TWI755213B - Method and device for detecting defections, computer device and storage medium - Google Patents

Abstract

The present application relates to an image detection technique, and the present application provides a method and a device for detecting defections, a computer device and a storage medium. The method can obtain a detection image and a plurality of positive sample images, encode the detection image and the plurality of positive sample images to obtain a target vector and a plurality of subtexts, and decode the target vector and the plurality of subtexts. The method further obtains a bull's-eye chart image and a plurality of reconstruction images, compares the bull's-eye chart image and the detection image to obtain a target error, compares each reconstruction image and each positive sample image to obtain a reconstruction error. The method further determines a test probability and a test error of the detection image, determines an estimated probability and a sample error of each positive sample image and determines an error threshold. The method further determines a test result according to the test error and the error threshold. The present application can detect subtle defects in the detection image and improve the accuracy of a defect detection.

Description

Defect detection method, device, computer device and storage medium

The present application relates to the technical field of image detection, and in particular, to a defect detection method, device, computer device and storage medium.

In order to improve the quality of industrial products, certain defects are usually detected on industrial products before they are packaged. Because the current defect detection method will generate certain errors in the process of reconstructing the image, it is impossible to detect products with subtle defects, thereby reducing the accuracy of defect detection.

In view of the above, it is necessary to provide a defect detection method, device, computer device and storage medium, which can detect whether there are subtle defects in the image to be inspected, thereby improving the accuracy of defect detection.

A first aspect of the present application provides a flaw detection method, the flaw detection method includes: when a flaw detection request is received, acquiring an image to be detected and a plurality of positive sample images according to the flaw detection request; The to-be-detected image is encoded to obtain a target vector corresponding to the to-be-detected image, and the encoder is used to encode the plurality of positive sample images to obtain a target vector corresponding to the plurality of positive samples. Multiple latent vectors corresponding to the image; The target vector is decoded by the decoder corresponding to the encoder to obtain a bullseye image corresponding to the to-be-detected image, and the decoder is used to decode the multiple latent vectors, Obtain multiple reconstructed images corresponding to the multiple positive sample images; compare the bullseye image with the to-be-detected image to obtain the target error, and determine the target error according to the multiple reconstructed images and the multiple images. The positive sample image determines the reconstruction error of each positive sample image; the target vector is input into the pre-trained Gaussian mixture model, the test probability of the to-be-detected image is obtained, and the multiple latent images are obtained. The vector is input into the Gaussian mixture model, and the estimated probability of each positive sample image is obtained; the test error of the to-be-detected image is determined according to the target error and the test probability, and according to each reconstruction error and each estimated probability to determine the sample error of each positive sample image; select an error threshold from the sample error, and determine the detection result of the to-be-detected image according to the test error and the error threshold.

A second aspect of the present application provides a defect detection device, the defect detection device includes: an acquisition unit configured to, when a defect detection request is received, acquire an image to be detected and a plurality of positive sample images according to the defect detection request The coding unit is used to encode the image to be detected using an encoder to obtain a target vector corresponding to the image to be detected, and to encode the multiple positive sample images using the encoder processing, to obtain a plurality of latent vectors corresponding to the plurality of positive sample images; a decoding unit, used for decoding the target vector with a decoder corresponding to the encoder, to obtain a picture corresponding to the to-be-detected image image corresponding bullseye images, and use the decoder to decode the multiple latent vectors to obtain multiple reconstructed images corresponding to the multiple positive sample images; The determining unit is configured to compare the bullseye image with the to-be-detected image to obtain a target error, and determine the value of each positive sample image according to the multiple reconstructed images and the multiple positive sample images. Reconstruction error; an input unit for inputting the target vector into a pre-trained Gaussian mixture model to obtain the test probability of the to-be-detected image, and inputting the multiple latent vectors into the Gaussian mixture In the model, the estimated probability of each positive sample image is obtained; the determining unit is further configured to determine the test error of the to-be-detected image according to the target error and the test probability, and according to each reconstruction The error and each estimated probability determine the sample error of each positive sample image; the determining unit is further configured to select an error threshold from the sample error, and determine the error threshold according to the test error and the error threshold The detection result of the image to be detected.

A third aspect of the present application provides a computer device, the computer device comprising: a storage for storing at least one instruction; and a processor for acquiring the instruction stored in the storage to implement the defect detection method.

A fourth aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores at least one instruction, and the at least one instruction is acquired by a processor in a computer device to implement the defect detection method.

