Glass substrate detection method, device, electronic equipment and storage medium

Abstract

The present invention provides a method and system for detecting a glass substrate, the method comprising: obtaining a first light image and a second light image of a glass substrate to be detected; fusing the first light image with the second light image to obtain a fused light image; performing defect detection on the fused light image through a pre-trained defect detection network, and outputting a detection result of the glass substrate to be detected. The fusion can be performed on the basis of the first light image and the second light image of the glass substrate to be detected to obtain a fused light image of the glass substrate to be detected, and then the fused light image is detected through image detection technology to obtain a detection result of the glass substrate to be detected, which can improve the detection accuracy of the glass substrate.

Description

Technical Field

[0001] The present invention relates to the field of glass substrate detection, and in particular to a glass substrate detection method, device, electronic equipment and storage medium. Background Art

[0002] Glass substrates are commonly used optical materials. The quality of glass substrates directly affects the quality of subsequent products. Common defects of glass substrates include cracks, breakage, bubbles, etc. The existing glass substrate detection method mainly uses photoelectric sensors for detection. However, photoelectric sensors can only detect larger defects and cannot accurately detect smaller defects. Therefore, the existing glass substrate detection accuracy is low, which affects the product yield. Summary of the invention

[0003] An embodiment of the present invention provides a method for detecting a glass substrate, which can fuse a first light image and a second light image of a glass substrate to be detected to obtain a fused light image of the glass substrate to be detected, and then detect the fused light image through image detection technology to obtain a detection result of the glass substrate to be detected, thereby improving the detection accuracy of the glass substrate.

[0004] In a first aspect, an embodiment of the present invention provides a method for detecting a glass substrate, the method comprising the following steps:

[0005] Acquire a first light image and a second light image of a glass substrate to be inspected, wherein the first light image is a light transmission image of a first side of the glass substrate to be inspected, and the second light image is a light transmission image of a second side of the glass substrate to be inspected, wherein the first side and the second side are opposite sides of the glass substrate to be inspected;

[0006] Fusing the first light image with the second light image to obtain a fused light image;

[0007] Defect detection is performed on the fused light image through a pre-trained defect detection network, and a detection result of the glass substrate to be detected is output.

[0008] Furthermore, the step of obtaining the first light image of the glass substrate to be inspected includes:

[0009] In a dark room, the glass substrate to be inspected is first irradiated with a first parallel light arranged on a first side of the glass substrate to be inspected;

[0010] In the first irradiation process, the light transmitted through the glass substrate to be inspected is digitally collected to obtain the first light image.

[0011] Furthermore, in the first irradiation process, a first filter is provided on the second side of the glass substrate to be inspected, and the step of first irradiating the glass substrate to be inspected with a first parallel light provided on one side of the glass substrate to be inspected comprises:

[0012] The glass substrate to be inspected is first irradiated at a first angle by a first parallel light arranged on one side of the glass substrate to be inspected, the first parallel light is perpendicular to the plane of the first filter, the first parallel light has a first wavelength, and the first filter allows light of the first wavelength to pass through.

[0013] In the first irradiation process, the step of digitally collecting the first parallel light passing through the glass substrate to be inspected to obtain the first light image includes:

[0014] During the first irradiation process, the light passing through the first filter is digitally collected to obtain the first light image.

[0015] Furthermore, the step of obtaining the second light image of the glass substrate to be inspected includes:

[0016] In a dark room, performing a second irradiation on the glass substrate to be inspected by a second parallel light arranged on a second side of the glass substrate to be inspected, wherein the first irradiation and the second irradiation are performed separately;

[0017] During the second irradiation process, the light transmitted through the glass substrate to be inspected is digitally collected to obtain the second light image.

[0018] Furthermore, in the second irradiation process, a second filter is provided on the first side of the glass substrate to be inspected, and the step of performing a second irradiation on the glass substrate to be inspected by a second parallel light provided on the second side of the glass substrate to be inspected comprises:

[0019] The glass substrate to be inspected is irradiated with a second light at a second angle by a second parallel light arranged on a second side of the glass substrate to be inspected, the second parallel light is perpendicular to the plane of the second filter, the second parallel light has a second wavelength, and the second filter allows light of the second wavelength to pass through.

[0020] In the second irradiation process, the step of digitally collecting the second parallel light passing through the glass substrate to be inspected to obtain the second light image includes:

[0021] During the second irradiation process, the light passing through the second filter is digitally collected to obtain the second light image.

