Industrial product quality detection method and system for artificial intelligence

Abstract

The present invention discloses an industrial product quality detection method and system for artificial intelligence, which relates to the field of data processing technology. The method includes: by obtaining the surface quality index of the industrial product, performing grayscale recognition feature analysis, generating the environmental constraint brightness, and the camera collecting constraint parameters and grayscale deviation thresholds. When the industrial product to be inspected enters the preset area of ​​the carrier, a first product image is obtained and grayscale processing is performed to obtain the first product grayscale image. For the first product grayscale image, defect identification is performed based on the grayscale deviation threshold to obtain suspicious abnormal areas. When the number of suspicious abnormal areas is not equal to 0, the first product image is sent to a cloud server. When the number of suspicious abnormal areas is equal to 0, the industrial product surface quality index of the industrial product to be inspected is marked as normal. The technical problem that the existing industrial product quality detection method has a large computing pressure of a large cloud model and is difficult to improve the overall detection efficiency and accuracy is solved.

Description

Technical Field

[0001] The present application relates to fields related to data processing technology, and specifically to industrial product quality detection methods and systems for artificial intelligence. Background Art

[0002] With the continuous development of industrial manufacturing technology, the quality inspection of industrial products plays a vital role in the production process. Industrial product quality inspection relies on large model inspection, and the user's local computing resources are usually insufficient to support the operation of large models. However, with the improvement of industrial product quality inspection accuracy and computing requirements, the computing pressure of large models deployed in the cloud is too high and cannot process all image data in real time, affecting the inspection efficiency and effect.

[0003] Therefore, in the existing technology, the industrial product quality inspection method has the technical problems of high computational pressure of large cloud models and difficulty in improving the overall inspection efficiency and accuracy. Summary of the invention

[0004] This application solves the technical problem that the existing industrial product quality detection methods have high computing pressure of large cloud models and difficulty in improving overall detection efficiency and accuracy by providing an industrial product quality detection method and system for artificial intelligence. By deploying edge servers locally to perform preliminary processing and screening of grayscale images, some images are sent to the cloud large model for processing, thereby effectively reducing the computing pressure of the cloud large model and improving the overall detection efficiency and accuracy.

[0005] The present application provides an industrial product quality detection method for artificial intelligence, which is applied to an edge server, including: obtaining an industrial product surface quality index, performing grayscale recognition feature analysis on the industrial product surface quality index, generating environmental constraint brightness, camera acquisition constraint parameters and grayscale deviation thresholds. When the industrial product to be inspected enters a preset area of ​​the carrier, the camera is controlled to shoot according to the environmental constraint brightness and the camera acquisition constraint parameters to obtain a first product image. The first product image is grayed to obtain a first product grayscale image. For the first product grayscale image, defect identification is performed based on the grayscale deviation threshold to obtain a suspected abnormal area. When the number of the suspected abnormal areas is not equal to 0, the first product image is sent to a cloud server. When the number of the suspected abnormal areas is equal to 0, the industrial product surface quality index of the industrial product to be inspected is marked as normal.

[0006] In the implementation, the grayscale recognition feature analysis is performed on the surface quality index of the industrial product to generate the environment constraint brightness, camera acquisition constraint parameter and grayscale deviation threshold, including: the surface quality index of the industrial product includes the first surface quality index until the Nth surface quality index. The grayscale recognition feature analysis is performed by traversing the first surface quality index until the Nth surface quality index to obtain the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold. According to the environment constraint brightness and the camera acquisition constraint parameter, the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold are clustered to generate the first cluster quality index until the Mth cluster quality index. The first cluster quality index has the first environment brightness label, the first camera acquisition constraint parameter label and the first grayscale deviation threshold label, and the Mth cluster quality index has the Mth environment brightness label, the Mth camera acquisition constraint parameter label and the Mth grayscale deviation threshold label. According to the first environment brightness tag, the first camera acquisition constraint parameter tag and the first grayscale deviation threshold tag, until the Mth environment brightness tag, the Mth camera acquisition constraint parameter tag and the Mth grayscale deviation threshold tag, the environment constraint brightness, the camera acquisition constraint parameter and the grayscale deviation threshold are configured.