It can be seen from the above technical solutions that the present application can accurately determine the error threshold by determining the reconstruction error generated during image reconstruction and by determining the estimated probability generated by the Gaussian mixture model, and then by comparing the test error with the above. As for the error threshold, since the test error is numerically compared with the error threshold, it is possible to detect whether there are subtle defects in the to-be-detected image, thereby improving the accuracy of defect detection.

1: Computer device

12: Storage

13: Processor

11: Defect detection device

110: Get Unit

111: coding unit

112: decoding unit

113: Determine unit

114: Input unit

115: Generate unit

FIG. 1 is a flowchart of a preferred embodiment of the flaw detection method of the present application.

FIG. 2 is a functional module diagram of a preferred embodiment of the defect detection device of the present application.

FIG. 3 is a schematic structural diagram of a computer device for implementing a preferred embodiment of the defect detection method of the present application.

In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be described in detail below with reference to the accompanying drawings and specific embodiments.

As shown in FIG. 1 , it is a flowchart of a preferred embodiment of the defect detection method of the present application. According to different requirements, the order of the steps in this flowchart can be changed, and some steps can be omitted.

The defect detection method is applied to one or more computer devices 1, and the computer device 1 is a device that can automatically perform numerical calculation and/or information processing according to pre-set or stored instructions, and its hardware includes but not Limited to microprocessors, Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), Digital Signal Processors (DSPs), embedded devices, etc.

The computer device 1 can be any electronic product that can interact with a user, such as a personal computer, a tablet computer, a smart phone, a personal digital assistant (PDA), a game console, and an interactive network television. (Internet Protocol Television, IPTV), smart wearable devices, etc.

The computer device 1 may also include network equipment and/or user equipment. Wherein, the network device includes, but is not limited to, a single network server, a server group formed by multiple network servers, or a cloud formed by a large number of hosts or network servers based on cloud computing (Cloud Computing).

The network where the computer device 1 is located includes, but is not limited to, the Internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (Virtual Private Network, VPN), and the like.

In step S10, when a defect detection request is received, an image to be detected and a plurality of positive sample images are acquired according to the defect detection request.

In at least one embodiment of the present application, the defect detection request may be triggered by a user (eg, triggered by a preset function button), or may be automatically triggered within a preset time, which is not limited in this application.

Wherein, the preset time may be a time point (for example, nine o'clock in the morning every day), or a time period.

In at least one embodiment of the present application, the information carried in the defect detection request includes, but is not limited to, a detection object and the like.

In at least one embodiment of the present application, the computer device acquiring the image to be detected and the plurality of positive sample images according to the flaw detection request includes: parsing the method body of the flaw detection request to obtain the flaw detection request data information carried; obtain a preset label, and obtain information corresponding to the preset label from the data information, as the detection object; obtain the to-be-detected image from the to-be-detected library according to the detection object image, and obtain the plurality of positive sample images from the sample library according to the detection object.

Wherein, the preset tag refers to a pre-defined tag, for example, the preset tag may be name.

Further, the to-be-detected images without flaw detection are stored in the to-be-detected library, and a plurality of flawless positive sample images are stored in the sample library.

By parsing the method body of the flaw detection request, the parsing time of the flaw detection request can be shortened, thereby improving parsing efficiency, and the detection object can be accurately determined through the mapping relationship between the preset label and the detection object, and then the detection object can be accurately determined. Accurately acquire the to-be-detected image and the multiple positive sample images.

Step S11, using an encoder to perform encoding processing on the image to be detected to obtain a target vector corresponding to the image to be detected, and using the encoder to perform encoding processing on the plurality of positive sample images to obtain: a plurality of latent vectors corresponding to the plurality of positive sample images.

In at least one embodiment of the present application, the encoder may be an encoder in an autoencoder (AE). Further, the encoder includes multiple hidden layers, and the number of the multiple hidden layers can be arbitrarily set according to application scenarios.

In at least one embodiment of the present application, the computer device performs encoding processing on the image to be detected by using an encoder, and obtaining a target vector corresponding to the image to be detected includes: performing an encoding process on the image to be detected. Vectorization processing is performed to obtain the first feature vector of the image to be detected; the hidden layer in the encoder is extracted; the first feature vector is operated by using the hidden layer to obtain the target vector.

Specifically, the computer device uses the hidden layer to operate on the first eigenvector, and obtaining the target vector includes: obtaining the weight matrix and offset value of the hidden layer; combining the first eigenvector with The weight matrix is multiplied to obtain an operation result; the operation result and the offset value are added to obtain the target vector.