[0022] Furthermore, the first light image and the second light image have the same size and resolution, and fusing the first light image and the second light image to obtain a fused light image includes:

[0023] aligning the first light pattern with the second light pattern;

[0024] Adding the light element value of each light element in the first light image to the light element value of the corresponding light element in the second light image to obtain a first fused image;

[0025] averaging the light element value of each light element in the first light image and the light element value of the corresponding light element in the second light image to obtain a second fused image;

[0026] Calculate the difference between the light element value of each light element in the first light image and the light element value of the corresponding light element in the second light image to obtain a third fused image;

[0027] The first fusion image, the second fusion image, and the third fusion image are superimposed as channels to obtain the fusion light image.

[0028] Furthermore, the glass substrate detection method further includes:

[0029] Constructing a defect detection network based on convolutional neural network technology, wherein the defect detection network includes an input layer, a convolutional layer, and an output layer;

[0030] Constructing a training data set, wherein the training data set includes a sample fused light image, and the sample fused light image and the fused light image are fused by the same fusion method;

[0031] The defect detection network is trained according to the training data set to obtain the pre-trained defect detection network.

[0032] In a second aspect, an embodiment of the present invention further provides a device for detecting a glass substrate, the device comprising:

[0033] an acquisition module, configured to acquire a first light image and a second light image of a glass substrate to be inspected, wherein the first light image is a light transmission image of a first side of the glass substrate to be inspected, and the second light image is a light transmission image of a second side of the glass substrate to be inspected, wherein the first side and the second side are opposite sides of the glass substrate to be inspected;

[0034] an integration module, used for fusing the first light image with the second light image to obtain a fused light image;

[0035] The detection module is used to perform defect detection on the fused light image through a pre-trained defect detection network, and output a detection result of the glass substrate to be detected.

[0036] In a third aspect, an embodiment of the present invention further provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps in the glass substrate detection method when executing the computer program.

[0037] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps in the method for detecting a glass substrate are implemented.

[0038] In an embodiment of the present invention, a first light image and a second light image of a glass substrate to be detected are obtained, wherein the first light image is a light passing image of a first side of the glass substrate to be detected, and the second light image is a light passing image of a second side of the glass substrate to be detected, and the first side and the second side are opposite sides of the glass substrate to be detected; the first light image and the second light image are fused to obtain a fused light image; a defect detection is performed on the fused light image through a pre-trained defect detection network, and the detection result of the glass substrate to be detected is output. The fusion can be performed on the basis of the first light image and the second light image of the glass substrate to be detected to obtain a fused light image of the glass substrate to be detected, and then the fused light image is detected through image detection technology to obtain the detection result of the glass substrate to be detected, which can improve the detection accuracy of the glass substrate. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 is a flow chart of a method for detecting a glass substrate provided by an embodiment of the present invention;

[0041] Figure 2 is a flow chart of a pre-trained defect detection network method provided by an embodiment of the present invention;

[0042] Figure 3 is a structural schematic diagram of a glass substrate detection device provided by an embodiment of the present invention;

[0043] Figure 4 is a structural diagram of an acquisition module provided by an embodiment of the present invention;

[0044] Figure 5 is a structural diagram of another acquisition module provided by an embodiment of the present invention;

[0045] Figure 6 is a structural schematic diagram of a first integration module provided by an embodiment of the present invention;

[0046] Figure 7 Schematic diagram of another glass substrate detection device provided by an embodiment of the present invention

[0047] Figure 8 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] In the present invention, a method for detecting a glass substrate is proposed, which can fuse the first light image and the second light image of the glass substrate to be detected to obtain a fused light image of the glass substrate to be detected, and then detect the fused light image through image detection technology to obtain a detection result of the glass substrate to be detected, which can improve the detection accuracy of the glass substrate.

[0050] See also Figure 1 , Figure 1 is a flow chart of a method for detecting a glass substrate provided by an embodiment of the present invention, such as Figure 1 As shown, the following steps are included:

[0051] 101. Obtain a first light image and a second light image of a glass substrate to be inspected.

[0052] In an embodiment of the present invention, the first light image is a light transmission image of the first side of the glass substrate to be detected, and the second light image is a light transmission image of the second side of the glass substrate to be detected. The first side and the second side are opposite sides of the glass substrate to be detected. Specifically, the light transmission image refers to a light image after the light passes through the glass substrate to be detected. The light image includes light elements. The light elements are similar to pixels in an image. Each light element in the light image corresponds to a light element value. The light element value is the intensity of light received by the light element point. The light intensity can be the luminous flux per unit area. For example, the light element value of a light element is the luminous flux of the light element unit.

[0053] The above-mentioned first light pattern is obtained in a dark room by first irradiating the above-mentioned glass substrate to be inspected with a first parallel light set on the first side of the above-mentioned glass substrate to be inspected. During the first irradiation process, the light passing through the above-mentioned glass substrate to be inspected is digitally collected, and the luminous flux on each light element point is collected to obtain the above-mentioned first light pattern.

