the size detection model 112. More specifically, the size detection unit 103 of the present embodiment detects the size of a rectangle in the estimation result of the reliability equal to or higher than the predetermined threshold value as the size of a defect generated in the glass sheet to be imaged.

The reliability is a value indicating the certainty of the estimation result, and may be a value of 0 to 1. In the reliability of the present embodiment, a larger value indicates a greater likelihood that the result of the estimation is an actual defect.

The reliability can be calculated by, for example, performing numerical analysis on pixel values of respective pixels constituting a captured image. In the photographed image, there is a large variation in pixel values (brightness) of a region where a defect exists and a region where a defect does not exist. Therefore, the reliability can be calculated from the amount of change in the pixel value, for example, the reliability is low when the amount of change in the pixel value is small on the boundary between the inside and the outside of the range indicated by the estimation result of the size detection model 112, the reliability is high when the amount of change in the pixel value is large, and the like. Thereby, the reliability of the value corresponding to the amount of change in the pixel value is calculated. For example, the size detection model 112 may be used, which outputs a numerical value indicating the reliability of the estimation result together with the estimation result. In this case, processing such as numerical analysis is not required.

(example of reliability-based size detection)

Fig. 5 is a diagram showing an example of size detection of a defect based on reliability. The captured image 32 shown in fig. 5 is a diagram in which rectangles 42 to 44 indicating the estimation results of the size detection model 112 and numerical values 52 to 54 indicating the reliability of these estimation results are drawn in the captured image of the end face of the glass plate in which the defect exists.

As shown, in this example, the dimension detection model 112 concludes that the ranges shown by rectangles 42-44, respectively, are defects, and these inferred reliability values 52-54 are 0.95, 0.90, and 0.75, respectively.

Here, for example, the threshold value of the reliability is set to 0.80. In this case, the size detection unit 103 detects the sizes of the rectangles 42 and 43 having a reliability of 0.80 or more among the rectangles 42 to 44 as the sizes of the defects, respectively. On the other hand, the size detection section 103 does not detect the size of the rectangle 44 having a reliability of less than 0.80 as the size of the defect.

Therefore, as shown in the lower side of fig. 5, when the result of the final size detection based on the size detection section 103 is displayed on the captured image 32, only the rectangles 42 and 43 are displayed. In this case, the size detection unit 103 transmits and stores size information indicating the width and height of the rectangle 42 and size information indicating the width and height of the rectangle 43 to the detection result storage device 4.

(flow of inspection Process)

Fig. 6 is a flowchart showing an example of the inspection process according to the present embodiment. In this flowchart, the same processing as that of the flowchart of fig. 4 is given the same reference numerals as those of fig. 4, and the repetitive description thereof is omitted.

In S11, the size detection unit 103 inputs the captured image 21 to the size detection model 112 determined in S3, and obtains the estimation result of the size of the defect. The inferred result includes the reliability of the inference in addition to the position and the size of the range inferred by the size detection model 112 as the existence of the defect.

In S12, the size detection unit 103 detects the size of the range in the estimation result in which the reliability output in S11 is equal to or greater than the threshold value as the size of the defect.

(effects of action)

As described above, in the inspection apparatus 1 of the present embodiment, the size detection model 112 also outputs the reliability of the estimation result. The size detection unit 103 detects the size of the range of defects in the estimation result of the reliability equal to or greater than the predetermined threshold as the size of defects generated in the captured glass sheet.

According to this, since the size of the range in the highly reliable estimation result is detected as the size of the defect, even if the position and the range are obtained as the estimation result for the portion that is not the defect, the possibility of erroneously detecting the size of the range as the size of the defect can be reduced.

In particular, the irradiation pattern of light at the time of photographing may be changed due to a deviation in the conveying direction of the glass plate or the like, and color unevenness may occur in a portion other than the defect. For such a portion where color unevenness occurs, the size detection model 112 may misjudge as a defective portion, and by appropriately setting a threshold value, it is possible to perform reasonable detection by distinguishing a color unevenness portion having no problem in quality of a product from a true defective portion.

[ embodiment 3 ]

When a plurality of estimation results are obtained, the size detection unit 103 of the present embodiment detects the size of a region including a plurality of ranges in the plurality of estimation results as the size of a defect generated in the captured glass sheet.