In other embodiments, the computer device uses the hidden layer to operate each second feature vector in the same manner as the computer device uses the hidden layer to operate the first feature vector. The application will not repeat this.

Step S12, use the decoder corresponding to the encoder to decode the target vector to obtain a bullseye image corresponding to the to-be-detected image, and use the decoder to perform decoding on the multiple latent vectors. A decoding process is performed to obtain a plurality of reconstructed images corresponding to the plurality of positive sample images.

In at least one embodiment of the present application, the decoder may be a decoder in the auto-encoder. Further, the decoder includes an operation layer corresponding to the hidden layer in the encoder.

In at least one embodiment of the present application, the computer apparatus operates on the target vector by using the operation layer, and performs reduction processing on the vector obtained after the operation to obtain the target vector.

In other embodiments, the manner in which the computer device obtains the plurality of reconstructed images is the same as the manner in which the target vector is obtained, which will not be repeated in this application.

Step S13, compare the bullseye image with the to-be-detected image to obtain the target error, and determine the reconstruction of each positive sample image according to the multiple reconstructed images and the multiple positive sample images error.

In at least one embodiment of the present application, the target error refers to an error generated during reconstruction of the image to be detected.

In at least one embodiment of the present application, the computer device compares the bullseye image with the to-be-detected image, and obtaining the target error includes: extracting all the primitive points of the to-be-detected image to obtain multiple each primitive point to be detected, and extract all primitive points of the bullseye image to obtain a plurality of target primitive points; compare each target primitive point with each primitive point to be detected to obtain a comparison result; When the comparison result shows that the target primitive point is different from the to-be-detected primitive point, calculate the number of the target primitive point and the to-be-detected primitive point different as the first number, and calculate the number of the multiple target primitive points As a second quantity; the target error is obtained by dividing the first quantity by the second quantity.

Through the above-described embodiments, the target error can be accurately determined.

In other embodiments, the method of determining the reconstruction error of each positive sample image by the computer device is the same as the method of determining the target error, which will not be repeated in this application.

Step S14, inputting the target vector into a pre-trained Gaussian Mixture Model (GMM) to obtain the test probability of the image to be detected, and inputting the multiple latent vectors into the Gaussian Mixture Model (GMM) In the mixed model, the estimated probability of each positive sample image is obtained.

In at least one embodiment of the present application, the Gaussian mixture model refers to an open-source mixture model, and the Gaussian mixture model includes a plurality of single Gaussian models.

In at least one embodiment of the present application, the computer device inputs the target vector into a pre-trained Gaussian mixture model, obtains the test probability of the to-be-detected image, and inputs the multiple latent vectors into In the Gaussian mixture model, obtaining the estimated probability of each positive sample image includes: inputting the multiple latent vectors into the Gaussian mixture model to obtain the feature distribution of the multiple positive sample images; Determine the mean value and covariance of the multiple latent vectors according to the feature distribution, and obtain the mixture coefficient of the Gaussian mixture model; according to the target vector, the mean value, the covariance and the mixture coefficient The test probability of the image to be detected is determined, and the estimated probability of each positive sample image is determined according to each latent vector, the average value, the covariance and the mixing coefficient.

Through the above embodiment, the test probability and the estimated probability can be accurately determined.

Step S15: Determine the test error of the to-be-detected image according to the target error and the test probability, and determine the sample error of each positive sample image according to each reconstruction error and each estimated probability.

In at least one embodiment of the present application, the computer device determining the sample error of each positive sample image according to each reconstruction error and each estimated probability includes: calculating the logarithm of each estimated probability to obtain each Estimate the log value of the probability; perform a weighted sum operation on the inverse of each log value and each reconstruction error to obtain the sample error.

For example: the estimated probability is 0.01, the reconstruction error is 0.03, the logarithm of the estimated probability is calculated, and the logarithm value is: log(0.01)=-2, the inverse of the logarithm value is calculated, and the value is 2, calculate 2 and 0.03 The weighted sum of , when the estimated probability accounts for 20% of the sample error, and the reconstruction error accounts for 80% of the sample error, the calculated sample error is: 2*20%+0.03*80%=0.424.

Through the above-mentioned embodiments, the error interval generated by the image reconstruction process and the probability distribution can be determined.

Step S16, selecting an error threshold from the sample errors, and determining the detection result of the to-be-detected image according to the test error and the error threshold.