[0054] The second light pattern is obtained by irradiating the glass substrate to be inspected with a second parallel light arranged on the second side of the glass substrate to be inspected in a dark room, and the first irradiation is performed separately from the second irradiation. During the second irradiation, the light passing through the glass substrate to be inspected is digitally collected, and the luminous flux at each light element point is collected to obtain the second light pattern.

[0055] The light energies of the first parallel light source and the second parallel light source are stable, that is, when the first parallel light source and the second parallel light source are measured, the light energies at each position are the same.

[0056] In a possible embodiment, the first parallel light and the second parallel light have the same irradiation parameters, that is, the first parallel light and the second parallel light respectively irradiate the first side and the second side of the glass substrate to be inspected with the same irradiation parameters.

[0057] It should be noted that if the quality of the glass substrate to be inspected is good, the first light image and the second light image obtained by the same irradiation parameters should be extremely similar in light element distribution.

[0058] 102. Fuse the first light map with the second light map to obtain a fused light map.

[0059] In an embodiment of the present invention, the first optical image and the second optical image have the same size and resolution. The above-mentioned fusing of the first optical image and the second optical image to obtain a fused optical image includes: aligning the first optical image with the second optical image; adding the optical element value of each optical element in the first optical image with the optical element value of the corresponding optical element in the second optical image to obtain a first fused image; averaging the optical element value of each optical element in the first optical image with the optical element value of the corresponding optical element in the second optical image to obtain a second fused image; performing difference calculation between the optical element value of each optical element in the first optical image and the optical element value of the corresponding optical element in the second optical image to obtain a third fused image; and superimposing the first fused image, the second fused image, and the third fused image as channels to obtain the fused optical image.

[0060] In the first fusion image, the addition of light element values ​​can amplify the global information of the first and second light images, so that the extracted implicit information can better express the defect characteristics. In the second fusion image, the averaging of light element values ​​can eliminate the interference caused by noise to a certain extent, which can assist the first fusion image to more accurately extract the defect characteristics. In the third fusion image, the difference of light element values ​​can amplify the local difference information, thereby assisting the first fusion image to more accurately extract the defect characteristics.

[0061] 103. Defect detection is performed on the fused light image through a pre-trained defect detection network, and the detection result of the glass substrate to be detected is output.

[0062] In an embodiment of the present invention, the defect detection network may be a convolutional neural network, and by training the convolutional neural network, a defect detection network capable of detecting the glass substrate may be obtained. Specifically, the convolutional neural network is used to extract implicit features of the fused light image, and then linear regression is performed on the implicit features to obtain a detection result.

[0063] Optionally, in a possible embodiment, the first parallel light and the second parallel light have the same irradiation parameters, that is, the first parallel light and the second parallel light respectively irradiate the first side and the second side of the glass substrate to be inspected with the same irradiation parameters. A defect detection network can be constructed based on convolutional neural network technology, and the defect detection network includes an input layer, a convolution layer, and an output layer.

[0064] A defect detection network is constructed based on convolutional neural network technology, and the defect detection network includes an input layer, a convolution layer and an output layer; a training data set is constructed, and the training data set includes a sample fusion light image, and the sample fusion light image and the fusion light image are fused by the same fusion method; the defect detection network is trained according to the training data set to obtain the pre-trained defect detection network.

[0065] The sample fused light images include positive samples and negative samples. The positive sample is a sample fused light image obtained by fusing the first light image of a glass substrate without defects with the second light image. The negative sample is a sample fused light image obtained by fusing the first light image of a glass substrate with defects with the second light image. The defect detection network is trained by the sample fused light images. During the training, the parameters of the defect detection network are continuously adjusted so that the defect detection network can classify the sample fused light images by the trained parameters. The classification results may include defective and non-defective.

[0066] In an embodiment of the present invention, a first light image and a second light image of a glass substrate to be detected are obtained, wherein the first light image is a light passing image of a first side of the glass substrate to be detected, and the second light image is a light passing image of a second side of the glass substrate to be detected, and the first side and the second side are opposite sides of the glass substrate to be detected; the first light image and the second light image are fused to obtain a fused light image; a defect detection is performed on the fused light image through a pre-trained defect detection network, and the detection result of the glass substrate to be detected is output. The fusion can be performed on the basis of the first light image and the second light image of the glass substrate to be detected to obtain a fused light image of the glass substrate to be detected, and then the fused light image is detected through image detection technology to obtain the detection result of the glass substrate to be detected, which can improve the detection accuracy of the glass substrate.