[0007] In the implementation method, the first surface quality index is traversed until the Nth surface quality index is subjected to grayscale recognition feature analysis, and the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold are obtained, including: setting the first environment test brightness and the first camera acquisition test parameter. According to the first environment test brightness and the first camera acquisition test parameter, grayscale acquisition of defective areas is performed on a number of first surface quality index defective samples to obtain a comparison grayscale set. According to the first environment test brightness and the first camera acquisition test parameter, grayscale acquisition of preset areas is performed on a number of first surface quality index healthy samples to obtain a reference grayscale set, wherein the coordinates of the preset area and the defective area are the same. The grayscale deviation of the comparison grayscale set is compared with the grayscale deviation of the reference grayscale set to obtain a grayscale deviation set. Minimum value sorting is performed on the grayscale deviation set to obtain a minimum grayscale deviation. When the minimum grayscale deviation is less than or equal to the recognizable grayscale accuracy threshold, the first environment test brightness and the first camera acquisition test parameter are updated to repeat the test. When the minimum grayscale deviation is greater than the recognizable grayscale accuracy threshold, the first environment test brightness is set to the first environment constraint brightness of the first surface quality index, the first camera acquisition test parameter is set to the first camera acquisition constraint parameter of the first surface quality index, and the minimum grayscale deviation is set to the first grayscale deviation threshold of the first surface quality index. Traverse and analyze until the Nth surface quality index, obtain the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold.

[0008] In an implementation, the grayscale deviation set is sorted for minimum value to obtain the minimum grayscale deviation, including: sorting the grayscale deviation set in order from small to large, extracting the grayscale deviation of the first quarter sequence number and the grayscale deviation of the third quarter sequence number. Subtracting the grayscale deviation of the first quarter sequence number from the grayscale deviation of the third quarter sequence number to obtain a first range. Subtracting 1.5 times the first range from the grayscale deviation of the first quarter sequence number to obtain a grayscale deviation lower limit. Extracting the minimum value of the grayscale deviation set that is greater than or equal to the grayscale deviation lower limit, and setting it as the minimum grayscale deviation.

[0009] In an implementation, clustering the first environment constraint brightness, the first camera acquisition constraint parameter, and the first grayscale deviation threshold until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter, and the Nth grayscale deviation threshold according to the environment constraint brightness and the camera acquisition constraint parameter to generate a first cluster quality indicator until the Mth cluster quality indicator includes: clustering the first environment constraint brightness, the first camera acquisition constraint parameter, and the first grayscale deviation threshold until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter, and the Nth grayscale deviation threshold according to the environment constraint brightness and the camera acquisition constraint parameter to obtain a first cluster quality indicator and a first cluster grayscale deviation threshold. Minimum value sorting is performed on the first cluster grayscale deviation threshold, and the grayscale deviation threshold is set as the first grayscale deviation threshold label of the first cluster quality indicator. Until the first environment constraint brightness, the first camera acquisition constraint parameter, and the first grayscale deviation threshold until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter, and the Nth grayscale deviation threshold are clustered according to the environment constraint brightness and the camera acquisition constraint parameter to obtain an Mth cluster quality indicator and an Mth cluster grayscale deviation threshold. The grayscale deviation threshold of the Mth cluster is sorted for minimum value, and is set as the Mth grayscale deviation threshold label of the Mth cluster quality indicator.

[0010] In an implementation, defect identification is performed on the first product grayscale image based on the grayscale deviation threshold to obtain a suspicious abnormal area, including: obtaining a healthy product grayscale image, wherein the healthy product grayscale image satisfies the environmental constraint brightness and the camera acquisition constraint parameter. The healthy product grayscale image is compared with the first product grayscale image to obtain grayscale deviation distribution information. Feature pixel point distribution information that satisfies the grayscale deviation threshold is extracted from the grayscale deviation distribution information. Neighborhood hierarchical clustering is performed on the feature pixel point distribution information to obtain the suspicious abnormal area.

[0011] In the implementation method, the characteristic pixel distribution information is subjected to neighborhood hierarchical clustering to obtain the suspected abnormal area, including: obtaining a first characteristic pixel coordinate and a second characteristic pixel coordinate. When the pixel coordinate distance between the first characteristic pixel coordinate and the second characteristic pixel coordinate is less than or equal to a pixel coordinate distance threshold, the first characteristic pixel coordinate and the second characteristic pixel coordinate are connected by a straight line to obtain straight line associated pixel coordinates. The straight line associated pixel coordinates are added to the first suspected abnormal area, and the first suspected abnormal area is added to the suspected abnormal area. Repeat clustering, and when the pixel coordinate distance between any two characteristic pixel coordinates in different suspected abnormal areas is greater than the pixel coordinate distance threshold, the suspected abnormal area is output.

[0012] This application also provides an industrial product quality detection system for artificial intelligence, including:

[0013] The quality index acquisition module is used to obtain the surface quality index of industrial products.

[0014] The constraint parameter acquisition module is used to perform grayscale recognition feature analysis on the surface quality index of the industrial product, generate environmental constraint brightness, and collect constraint parameters and grayscale deviation thresholds by a camera.