In at least one embodiment of the present application, the detection result includes that the image to be inspected is defective and the image to be inspected is flawless.

In at least one embodiment of the present application, selecting the error threshold from the sample errors by the computer device includes: sorting the sample errors in ascending order to obtain an error list and a sample serial number of each sample error ; Calculate the quantity of the sample error, and multiply the quantity by the configuration value to obtain the target value; select the sample error whose sample serial number is equal to the target value from the error list as the error threshold.

Through the above-described embodiments, errors generated when the image reconstruction process and probability distribution are affected can be determined.

In at least one embodiment of the present application, the computer device determining the detection result of the to-be-detected image according to the test error and the error threshold includes: when the test error is less than the error threshold, setting the The detection result is determined as the image to be detected is flawless; or when the test error is greater than or equal to the error threshold, the detection result is determined as the image to be detected as defective.

By comparing the test error with the error threshold, since the test error is numerically compared with the error threshold, it is possible to detect whether there are subtle flaws in the image to be detected, thereby improving the performance of the image. Accuracy of flaw detection.

In at least one embodiment of the present application, when the image to be detected is defective, the computer device generates reminder information according to the image to be detected, and sends the reminder information to a terminal device of a designated contact person middle.

Wherein, the designated contact person may be a quality person responsible for testing the testing object.

Through the above-mentioned embodiment, when there is a defect in the image to be detected, the designated contact person can be notified in time.

It can be seen from the above technical solutions that the present application can accurately determine the error threshold by determining the reconstruction error generated during image reconstruction and by determining the estimated probability generated by the Gaussian mixture model, and then by comparing the test error with the above. As for the error threshold, since the test error is numerically compared with the error threshold, it is possible to detect whether there are subtle defects in the to-be-detected image, thereby improving the accuracy of defect detection.

As shown in FIG. 2 , it is a functional module diagram of a preferred embodiment of the defect detection device of the present application. The defect detection device 11 includes an acquisition unit 110 , an encoding unit 111 , a decoding unit 112 , a determination unit 113 , an input unit 114 and a generation unit 115 . The module/unit referred to in this application refers to a series of computer program segments that can be acquired by the processor 13 and can perform fixed functions, and are stored in the storage 12 . In this embodiment, the functions of each module/unit will be described in detail in subsequent embodiments.

When receiving the defect detection request, the acquiring unit 110 acquires the image to be detected and a plurality of positive sample images according to the defect detection request.

In at least one embodiment of the present application, the defect detection request may be triggered by a user (eg, triggered by a preset function button), or may be automatically triggered within a preset time, which is not limited in this application.

Wherein, the preset time may be a time point (for example, nine o'clock in the morning every day), or a time period.

In at least one embodiment of the present application, the information carried in the defect detection request includes, but is not limited to, a detection object and the like.

In at least one embodiment of the present application, the acquiring unit 110 acquiring the image to be detected and a plurality of positive sample images according to the flaw detection request includes: parsing the method body of the flaw detection request to obtain the flaw detection request Request the carried data information; obtain a preset label, and obtain the information corresponding to the preset label from the data information as the detection object; obtain the to-be-detected object from the to-be-detected library according to the detection object image, and obtain the plurality of positive sample images from the sample library according to the detection object.

Wherein, the preset tag refers to a pre-defined tag, for example, the preset tag may be name.

Further, the to-be-detected images without flaw detection are stored in the to-be-detected library, and a plurality of flawless positive sample images are stored in the sample library.

By parsing the method body of the flaw detection request, the parsing time of the flaw detection request can be shortened, thereby improving parsing efficiency, and the detection object can be accurately determined through the mapping relationship between the preset label and the detection object, and then the detection object can be accurately determined. Accurately acquire the to-be-detected image and the multiple positive sample images.

The encoding unit 111 uses an encoder to perform encoding processing on the image to be detected to obtain a target vector corresponding to the image to be detected, and uses the encoder to perform encoding processing on the plurality of positive sample images to obtain a plurality of latent vectors corresponding to the plurality of positive sample images.

In at least one embodiment of the present application, the encoder may be an encoder in an autoencoder (AE). Further, the encoder includes multiple hidden layers, and the number of the multiple hidden layers can be arbitrarily set according to application scenarios.

In at least one embodiment of the present application, the encoding unit 111 uses an encoder to perform encoding processing on the image to be detected, and obtaining a target vector corresponding to the image to be detected includes: encoding the image to be detected Perform vectorization processing to obtain the first feature vector of the to-be-detected image; Extracting the hidden layer in the encoder; using the hidden layer to operate on the first feature vector to obtain the target vector.