[0067] Optionally, in an embodiment of the present invention, during the first irradiation process, a first filter is provided on the second side of the glass substrate to be inspected, and the step of performing a first irradiation on the glass substrate to be inspected by using a first parallel light provided on one side of the glass substrate to be inspected includes: performing a first irradiation on the glass substrate to be inspected at a first angle by using a first parallel light provided on one side of the glass substrate to be inspected, the first parallel light being perpendicular to the plane of the first filter, the first parallel light having a first wavelength, and the first filter allowing light of the first wavelength to pass through.

[0068] The first angle may be any angle value between 30 degrees and 60 degrees. In the embodiment of the present invention, the first angle is preferably between 40 degrees and 50 degrees. The first angle is used as the incident angle of the first parallel light. When there are defects in the glass substrate, the refraction angle of the first parallel light will be changed, so that the corresponding light element position in the first light image corresponding to the first parallel light does not receive light information, resulting in a light element value of 0 or null.

[0069] The first filter allows light within the first wavelength range to pass through and can filter out stray light. Arranging the first parallel light and the first filter on both sides of the glass substrate to be inspected can better filter out stray light of different wavelengths passing through the glass substrate to be inspected and reduce the noise of the first light pattern.

[0070] Furthermore, during the first irradiation process, the first parallel light passing through the glass substrate to be inspected is digitally collected to obtain the first light pattern, including: during the first irradiation process, the light passing through the first filter is digitally collected to obtain the first light pattern.

[0071] It should be noted that the above-mentioned digital acquisition refers to the acquisition of the numerical value of the light intensity, that is, the luminous flux data is used as the light element value, so that the light image can be digitally processed by a computer.

[0072] Optionally, in an embodiment of the present invention, during the second irradiation process, a second filter is provided on the first side of the glass substrate to be inspected, and the step of performing second irradiation on the glass substrate to be inspected by using the second parallel light provided on the second side of the glass substrate to be inspected includes: performing second irradiation on the glass substrate to be inspected at a second angle by using the second parallel light provided on the second side of the glass substrate to be inspected, the second parallel light is perpendicular to the plane of the second filter, the second parallel light has a second wavelength, and the second filter allows light of the second wavelength to pass through.

[0073] The second angle may be any angle value between 30 degrees and 60 degrees. In the embodiment of the present invention, the second angle is preferably between 40 degrees and 50 degrees. As the incident angle of the second parallel light, the second angle may change the refraction angle of the second parallel light when there are defects in the glass substrate, so that the corresponding light element position in the second light image corresponding to the second parallel light does not receive light information, resulting in a light element value of 0 or null.

[0074] The second filter allows light within the second wavelength range to pass through and can filter out stray light. Arranging the second parallel light and the second filter on both sides of the glass substrate to be inspected can better filter out stray light of different wavelengths passing through the glass substrate to be inspected and reduce the noise of the second light pattern.

[0075] Furthermore, during the second irradiation process, the step of digitally collecting the second parallel light passing through the glass substrate to be inspected to obtain the second light pattern includes: during the second irradiation process, digitally collecting the light passing through the second filter to obtain the second light pattern.

[0076] Optionally, the first angle and the second angle may have the same angle value.

[0077] Optionally, in another possible embodiment, the first angle and the second angle may be different angles, and the angle value of the first angle plus the second angle is 90 degrees, and the first angle and the second angle are both angles with reference to the normal of the glass substrate to be detected.

[0078] In an embodiment of the present invention, a fusion can be performed on the basis of the first light image and the second light image of the glass substrate to be inspected to obtain a fused light image of the glass substrate to be inspected, and then the fused light image is detected by image detection technology to obtain a detection result of the glass substrate to be inspected, which can improve the detection accuracy of the glass substrate.

[0079] See also Figure 2 , Figure 2 is a flow chart of a pre-trained defect detection network method provided by an embodiment of the present invention, such as Figure 2 As shown, including:

[0080] 201. A defect detection network is constructed based on convolutional neural network technology. The defect detection network includes an input layer, a convolutional layer and an output layer.

[0081] In an embodiment of the present invention, the defect detection network can effectively extract defect features, and the defect detection network includes: an input layer, a convolution layer, and an output layer. The defect detection network can be implemented by convolutional neural network technology, and it can also be understood that the defect detection network is a convolutional neural network. The input layer is used to input the fused light image, the convolution layer is used to extract implicit information in the fused light image, and the output layer is used to output the classification results, and the classification results include defective and non-defective.

[0082] 202. Build a training dataset.

[0083] In the embodiment of the present invention, the training data set includes a sample fused light image, and the sample fused light image and the fused light image are fused by the same fusion method.