Fig. 7 is a diagram showing an outline of the defect size detection performed by the size detection unit 103 according to the present embodiment. The captured image 33 shown in fig. 7 is a diagram in which rectangles 45 to 48 indicating the estimation result of the size detection model 112 are drawn from a captured image obtained by capturing an end surface of a glass plate including a defective portion.

As described above, the size detection unit 103 according to the present embodiment obtains the size of the smallest area including a plurality of defective portions when the areas are detected in one captured image. For example, the size detection unit 103 may specify the uppermost side of the sides constituting the rectangles 45 to 48 and the lowermost side of the rectangles 45 to 48, and set the distance between these sides to the height of the obtained region. The size detection unit 103 may specify the leftmost side of the sides constituting the rectangles 45 to 48 and the rightmost side of the rectangles 45 to 48, and set the distance between these sides to the width of the obtained region.

The size detection unit 103 can specify the containing region 61 containing the rectangles 45 to 48 by this processing. Then, the size detection unit 103 transmits and stores size information indicating the width and height of the containing region 61 to the detection result storage device 4.

In the example of fig. 7, the containing regions 61 including the rectangles 45 to 48 that are in contact with each other are specified, but the present application is not limited to this. That is, the size detection unit 103 may detect the size of the region including a plurality of areas not in contact with each other as the size of the defect.

This configuration is effective, for example, when only both ends of an actual defect are inferred to be defects, and the central portion of the defect is not inferred to be defects. In this case, the size detection unit 103 can detect the size of the inclusion region having a small error from the actual size of the defect as the size of the defect by specifying the inclusion region including the range of both the detected ends.

(flow of inspection Process)

Fig. 8 is a flowchart showing an example of the inspection process according to the present embodiment. In this flowchart, the same processing as in the flowchart of fig. 4 is given the same reference numerals as in fig. 4. In addition, the same processing as the flowchart of fig. 6 is given the same reference numerals as in fig. 6. For the same processing, a repeated explanation thereof is omitted.

In S21, the size detection unit 103 determines whether or not the number of estimation results obtained in S11 is plural. If a plurality of the inspection steps are determined (yes in S21), the inspection step proceeds to S22. On the other hand, if it is determined that there are not a plurality of the inspection steps (no in S21), the inspection steps proceed to S23.

In S22, the size detection unit 103 obtains the size of the included region including the plurality of ranges indicated by the estimation result obtained in S11, and detects the size as the size of the defect. The method of determining the size of the included region is as described with reference to fig. 7.

In S23, the size detection unit 103 detects the size of the range in the obtained estimation result as the size of the defect. That is, the size detection unit 103 detects the width and the height indicated in the one estimation result obtained in S11 as the size of the defect.

(effects of action)

As described above, in the inspection apparatus 1 according to the present embodiment, when a plurality of estimation results are obtained, the size detection unit 103 detects the size of the included region including a plurality of ranges among the plurality of estimation results as the size of the defect generated in the captured glass sheet.

According to this configuration, since the size of the included region is detected as the size of the defect, even when a plurality of ranges having a large error from the actual defect size are obtained, the plurality of ranges can be corrected to the included region having a small error from the actual defect size. As a result, even when a plurality of ranges are obtained, the error between the size detection and the actual size of the defect using the size detection model 112 can be reduced.

In addition, even if a plurality of defects are detected in the glass sheet, if the size of each detected defect is small, the inspection apparatus 1 may finally determine the glass sheet as good. In the case where the size of the actual defect is small, the judgment result is reasonable, but in the case where a defect of a large size is actually present and a part thereof is detected as a broken defect, the judgment result is misjudgment. According to the configuration of the present embodiment, since the size of the region including the range including the plurality of defects is detected, the possibility of occurrence of erroneous judgment as described above can be reduced.

(modification)

This embodiment mode can be combined with embodiment mode 2. Specifically, when there are a plurality of estimation results whose reliability is equal to or higher than a predetermined threshold, the size detection unit 103 may specify an inclusion region including a range in the estimation results, and detect the size of the inclusion region as the size of a defect generated in the captured glass sheet.