Specifically, the encoding unit 111 uses the hidden layer to operate on the first feature vector, and obtaining the target vector includes: acquiring the weight matrix and offset value of the hidden layer; Perform a multiplication operation with the weight matrix to obtain an operation result; perform an addition operation on the operation result and the offset value to obtain the target vector.

In other embodiments, the encoding unit 111 uses the hidden layer to operate each second feature vector in the same manner as the encoding unit 111 uses the hidden layer to operate the first feature vector , which will not be repeated in this application.

The decoding unit 112 uses the decoder corresponding to the encoder to decode the target vector to obtain a bullseye image corresponding to the image to be detected, and uses the decoder to perform decoding on the multiple latent vectors. A decoding process is performed to obtain a plurality of reconstructed images corresponding to the plurality of positive sample images.

In at least one embodiment of the present application, the decoder may be a decoder in the auto-encoder. Further, the decoder includes an operation layer corresponding to the hidden layer in the encoder.

In at least one embodiment of the present application, the decoding unit 112 uses the operation layer to operate on the target vector, and performs restoration processing on the vector obtained after the operation to obtain the target vector.

In other embodiments, the manner in which the decoding unit 112 obtains the plurality of reconstructed images is the same as the manner in which the target vector is obtained, which will not be repeated in this application.

The determining unit 113 compares the bullseye image with the to-be-detected image to obtain the target error, and determines the reconstruction of each positive sample image according to the multiple reconstructed images and the multiple positive sample images error.

In at least one embodiment of the present application, the target error refers to an error generated during reconstruction of the image to be detected.

In at least one embodiment of the present application, the determining unit 113 compares the bullseye image with the image to be detected, and obtaining the target error includes: extracting all the primitive points of the image to be detected, and obtaining multiple primitive points to be detected, and extract all primitive points of the bullseye image to obtain multiple target primitive points; compare each target primitive point with each primitive point to be detected to obtain a comparison result ; When the comparison result shows that the target primitive point is different from the primitive point to be detected, calculate the different quantity of the target primitive point and the primitive point to be detected, and use it as the first quantity, and calculate the number of the plurality of target primitive points. Quantity as the second quantity; dividing the first quantity by the second quantity to obtain the target error.

Through the above-described embodiments, the target error can be accurately determined.

In other embodiments, the manner in which the determining unit 113 determines the reconstruction error of each positive sample image is the same as the manner in which the target error is determined, which will not be repeated in this application.

The input unit 114 inputs the target vector into a pre-trained Gaussian Mixture Model (GMM), obtains the test probability of the image to be detected, and inputs the multiple latent vectors into the Gaussian In the mixed model, the estimated probability of each positive sample image is obtained.

In at least one embodiment of the present application, the Gaussian mixture model refers to an open-source mixture model, and the Gaussian mixture model includes a plurality of single Gaussian models.

In at least one embodiment of the present application, the input unit 114 inputs the target vector into a pre-trained Gaussian mixture model to obtain the test probability of the to-be-detected image, and converts the multiple latent vectors Input into the Gaussian mixture model, and obtaining the estimated probability of each positive sample image includes: inputting the multiple latent vectors into the Gaussian mixture model to obtain the feature distribution of the multiple positive sample images ; Determine the mean value and covariance of the multiple latent vectors according to the feature distribution, and obtain the mixture coefficient of the Gaussian mixture model; according to the target vector, the mean value, the covariance and the mixture coefficient The test probability of the image to be detected is determined, and the estimated probability of each positive sample image is determined according to each latent vector, the average value, the covariance and the mixing coefficient.

Through the above embodiment, the test probability and the estimated probability can be accurately determined.

The determining unit 113 determines the test error of the to-be-detected image according to the target error and the test probability, and determines the sample error of each positive sample image according to each reconstruction error and each estimated probability.

In at least one embodiment of the present application, the determining unit 113 determining the sample error of each positive sample image according to each reconstruction error and each estimated probability includes: calculating the logarithm of each estimated probability to obtain each logarithmic values of the estimated probabilities; perform a weighted sum operation on the inverse of each logarithmic value and each reconstruction error to obtain the sample error.