[0084] In an embodiment of the present invention, the sample fused light image refers to a fused light image used as a sample. The training data set includes the sample fused light image and the corresponding annotation information, and the sample fused light image and the fused light image are fused by the same fusion method. The fused light image is a fused light image obtained by fusing the first light image with the second light image.

[0085] The above-mentioned sample fusion light image includes a positive sample and a negative sample. The positive sample is a sample fusion light image obtained by fusing the first sample light image of a defective glass substrate with the second sample light image, and the negative sample is a sample fusion light image obtained by fusing the third sample light image of a defective glass substrate with the fourth sample light image.

[0086] The first sample light pattern is obtained by first irradiating the defect-free glass substrate with a first parallel light arranged on a first side of the defect-free glass substrate in a dark room. During the first irradiation, light passing through the defect-free glass substrate is digitally collected to obtain the first sample light pattern.

[0087] The second sample light pattern is obtained by performing a second irradiation on the defect-free glass substrate in a dark room by a second parallel light arranged on the second side of the defect-free glass substrate, and the first irradiation is performed separately from the second irradiation. During the second irradiation, the light passing through the defect-free glass substrate is digitally collected to obtain the second sample light pattern.

[0088] The third sample light pattern is obtained by first irradiating the defective glass substrate with a first parallel light set on a first side of the defective glass substrate in a dark room. During the first irradiation, light passing through the defective glass substrate is digitally collected to obtain the third sample light pattern.

[0089] The fourth sample light pattern is obtained by performing a second irradiation on the defective glass substrate in a dark room by a second parallel light arranged on the second side of the defective glass substrate, and the first irradiation is performed separately from the second irradiation. During the second irradiation, the light passing through the defective glass substrate is digitally collected to obtain the fourth sample light pattern.

[0090] The first sample light image, the second sample light image, the third sample light image, and the fourth sample light image have the same size and resolution.

[0091] The first sample light image and the second sample light image may be fused to obtain a fused light image of a positive sample, specifically comprising: aligning the first sample light image and the second sample light image; adding the light element value of each light element in the first sample light image with the light element value of the corresponding light element in the second sample light image to obtain a first sample fused image; averaging the light element value of each light element in the first sample light image with the light element value of the corresponding light element in the second sample light image to obtain a second sample fused image; performing difference calculation between the light element value of each light element in the first sample light image and the light element value of the corresponding light element in the second sample light image to obtain a third sample fused image; and superimposing the first sample fused image, the second sample fused image and the third sample fused image as channels to obtain the fused light image of the positive sample.

[0092] Similarly, the third sample light map and the fourth sample light map can be fused to obtain a fused light map of the negative sample, specifically including: aligning the third sample light map and the fourth sample light map; adding the light element value of each light element in the third sample light map with the light element value of the corresponding light element in the fourth sample light map to obtain a fourth sample fused map; averaging the light element value of each light element in the third sample light map with the light element value of the corresponding light element in the fourth sample light map to obtain a fifth sample fused map; performing difference calculation between the light element value of each light element in the third sample light map and the light element value of the corresponding light element in the fourth sample light map to obtain a sixth sample fused map; superimposing the fourth sample fused map, the fifth sample fused map and the sixth sample fused map as channels to obtain the fused light map of the negative sample.

[0093] 203. Train the defect detection network according to the training data set to obtain a pre-trained defect detection network.

[0094] In an embodiment of the present invention, the above-mentioned defect detection network can be trained through sample fusion light images, and the parameters of the defect detection network are continuously adjusted during the training, so that the defect detection network is adjusted to classify the sample fusion light images through the trained parameters, and the classification results may include defects and no defects. The training process includes a forward process and a reverse process. The forward process classifies the sample fusion light images to obtain a forward classification result. The forward classification result and the corresponding annotation information are subjected to error calculation to obtain the error loss. In the reverse process, back propagation is performed according to the error loss to adjust the parameters of the defect detection network until the preset number of iterations is completed or the error loss of the defect detection network converges to a preset value. The training can be stopped to obtain a trained defect detection network.

[0095] In an embodiment of the present invention, a defect detection network is constructed based on convolutional neural network technology, and the defect detection network includes an input layer, a convolutional layer, and an output layer; a training data set is constructed, and the training data set includes a sample fusion light image, and the sample fusion light image and the fusion light image are fused by the same fusion method; the defect detection network is trained according to the training data set to obtain a pre-trained defect detection network. A defect detection network is constructed by convolutional neural network technology, a training data set is constructed, and then the defect detection network is trained according to the training data set to obtain a pre-trained defect detection network.

[0096] It should be noted that the glass substrate detection method provided in the embodiment of the present invention can be applied to devices such as smart phones, computers, servers, etc. that can perform data query.