[ embodiment 4 ]

When the size of the detected defect is out of the normal range, the size detection unit 103 of the present embodiment re-detects the size of the defect by performing numerical analysis on the pixel values of the pixels constituting the captured image. The normal range may be set in advance based on, for example, the size of the glass plate, the size of a general defect, and the like. The size of the defect outside the normal range may be, for example, a size satisfying at least one of (1) a width of the rectangle as the result of the inference being outside a prescribed first numerical range and (2) a height of the rectangle as the result of the inference being outside a prescribed second numerical range.

The size detection unit 103 of the present embodiment also performs size detection based on the estimation result output from the size detection model 112, similarly to the size detection unit 103 of each of the above embodiments. The size detection unit 103 of the present embodiment is different from the size detection unit 103 of each embodiment in determining whether the detected size is within a normal range.

When the detected size is out of the normal range, the size detection unit 103 of the present embodiment re-detects the size of the defect by performing numerical analysis on the pixel values of the pixels constituting the captured image. In this respect, the size detection unit 103 of the present embodiment is also different from the size detection unit 103 of each of the above embodiments. The processing block for performing the numerical analysis may be set to a processing block different from the size detection unit 103.

The method for detecting the size of the defect by numerical analysis is not particularly limited, and various methods can be applied. For example, as shown in fig. 2 and the like, in a captured image, pixel values of a boundary portion between an area in which an end surface of a glass plate is captured and a background area thereof are greatly changed. Therefore, the size detection unit 103 can first extract a region in which the end surface of the glass plate is photographed based on the change in the pixel value.

As shown in fig. 2, the end face of the glass plate on which the shot image is shot has different pixel values between the shot defect portion and the non-defect portion. Therefore, if the region including the unique pixel value is included in the region of the end face of the glass plate extracted as described above, the size detection unit 103 may detect the width and the height of the region as the size of the defect. After that, the size detection unit 103 transmits size information indicating the detected width and height to the detection result storage device 4 and stores the same.

(flow of inspection Process)

Fig. 9 is a flowchart showing an example of the inspection process according to the present embodiment. Note that, in this flowchart, the same processing as in the flowchart of fig. 4 is given the same reference numerals as in fig. 4, and the repeated description thereof is omitted.

In S31, the size detection unit 103 determines whether the size of the defect detected in S4 is within the normal range. If it is determined that the detection result is within the normal range (yes in S31), the inspection process proceeds to S5. If it is determined that the detection result is not within the normal range, that is, outside the normal range (no in S31), the inspection process proceeds to S32.

In S32, the size detection unit 103 discards the detection result of the size based on the estimation result output from the size detection model 112, and in S33, the size detection unit 103 re-detects the size of the defect by performing numerical analysis on the pixel values of the respective pixels constituting the captured image. After that, the inspection process advances to S5.

In S5 shifted from S33, the size detection unit 103 may store the information indicating that the re-detection has been performed in the detection result storage device 4 in association with the size information. Thus, the user of the inspection system 100 can specify a captured image to be re-detected.

In addition, S32 may be omitted. In this case, in S5, the size detection unit 103 may store size information indicating the detection result of S4 and size information indicating the detection result of S33 in the detection result storage device 4.

(effects of action)

As described above, in the inspection apparatus 1 of the present embodiment, when the size of the detected defect is out of the normal range, the size detection unit 103 re-detects the size of the defect by performing numerical values on the pixel values of the respective pixels constituting the captured image.

According to this configuration, the size of the defect whose size is out of the normal range is re-detected by performing numerical analysis on the pixel values of the pixels constituting the captured image. That is, since the re-detection of the size is performed by a method different from the size detection of the defect using the trained model, the size can be corrected to an appropriate size.

[ modification ]

In the above embodiments, the example of detecting the size of the defect generated in the end face of the glass plate has been described, but the size of the defect generated in the portion other than the end face of the glass plate may be detected. In the above embodiments, the example of detecting the size of the defect generated in the glass sheet cut from the glass raw sheet has been described, but the glass sheet to be subjected is not limited to the glass sheet cut from the glass raw sheet. For example, it is also possible to detect the size of a defect generated in a glass plate portion in a product including a glass plate, or the like.

In the above embodiments, the example of determining the presence or absence (or type) of a defect and detecting the size was described using one inspection apparatus, but these processes may be performed using other apparatuses. That is, the inspection method described in the above embodiments may be performed by one inspection apparatus or may be performed by a plurality of inspection apparatuses.