For example: the estimated probability is 0.01, the reconstruction error is 0.03, the logarithm of the estimated probability is calculated, and the logarithm value is: log(0.01)=-2, the inverse of the logarithm value is calculated, and the value is 2, calculate 2 and 0.03 The weighted sum of , when the estimated probability accounts for 20% of the sample error, and the reconstruction error accounts for 80% of the sample error, the calculated sample error is: 2*20%+0.03*80%=0.424.

Through the above-mentioned embodiments, the error interval generated by the image reconstruction process and the probability distribution can be determined.

The determining unit 113 selects an error threshold from the sample errors, and determines a detection result of the to-be-detected image according to the test error and the error threshold.

In at least one embodiment of the present application, the detection result includes that the image to be inspected is defective and the image to be inspected is flawless.

In at least one embodiment of the present application, the determining unit 113 selecting an error threshold from the sample errors includes: Sort the sample errors in ascending order to obtain the error list and the sample serial number of each sample error; calculate the number of the sample errors, and multiply the number by the configuration value to obtain the target value; The sample error whose sample serial number is equal to the target value is selected from the error list as the error threshold.

Through the above-described embodiments, errors generated when the image reconstruction process and probability distribution are affected can be determined.

In at least one embodiment of the present application, the determining unit 113 determining the detection result of the to-be-detected image according to the test error and the error threshold includes: when the test error is less than the error threshold, setting The detection result is determined that the image to be detected is flawless; or when the test error is greater than or equal to the error threshold, the detection result is determined that the image to be detected is defective.

By comparing the test error with the error threshold, since the test error is numerically compared with the error threshold, it is possible to detect whether there are subtle flaws in the image to be detected, thereby improving the performance of the image. Accuracy of flaw detection.

In at least one embodiment of the present application, when the image to be detected is defective, the generating unit 115 generates reminder information according to the image to be detected, and sends the reminder information to the terminal device of the designated contact person .

Wherein, the designated contact person may be a quality person responsible for testing the testing object.

Through the above-mentioned embodiment, when there is a defect in the image to be detected, the designated contact person can be notified in time.

It can be seen from the above technical solutions that the present application can accurately determine the error threshold by determining the reconstruction error generated during image reconstruction and by determining the estimated probability generated by the Gaussian mixture model, and then by comparing the test error with the above. error threshold, due to the numerical error on the test By comparing with the error threshold, it is possible to detect whether there are subtle flaws in the to-be-detected image, thereby improving the accuracy of flaw detection.

As shown in FIG. 3 , it is a schematic structural diagram of a computer device according to a preferred embodiment of the flaw detection method of the present application.

In one embodiment of the present application, the computer device 1 includes, but is not limited to, a storage 12 , a processor 13 , and a computer program stored in the storage 12 and running on the processor 13 , such as defect detection programs.

Those skilled in the art can understand that the schematic diagram is only an example of the computer device 1, and does not constitute a limitation on the computer device 1. It may include more or less components than the one shown, or combine some components, or different Components, such as the computer device 1, may also include input and output devices, network access devices, bus bars, and the like.

The processor 13 may be a Central Processing Unit (CPU), other general-purpose processors, a Digital Signal Processor (DSP), or an Application Specific Integrated Circuit (ASIC). , Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The processor 13 is the computing core and control center of the computer device 1, and uses various interfaces and lines to connect the entire computer device 1 , and obtain the operating system of the computer device 1 and various installed applications, code, etc.

The processor 13 acquires the operating system of the computer device 1 and various installed applications. The processor 13 acquires the application program to implement the steps in each of the above-mentioned embodiments of the defect detection method, such as the steps shown in FIG. 1 .

Exemplarily, the computer program can be divided into one or more modules/units, and the one or more modules/units are stored in the storage 12 and acquired by the processor 13, to complete this application. The one or more modules/units may be a series of computer program instruction segments capable of accomplishing specific functions, and the instruction segments are used to describe the acquisition process of the computer program in the computer device 1. Procedure. For example, the computer program may be divided into an acquisition unit 110 , an encoding unit 111 , a decoding unit 112 , a determination unit 113 , an input unit 114 , and a generation unit 115 .

The storage 12 can be used to store the computer programs and/or modules, and the processor 13 executes or obtains the computer programs and/or modules stored in the storage 12, and calls the computer programs and/or modules stored in the storage 12. 12 to realize various functions of the computer device 1 . The storage 12 may mainly include a program storage area and a data storage area, wherein the storage program area may store an operating system, an application program required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; storage data The area can store data etc. created according to the use of the computer device. In addition, the storage 12 may include non-volatile storage such as hard disk, storage, plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, flash memory A memory card (Flash Card), at least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device.