[0097] Optional, see Figure 3 , Figure 3 is a schematic diagram of the structure of a glass substrate detection device provided by an embodiment of the present invention, such as Figure 3 As shown, the device comprises:

[0098] An acquisition module 301 is used to acquire a first light image and a second light image of a glass substrate to be inspected, wherein the first light image is a light transmission image of a first side of the glass substrate to be inspected, and the second light image is a light transmission image of a second side of the glass substrate to be inspected, wherein the first side and the second side are opposite sides of the glass substrate to be inspected;

[0099] A first integration module 302, configured to fuse the first light image with the second light image to obtain a fused light image;

[0100] The detection module 303 is used to perform defect detection on the fused light image through a pre-trained defect detection network, and output a detection result of the glass substrate to be detected.

[0101] Further, such as Figure 4As shown, the acquisition module 301 includes:

[0102] A first irradiation unit 3011 is used to perform a first irradiation on the glass substrate to be inspected in a dark room by using a first parallel light arranged on a first side of the glass substrate to be inspected;

[0103] The first collection unit 3012 is used to digitally collect the light passing through the glass substrate to be inspected during the first irradiation process to obtain the first light image.

[0104] Furthermore, during the first irradiation process, a first filter is provided on the second side of the glass substrate to be inspected, and the first irradiation unit 3011 is further used to perform a first irradiation on the glass substrate to be inspected at a first angle using a first parallel light provided on one side of the glass substrate to be inspected, wherein the first parallel light is perpendicular to the plane of the first filter, the first parallel light has a first wavelength, and the first filter allows light of the first wavelength to pass through.

[0105] The first acquisition unit 3012 is further used to digitally acquire the light passing through the first filter during the first irradiation process to obtain the first light image.

[0106] Further, such as Figure 5 As shown, the acquisition module 301 also includes:

[0107] A second irradiation unit 3013 is used to perform a second irradiation on the glass substrate to be inspected in a dark room by using a second parallel light arranged on a second side of the glass substrate to be inspected, wherein the first irradiation and the second irradiation are performed separately;

[0108] The second collection unit 3014 is used for digitally collecting the light passing through the glass substrate to be inspected during the second irradiation process to obtain the second light image.

[0109] Furthermore, in the second irradiation process, a second filter is provided on the first side of the glass substrate to be inspected, and the second irradiation unit 3013 is further used to perform a second irradiation on the glass substrate to be inspected at a second angle using a second parallel light provided on the second side of the glass substrate to be inspected, the second parallel light is perpendicular to the plane of the second filter, the second parallel light has a second wavelength, and the second filter allows light of the second wavelength to pass through.

[0110] The second acquisition unit 3014 is further used to digitally acquire the light passing through the second filter during the second irradiation process to obtain the second light image.

[0111] Further, such as Figure 6As shown, the first light image and the second light image have the same size and resolution, and the first integration module 302 includes:

[0112] An alignment unit 3021, configured to align the first light image with the second light image;

[0113] A first calculation unit 3022, configured to add the light element value of each light element in the first light image to the light element value of the corresponding light element in the second light image to obtain a first fused image;

[0114] A second calculation unit 3023 is used to average the light element value of each light element in the first light image and the light element value of the corresponding light element in the second light image to obtain a second fused image;

[0115] A third calculation unit 3024 is used to perform difference calculation between the light element value of each light element in the first light image and the light element value of the corresponding light element in the second light image to obtain a third fused image;

[0116] The superposition unit 3025 is used to superimpose the first fusion image, the second fusion image and the third fusion image as channels to obtain the fusion light image.

[0117] Further, such as Figure 7 As shown, the device also includes:

[0118] A construction module 304 is used to construct a defect detection network based on a convolutional neural network technology, wherein the defect detection network includes an input layer, a convolutional layer, and an output layer;

[0119] A second integration module 305 is used to construct a training data set, wherein the training data set includes a sample fusion light image, and the sample fusion light image and the fusion light image are fused by the same fusion method;

[0120] The training module 306 is used to train the defect detection network according to the training data set to obtain the pre-trained defect detection network.

[0121] It should be noted that the glass substrate detection device provided in the embodiment of the present invention can be applied to smart phones, computers, servers and other devices that can perform business analysis at the image level.

[0122] The glass substrate detection device provided in the embodiment of the present invention can implement each process implemented by the glass substrate detection method in the above method embodiment and can achieve the same beneficial effects. To avoid repetition, it will not be described here.