[0015] The product image acquisition module is used to control the camera to shoot according to the environmental constraint brightness and the camera acquisition constraint parameters to obtain a first product image when the industrial product to be inspected enters the preset area of ​​the carrier.

[0016] The grayscale image acquisition module is used to perform grayscale processing on the first product image to obtain a first product grayscale image.

[0017] The abnormal area acquisition module is used to perform defect identification on the first product grayscale image based on the grayscale deviation threshold to obtain a suspicious abnormal area.

[0018] The exception handling module is used to send the first product image to a cloud server when the number of the suspected abnormal areas is not equal to 0.

[0019] The no-abnormality processing module is used to mark the industrial product surface quality index as no-abnormality for the industrial product to be inspected when the number of the suspected abnormality areas is equal to 0.

[0020] The industrial product quality detection method and system for artificial intelligence proposed in this application are proposed to obtain the surface quality index of the industrial product. The grayscale recognition feature analysis of the surface quality index of the industrial product is performed to generate the environmental constraint brightness, the camera acquisition constraint parameter and the grayscale deviation threshold. When the industrial product to be inspected enters the preset area of ​​the carrier, the camera is controlled to shoot according to the environmental constraint brightness and the camera acquisition constraint parameter to obtain the first product image. The first product image is grayed to obtain the first product grayscale image. For the first product grayscale image, defect identification is performed based on the grayscale deviation threshold to obtain suspicious abnormal areas. When the number of the suspicious abnormal areas is not equal to 0, the first product image is sent to the cloud server. When the number of the suspicious abnormal areas is equal to 0, the industrial product surface quality index of the industrial product to be inspected is marked as normal. The technical problem that the industrial product quality detection method in the prior art has the large computing pressure of the cloud large model and the overall detection efficiency and accuracy are difficult to improve is solved. By deploying edge servers locally to perform preliminary processing and screening of grayscale images, some images are sent to the large cloud model for processing, thereby effectively reducing the computing pressure of the large cloud model and improving the overall detection efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solution of the embodiment of the present disclosure, the accompanying drawings of the embodiment of the present disclosure will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0022] Figure 1 A schematic diagram of the process flow of an industrial product quality detection method for artificial intelligence provided in an embodiment of the present application;

[0023] Figure 2 A schematic diagram of the structure of an industrial product quality inspection system for artificial intelligence provided in an embodiment of the present application.

[0024] Explanation of reference numerals: quality index acquisition module 11 , constraint parameter acquisition module 12 , product image acquisition module 13 , grayscale image acquisition module 14 , abnormal region acquisition module 15 , abnormality processing module 16 , no abnormality processing module 17 . DETAILED DESCRIPTION

[0025] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0026] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0027] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0028] The present application provides an industrial product quality detection method and system for artificial intelligence, such as Figure 1 As shown, the method includes:

[0029] The surface quality index of the industrial product is obtained, the grayscale recognition feature analysis is performed on the surface quality index of the industrial product, and the environmental constraint brightness, camera acquisition constraint parameters and grayscale deviation threshold are generated.

[0030] When the industrial product to be inspected enters the preset area of ​​the carrier, the camera is controlled to shoot according to the environmental constraint brightness and the camera acquisition constraint parameters to obtain a first product image.

[0031] Obtain the surface quality index of industrial products, which is the surface quality detection index corresponding to the industrial products, including but not limited to surface smoothness, glossiness, color difference, etc., and each industrial product has multiple surface quality detection indexes. Perform grayscale recognition feature analysis on the surface quality index of industrial products to generate environmental constraint brightness, camera acquisition constraint parameters and grayscale deviation thresholds, which are the preferred detection parameters corresponding to each surface quality detection index. When the industrial product to be inspected enters the preset area of ​​the carrier, i.e., the image detection area, the camera is controlled to shoot according to the environmental constraint brightness and the camera acquisition constraint parameters to obtain the first product image.

[0032] The method provided in the embodiment of the present application also includes:

[0033] The industrial product surface quality index includes a first surface quality index to an Nth surface quality index.

[0034] Traverse the first surface quality index until the Nth surface quality index to perform grayscale recognition feature analysis, obtain the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold, until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold.

[0035] According to the environmental constraint brightness and the camera acquisition constraint parameter, the first environmental constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold are clustered until the Nth environmental constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold to generate a first cluster quality indicator until the Mth cluster quality indicator.

[0036] The first cluster quality indicator has a first environment brightness label, a first camera acquisition constraint parameter label and a first grayscale deviation threshold label, and the Mth cluster quality indicator has an Mth environment brightness label, an Mth camera acquisition constraint parameter label and an Mth grayscale deviation threshold label.