The storage 12 may be an external storage and/or an internal storage of the computer device 1 . Further, the storage 12 may be a storage in physical form, such as a storage bar, a TF card (Trans-flash Card), and the like.

If the modules/units integrated in the computer device 1 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the present application realizes all or part of the processes in the methods of the above embodiments, and can also be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the computer program is acquired by the processor, the steps of each of the above method embodiments can be implemented.

Wherein, the computer program includes computer program code, and the computer program code may be in the form of original code, object code, obtainable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, pen drive, removable hard disk, magnetic disk, optical disk, computer storage, read-only storage (ROM, Read-only storage) Only Memory).

Referring to FIG. 1 , the storage 12 in the computer device 1 stores a plurality of instructions to implement a defect detection method, and the processor 13 can obtain the plurality of instructions to implement: When a flaw detection request is received, the image to be detected and a plurality of positive sample images are obtained according to the flaw detection request; and use the encoder to encode the multiple positive sample images to obtain multiple latent vectors corresponding to the multiple positive sample images; use the decoder corresponding to the encoder Perform decoding processing on the target vector to obtain a bullseye image corresponding to the image to be detected, and use the decoder to perform decoding processing on the multiple latent vectors to obtain images corresponding to the multiple positive samples Corresponding multiple reconstructed images; compare the bullseye image with the to-be-detected image to obtain the target error, and determine each positive sample image according to the multiple reconstructed images and the multiple positive sample images image reconstruction error; input the target vector into the pre-trained Gaussian mixture model to obtain the test probability of the to-be-detected image, and input the multiple latent vectors into the Gaussian mixture model, Obtain the estimated probability of each positive sample image; determine the test error of the to-be-detected image according to the target error and the test probability, and determine each positive image according to each reconstruction error and each estimated probability. The sample error of the sample image; an error threshold is selected from the sample error, and the detection result of the to-be-detected image is determined according to the test error and the error threshold.

Specifically, for the specific implementation method of the above-mentioned instruction by the processor 13, reference may be made to the description of the relevant steps in the embodiment corresponding to FIG. 1 , which is not repeated here.

In the several embodiments provided in this application, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the device embodiments described above are only illustrative. For example, the division of the modules is only a logical function division, and other division methods may be used in actual implementation.

The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place. It can also be distributed over multiple network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

In addition, each functional module in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware, or can be implemented in the form of hardware plus software function modules.

Accordingly, the embodiments are to be regarded in all respects as illustrative and not restrictive, and the scope of this application is defined by the appended claims rather than the foregoing description, and is therefore intended to fall within the scope of the claims. All changes within the meaning and scope of the equivalents of , are included in this application. Any associated icon indicia in a claim should not be considered to limit the claim to which it relates.

Furthermore, it is clear that the word "comprising" does not exclude other elements or steps, and the singular does not exclude the plural. A plurality of units or devices stated in this application may also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to denote names and do not denote any particular order.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application rather than limitations. Although the present application has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present application can be Modifications or equivalent substitutions can be made without departing from the spirit and scope of the technical solutions of the present application.

Claims (9)