[0123] See also Figure 8 , Figure 8 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 8 As shown, it includes: a memory 802, a processor 801, and a computer program of a method for serializing and combining cases stored in the memory 802 and executable on the processor 801, wherein:

[0124] The processor 801 is used to call the computer program stored in the memory 802 and execute the following steps:

[0125] Acquire a first light image and a second light image of a glass substrate to be inspected, wherein the first light image is a light transmission image of a first side of the glass substrate to be inspected, and the second light image is a light transmission image of a second side of the glass substrate to be inspected, wherein the first side and the second side are opposite sides of the glass substrate to be inspected;

[0126] Fusing the first light image with the second light image to obtain a fused light image;

[0127] Defect detection is performed on the fused light image through a pre-trained defect detection network, and a detection result of the glass substrate to be detected is output.

[0128] Further, the step of acquiring the first light image of the glass substrate to be inspected performed by the processor 801 includes:

[0129] In a dark room, the glass substrate to be inspected is first irradiated with a first parallel light arranged on a first side of the glass substrate to be inspected;

[0130] In the first irradiation process, the light transmitted through the glass substrate to be inspected is digitally collected to obtain the first light image.

[0131] Furthermore, in the first irradiation process, a first filter is provided on the second side of the glass substrate to be inspected, and the step of first irradiating the glass substrate to be inspected with a first parallel light provided on one side of the glass substrate to be inspected, performed by the processor 801, includes:

[0132] The glass substrate to be inspected is first irradiated at a first angle by a first parallel light arranged on one side of the glass substrate to be inspected, the first parallel light is perpendicular to the plane of the first filter, the first parallel light has a first wavelength, and the first filter allows light of the first wavelength to pass through.

[0133] In the first irradiation process, the step of digitally collecting the first parallel light passing through the glass substrate to be inspected to obtain the first light image includes:

[0134] During the first irradiation process, the light passing through the first filter is digitally collected to obtain the first light image.

[0135] Further, the step of acquiring the first light image of the glass substrate to be inspected performed by the processor 801 includes:

[0136] In a dark room, performing a second irradiation on the glass substrate to be inspected by a second parallel light arranged on a second side of the glass substrate to be inspected, wherein the first irradiation and the second irradiation are performed separately;

[0137] During the second irradiation process, the light transmitted through the glass substrate to be inspected is digitally collected to obtain the second light image.

[0138] Furthermore, in the second irradiation process, a second filter is provided on the first side of the glass substrate to be inspected, and the step of performing a second irradiation on the glass substrate to be inspected by the processor 801 through the second parallel light provided on the second side of the glass substrate to be inspected includes:

[0139] The glass substrate to be inspected is irradiated with a second light at a second angle by a second parallel light arranged on a second side of the glass substrate to be inspected, the second parallel light is perpendicular to the plane of the second filter, the second parallel light has a second wavelength, and the second filter allows light of the second wavelength to pass through.

[0140] In the second irradiation process, the step of digitally collecting the second parallel light passing through the glass substrate to be inspected to obtain the second light image includes:

[0141] During the second irradiation process, the light passing through the second filter is digitally collected to obtain the second light image.

[0142] Further, the first light image and the second light image have the same size and resolution, and the processor 801 performs the step of fusing the first light image with the second light image to obtain a fused light image, including:

[0143] aligning the first light pattern with the second light pattern;

[0144] Adding the light element value of each light element in the first light image to the light element value of the corresponding light element in the second light image to obtain a first fused image;

[0145] averaging the light element value of each light element in the first light image and the light element value of the corresponding light element in the second light image to obtain a second fused image;

[0146] Calculate the difference between the light element value of each light element in the first light image and the light element value of the corresponding light element in the second light image to obtain a third fused image;

[0147] The first fusion image, the second fusion image, and the third fusion image are superimposed as channels to obtain the fusion light image.

[0148] Furthermore, the glass substrate detection method executed by the processor 801 further includes:

[0149] Constructing a defect detection network based on convolutional neural network technology, wherein the defect detection network includes an input layer, a convolutional layer, and an output layer;

[0150] Constructing a training data set, wherein the training data set includes a sample fused light image, and the sample fused light image and the fused light image are fused by the same fusion method;

[0151] The defect detection network is trained according to the training data set to obtain the pre-trained defect detection network.

[0152] It should be noted that the electronic device provided in the embodiment of the present invention can be applied to devices such as smart phones, computers, servers, etc. that can import data.

[0153] The electronic device provided by the embodiment of the present invention can implement each process implemented by the method for detecting a glass substrate in the above method embodiment, and can achieve the same beneficial effects. To avoid repetition, it will not be described here.