In the above embodiments, the defect determination model 111 and the size detection model 112 were set as different models, but one model may be used in which the presence or absence of a defect and the position and range of a defect are learned. In this case, if a captured image is input to one of the models, an estimation result indicating the presence or absence of a defect is output. If there is a defect, the estimation result indicating the position and the range is output together. In addition, the model may be used to infer the type of defect.

(software-based implementation example)

The function of the inspection apparatus 1 (hereinafter referred to as "apparatus") can be realized by a program (inspection program) for causing a computer to function as the apparatus, and for causing the computer to function as each control block (in particular, each portion included in the control portion 10) of the apparatus.

In this case, as hardware for executing the above-described program, the above-described apparatus may include a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory). The functions described in the above embodiments are realized by executing the program by the control device and the storage device.

The above-described program may be stored in one or more storage media readable by a computer, instead of being temporary. The storage medium may or may not belong to the above-described apparatus. In the latter case, the program may be provided to the apparatus via any of wired or wireless transmission media.

Further, some or all of the functions of the control blocks may be realized by logic circuits. For example, an integrated circuit in which logic circuits functioning as the control blocks described above are formed is also within the scope of the present application. In addition, the functions of the control blocks described above may be realized by, for example, a quantum computer.

The present application is not limited to the above embodiments, and various modifications can be made within the scope shown in the claims, and embodiments obtained by appropriately combining the technical means disclosed in the respective different embodiments also fall within the technical scope of the present application.

Description of the reference numerals

1: inspection apparatus

101: image acquisition unit

103: size detecting unit

112: size detection model

Claims (9)

1. An inspection apparatus, comprising:

an image acquisition unit that acquires a captured image obtained by capturing a glass plate; and

and a size detection unit that detects, as a size of a defect generated in the glass plate, a size of a range in an estimation result obtained by inputting the captured image to a trained model in which a position and a range of the defect generated in the glass plate are learned to estimate the position and the range.

2. The inspection apparatus according to claim 1, wherein,

the trained model also outputs reliability for the inferred results,

the size detection unit detects the size of the range in the estimation result in which the reliability is equal to or greater than a predetermined threshold as the size of the defect generated in the glass sheet.

3. The inspection apparatus according to claim 1 or 2, wherein,

when a plurality of the estimation results are obtained, the size detection unit detects the size of a region including a plurality of ranges in the plurality of estimation results as the size of the defect generated in the glass plate that is photographed.

4. An inspection apparatus according to any one of claims 1 to 3, wherein,

in the case where the size of the detected defect is out of the normal range, the size detecting section re-detects the size of the defect by performing numerical analysis on pixel values of respective pixels constituting the captured image.

5. The inspection apparatus according to any one of claims 1 to 4, wherein,

the image acquisition unit acquires the captured image obtained by capturing an image of an end surface of the glass plate.

6. The inspection apparatus according to claim 1, wherein,

the size detection unit detects the size of the range in the estimation result in which the amount of change in the pixel value on the boundary between the inside and outside of the range indicated by the estimation result is equal to or greater than a predetermined threshold as the size of the defect generated in the glass plate.

7. An inspection method performed by an inspection apparatus, comprising:

an image acquisition step of acquiring a photographed image obtained by photographing the glass plate; and

and a size detection step of detecting, as a size of a defect generated in the glass plate, a size of a range in an estimation result obtained by inputting the captured image to a trained model in which a position and a range of the defect generated in the glass plate are learned to estimate the position and the range.

8. A method for manufacturing a glass plate, comprising a step of molding a glass raw plate into a glass plate of a predetermined size; and an inspection process of the glass sheet performed by an inspection device, wherein,

the inspection process includes:

an image acquisition step of acquiring a photographed image obtained by photographing the glass plate;

and a size detection step of detecting, as a size of a defect generated in the glass plate, a size of a range in an estimation result obtained by inputting the captured image to a trained model in which a position and a range of the defect generated in the glass plate are learned to estimate the position and the range.

9. An inspection program for causing a computer to function as the inspection apparatus according to claim 1, wherein,

the inspection program is for causing a computer to function as the image acquisition section and the size detection section.