A flaw detection method, wherein the flaw detection method comprises: when a flaw detection request is received, acquiring an image to be detected and a plurality of positive sample images according to the flaw detection request; performing encoding processing on the image to be detected to obtain a target vector corresponding to the image to be detected, including: performing vectorization processing on the image to be detected to obtain a first feature vector of the image to be detected; extracting the encoder The hidden layer in ; use the hidden layer to operate on the first feature vector to obtain the target vector, and use the encoder to encode the plurality of positive sample images to obtain the same multiple latent vectors corresponding to the positive sample image; decode the target vector by using the decoder corresponding to the encoder to obtain a bullseye image corresponding to the image to be detected, and use the decoding process to decode the target vector. The device decodes the multiple latent vectors to obtain multiple reconstructed images corresponding to the multiple positive sample images; compares the bullseye image with the to-be-detected image to obtain the target error, and Determine the reconstruction error of each positive sample image according to the multiple reconstructed images and the multiple positive sample images; input the target vector into a pre-trained Gaussian mixture model to obtain the image to be detected the test probability of the image, and input the multiple latent vectors into the Gaussian mixture model to obtain the estimated probability of each positive sample image; determine the to-be-detected image according to the target error and the test probability The test error of the image is determined, and the sample error of each positive sample image is determined according to each reconstruction error and each estimated probability; an error threshold is selected from the sample errors, and according to the test error and the error threshold A detection result of the to-be-detected image is determined. The defect detection method according to claim 1, wherein the comparing the bullseye image with the to-be-detected image to obtain the target error comprises: extracting all the primitive points of the to-be-detected image to obtain multiple primitive points to be detected, and extract all primitive points of the bullseye image to obtain multiple target primitive points; compare each target primitive point with each primitive point to be detected to obtain a comparison result ; When the comparison result shows that the target primitive point is different from the to-be-detected primitive point, calculate the number of the target primitive point and the to-be-detected primitive point different as the first number, and calculate the number of the multiple target primitive points As a second quantity; the target error is obtained by dividing the first quantity by the second quantity. The flaw detection method according to claim 1, wherein the target vector is input into a pre-trained Gaussian mixture model to obtain the test probability of the to-be-detected image, and the multiple latent vectors are Input into the Gaussian mixture model, and obtaining the estimated probability of each positive sample image includes: inputting the multiple latent vectors into the Gaussian mixture model to obtain the feature distribution of the multiple positive sample images ; Determine the mean value and covariance of the multiple latent vectors according to the feature distribution, and obtain the mixture coefficient of the Gaussian mixture model; According to the target vector, the mean value, the covariance and the mixture The coefficient determines the test probability of the image to be detected, and the estimated probability of each positive sample image is determined according to each latent vector, the average value, the covariance and the mixing coefficient. The flaw detection method according to claim 1, wherein the determining the sample error of each positive sample image according to each reconstruction error and each estimated probability includes: calculating the logarithm of each estimated probability, and obtaining each logarithmic values of the estimated probabilities; perform a weighted sum operation on the inverse of each logarithmic value and each reconstruction error to obtain the sample error. The defect detection method according to claim 1, wherein the selecting an error threshold from the sample errors comprises: sorting the sample errors in ascending order to obtain an error list and a sample of each sample error serial number; calculate the quantity of the sample error, and multiply the quantity by the configuration value to obtain the target value; A sample error whose sample serial number is equal to the target value is selected from the error list as the error threshold. The defect detection method according to claim 1, wherein the determining the detection result of the to-be-detected image according to the test error and the error threshold comprises: when the test error is less than the error threshold, determining The detection result is determined that the image to be detected is flawless; or when the test error is greater than or equal to the error threshold, the detection result is determined that the image to be detected is defective. A defect detection device, wherein the defect detection device comprises: an acquisition unit for acquiring an image to be detected and a plurality of positive sample images according to the defect detection request when a defect detection request is received; an encoding unit for using Using an encoder to perform encoding processing on the image to be detected to obtain a target vector corresponding to the image to be detected includes: performing vectorization processing on the image to be detected to obtain a vector of the image to be detected. the first feature vector; extract the hidden layer in the encoder; use the hidden layer to operate on the first feature vector to obtain the target vector, and use the encoder to perform operations on the multiple positive sample images image encoding processing to obtain a plurality of latent vectors corresponding to the plurality of positive sample images; a decoding unit for decoding the target vector by using a decoder corresponding to the encoder to obtain a decoding process corresponding to the the bullseye image corresponding to the image to be detected, and the decoder is used to decode the multiple latent vectors to obtain multiple reconstructed images corresponding to the multiple positive sample images; the determining unit is used for converting The bullseye image is compared with the to-be-detected image to obtain the target error, and the reconstruction error of each positive sample image is determined according to the plurality of reconstructed images and the plurality of positive sample images; the input unit , used to input the target vector into the pre-trained Gaussian mixture model to obtain the test probability of the image to be detected, and input the multiple latent vectors into the Gaussian mixture model to obtain each Estimated probability of positive sample images; The determining unit is further configured to determine the test error of the to-be-detected image according to the target error and the test probability, and determine the value of each positive sample image according to each reconstruction error and each estimated probability. Sample error; the determining unit is further configured to select an error threshold from the sample error, and determine the detection result of the to-be-detected image according to the test error and the error threshold. A computer device, wherein the computer device comprises: a storage for storing at least one instruction; and a processor for acquiring the instructions stored in the storage to implement defect detection according to any one of claim 1 to claim 6 method. A computer-readable storage medium, wherein: the computer-readable storage medium stores at least one instruction, and the at least one instruction is acquired by a processor in a computer device to implement any one of claim 1 to 6. defect detection method.

Images