[0154] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the various processes of the glass substrate detection method or the application-end glass substrate detection method provided in the embodiment of the present invention are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0155] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0156] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for detecting a glass substrate, characterized in that: The method comprises the following steps: In a darkroom, a first parallel light is arranged on a first side of the glass substrate to be inspected to perform a first irradiation on the glass substrate to be inspected at a first angle; during the first irradiation process, light passing through the glass substrate to be inspected is digitally collected to obtain a first light image; in a darkroom, a second parallel light is arranged on a second side of the glass substrate to be inspected to perform a second irradiation on the glass substrate to be inspected at a second angle, and the first irradiation and the second irradiation are performed separately; during the second irradiation process, light passing through the glass substrate to be inspected is digitally collected to obtain a second light image; the first light image and the second light image have the same size and resolution; the first side and the second side are opposite sides of the glass substrate to be inspected; the first angle and the second angle are any angle value between 30 degrees and 60 degrees; Align the first light image with the second light image; add the light element value of each light element in the first light image with the light element value of the corresponding light element in the second light image to obtain a first fused image; average the light element value of each light element in the first light image with the light element value of the corresponding light element in the second light image to obtain a second fused image; perform difference calculation between the light element value of each light element in the first light image and the light element value of the corresponding light element in the second light image to obtain a third fused image; superimpose the first fused image, the second fused image and the third fused image as channels to obtain a fused light image; Defect detection is performed on the fused light image through a pre-trained defect detection network, and a detection result of the glass substrate to be detected is output.

2. The method for detecting a glass substrate according to claim 1, wherein: In the first irradiation process, a first filter is arranged on the second side of the glass substrate to be inspected, and the step of first irradiating the glass substrate to be inspected at a first angle with a first parallel light arranged on the first side of the glass substrate to be inspected comprises: The glass substrate to be inspected is first irradiated at a first angle by a first parallel light arranged at one side of the glass substrate to be inspected, the first parallel light is perpendicular to the plane of the first filter, the first parallel light has a first wavelength, and the first filter allows light of the first wavelength to pass through; In the first irradiation process, the step of digitally collecting the first parallel light passing through the glass substrate to be inspected to obtain the first light image includes: During the first irradiation process, the light passing through the first filter is digitally collected to obtain the first light image.

3. The method for detecting a glass substrate according to claim 2, wherein: In the second irradiation process, a second filter is arranged on the first side of the glass substrate to be inspected, and the step of performing a second irradiation on the glass substrate to be inspected at a second angle by a second parallel light arranged on the second side of the glass substrate to be inspected comprises: The glass substrate to be inspected is irradiated with a second light at a second angle by a second parallel light arranged on a second side of the glass substrate to be inspected, wherein the second parallel light is perpendicular to the plane of the second filter, the second parallel light has a second wavelength, and the second filter allows light of the second wavelength to pass through; In the second irradiation process, the step of digitally collecting the second parallel light passing through the glass substrate to be inspected to obtain the second light image includes: During the second irradiation process, the light passing through the second filter is digitally collected to obtain the second light image.

4. The method for detecting a glass substrate according to claim 3, wherein: The glass substrate detection method further includes: Constructing a defect detection network based on convolutional neural network technology, wherein the defect detection network includes an input layer, a convolutional layer, and an output layer; Constructing a training data set, wherein the training data set includes a sample fused light image, and the sample fused light image and the fused light image are fused by the same fusion method; The defect detection network is trained according to the training data set to obtain the pre-trained defect detection network.

5. A glass substrate detection device, characterized in that: The device comprises: An acquisition module is used to, in a darkroom, perform a first irradiation on a glass substrate to be detected at a first angle by using a first parallel light set on a first side of the glass substrate to be detected; during the first irradiation process, digitally collect light passing through the glass substrate to be detected to obtain a first light image; in the darkroom, perform a second irradiation on the glass substrate to be detected at a second angle by using a second parallel light set on a second side of the glass substrate to be detected, wherein the first irradiation and the second irradiation are performed separately; during the second irradiation process, digitally collect light passing through the glass substrate to be detected to obtain a second light image; the first light image and the second light image have the same size and resolution; the first side and the second side are opposite sides of the glass substrate to be detected; the first angle and the second angle are any angle value between 30 degrees and 60 degrees; An integration module is used to align the first light image with the second light image; add the light element value of each light element in the first light image with the light element value of the corresponding light element in the second light image to obtain a first fused image; average the light element value of each light element in the first light image with the light element value of the corresponding light element in the second light image to obtain a second fused image; perform difference calculation between the light element value of each light element in the first light image and the light element value of the corresponding light element in the second light image to obtain a third fused image; superimpose the first fused image, the second fused image and the third fused image as channels to obtain a fused light image; The detection module is used to perform defect detection on the fused light image through a pre-trained defect detection network, and output a detection result of the glass substrate to be detected.

6. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the glass substrate detection method according to any one of claims 1 to 4 when executing the computer program.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the method for detecting a glass substrate according to any one of claims 1 to 4 are implemented.

Images