[0037] According to the first environment brightness tag, the first camera acquisition constraint parameter tag and the first grayscale deviation threshold tag, until the Mth environment brightness tag, the Mth camera acquisition constraint parameter tag and the Mth grayscale deviation threshold tag, the environment constraint brightness, the camera acquisition constraint parameter and the grayscale deviation threshold are configured.

[0038] The grayscale recognition feature analysis is performed on the surface quality index of the industrial product to generate the environment constraint brightness, the camera acquisition constraint parameter and the grayscale deviation threshold, including: the surface quality index of the industrial product includes the first surface quality index until the Nth surface quality index. Subsequently, the first surface quality index is traversed until the Nth surface quality index is subjected to grayscale recognition feature analysis to obtain the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold. The first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold are the preferred monitoring parameters corresponding to the first surface quality index.

[0039] According to the environmental constraint brightness and the camera acquisition constraint parameter, the first environmental constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold are clustered until the Nth environmental constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold, and a first cluster quality indicator is generated until the Mth cluster quality indicator, that is, the parameters corresponding to each surface quality indicator are clustered, the same environmental constraint brightness and camera acquisition constraint parameter are clustered into one category, and the first cluster quality indicator is obtained until the Mth cluster quality indicator. Each cluster corresponds to an environmental constraint brightness and a camera acquisition constraint parameter, and M is a positive integer less than or equal to N. Among them, the first cluster quality indicator has a first environmental brightness label, a first camera acquisition constraint parameter label and a first grayscale deviation threshold label, and the Mth cluster quality indicator has an Mth environmental brightness label, an Mth camera acquisition constraint parameter label and an Mth grayscale deviation threshold label, and each grayscale deviation threshold label is the minimum grayscale deviation threshold among the grayscale deviation thresholds corresponding to the surface quality indicator. According to the first environment brightness tag, the first camera acquisition constraint parameter tag and the first grayscale deviation threshold tag, until the Mth environment brightness tag, the Mth camera acquisition constraint parameter tag and the Mth grayscale deviation threshold tag, the environment constraint brightness, the camera acquisition constraint parameter and the grayscale deviation threshold are configured.

[0040] The method provided in the embodiment of the present application also includes:

[0041] Set the first environment test brightness and the first camera acquisition test parameters.

[0042] According to the first environment test brightness and the first camera acquisition test parameter, grayscale acquisition of defect areas is performed on a plurality of first surface quality index defect samples to obtain a comparison grayscale set.

[0043] According to the first environment test brightness and the first camera acquisition test parameters, grayscale acquisition is performed on a preset area of ​​several healthy samples of the first surface quality index to obtain a reference grayscale set, wherein the coordinates of the preset area and the defective area are the same.

[0044] The grayscale deviations of the comparison grayscale set and the reference grayscale set are compared to obtain a grayscale deviation set.

[0045] Minimum value sorting is performed on the grayscale deviation set to obtain the minimum grayscale deviation.

[0046] When the minimum grayscale deviation is less than or equal to the recognizable grayscale accuracy threshold, the first environment test brightness and the first camera acquisition test parameters are updated to repeat the test.

[0047] When the minimum grayscale deviation is greater than the recognizable grayscale accuracy threshold, the first environment test brightness is set to the first environment constraint brightness of the first surface quality index, the first camera acquisition test parameter is set to the first camera acquisition constraint parameter of the first surface quality index, and the minimum grayscale deviation is set to the first grayscale deviation threshold of the first surface quality index.

[0048] Traverse and analyze until the Nth surface quality index is reached, obtain the Nth environmental constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold.

[0049] Traversing the first surface quality index until the Nth surface quality index performs grayscale recognition feature analysis, and obtains the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold, including: setting the first environment test brightness and the first camera acquisition test parameter, the first environment test brightness and the first camera acquisition test parameter are test parameters randomly set based on the current environment, so as to find the acquisition parameters that match the current device to achieve the best defect recognition effect.

[0050] According to the first environment test brightness and the first camera acquisition test parameters, the defect area grayscale of several first surface quality indicator defect samples is collected to obtain a comparison grayscale set. According to the first environment test brightness and the first camera acquisition test parameters, the preset area grayscale of several first surface quality indicator healthy samples is collected to obtain a reference grayscale set, wherein the acquisition position coordinates of the preset area and the defect area are the same. Further, the grayscale deviation of the comparison grayscale set is compared with the grayscale deviation of the reference grayscale set to obtain the grayscale deviation of the surface quality indicator defect sample and the healthy sample, and obtain a grayscale deviation set. The grayscale deviation set is sorted for the minimum value to obtain the minimum grayscale deviation. The minimum grayscale deviation obtained at this time is the minimum judgment threshold of the grayscale deviation of the defective sample and the healthy sample, that is, the discrimination device needs to identify the minimum judgment threshold to achieve accurate judgment of the defect, otherwise the discrimination device cannot identify the corresponding defect grayscale deviation.

[0051] When the minimum grayscale deviation is less than or equal to the recognizable grayscale accuracy threshold, that is, the minimum recognizable grayscale deviation of the device, the device cannot recognize the minimum grayscale deviation caused by the defect, so the corresponding first environment test brightness and the first camera acquisition test parameter cannot meet the recognition requirements, and the first environment test brightness and the first camera acquisition test parameter are updated for repeated testing, and the grayscale deviation is repeatedly obtained for secondary verification. When the minimum grayscale deviation is greater than the recognizable grayscale accuracy threshold, the device can accurately identify the minimum grayscale deviation caused by the defect, and the first environment test brightness is set to the first environment constraint brightness of the first surface quality index, the first camera acquisition test parameter is set to the first camera acquisition constraint parameter of the first surface quality index, and the minimum grayscale deviation is set to the first grayscale deviation threshold of the first surface quality index. Finally, the traversal analysis is performed until the Nth surface quality index is completed, and the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold are obtained.

[0052] The method provided in the embodiment of the present application also includes:

[0053] The grayscale deviation set is sorted in order from small to large, and a quarter of the grayscale deviation and a three-quarters of the grayscale deviation are extracted.

[0054] The first range is obtained by subtracting one quarter of the serial number grayscale from the three quarters of the serial number grayscale deviation.

[0055] The lower limit value of the grayscale deviation is obtained by subtracting 1.5 times the first range from the one-quarter grayscale deviation.

[0056] A minimum value of the grayscale deviation set that is greater than or equal to the grayscale deviation lower limit is extracted and set as the grayscale deviation minimum value.

[0057] The grayscale deviation set is subjected to minimum value sorting to obtain the minimum grayscale deviation, including: sorting the grayscale deviation set from small to large, extracting the grayscale deviation of the quarter sequence number and the grayscale deviation of the three-quarter sequence number, the exemplary grayscale deviation set contains 40 data, the grayscale deviation of the quarter sequence number extracted is the grayscale deviation of the 10th sorting sequence number, and the grayscale deviation of the three-quarter sequence number is the grayscale deviation of the 30th sorting sequence number. The first range is obtained by subtracting the grayscale of the quarter sequence number from the grayscale deviation of the three-quarter sequence number. The lower limit value of the grayscale deviation is obtained by subtracting 1.5 times the first range from the grayscale deviation of the quarter sequence number. The minimum value of the grayscale deviation set that is greater than or equal to the lower limit value of the grayscale deviation is extracted and set as the minimum grayscale deviation value.

[0058] The method provided in the embodiment of the present application also includes:

[0059] According to the environmental constraint brightness and the camera acquisition constraint parameter, the first environmental constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold are clustered until the N-th environmental constraint brightness, the N-th camera acquisition constraint parameter and the N-th grayscale deviation threshold to obtain a first cluster quality indicator and a first cluster grayscale deviation threshold.

[0060] Minimum value sorting is performed on the first cluster grayscale deviation threshold, and the minimum value is set as the first grayscale deviation threshold label of the first cluster quality indicator.

[0061] Until the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold are clustered according to the environment constraint brightness and the camera acquisition constraint parameter, until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold are clustered to obtain the Mth cluster quality indicator and the Mth cluster grayscale deviation threshold.

[0062] The grayscale deviation threshold of the Mth cluster is sorted for minimum value, and is set as the Mth grayscale deviation threshold label of the Mth cluster quality indicator.

[0063] Generate the first cluster quality indicator until the Mth cluster quality indicator, including: clustering the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold according to the environment constraint brightness and the camera acquisition constraint parameter, and obtain the first cluster quality indicator and the corresponding first cluster grayscale deviation threshold. Further, perform minimum value sorting on the first cluster grayscale deviation threshold, the minimum value sorting scheme is consistent with the grayscale deviation minimum value acquisition scheme, and the sorting result is set as the first grayscale deviation threshold label of the first cluster quality indicator. Continue clustering until the clustering of all quality indicators environment constraint brightness and camera acquisition constraint parameters is completed, and obtain the Mth cluster quality indicator and the Mth cluster grayscale deviation threshold. Perform minimum value sorting on the Mth cluster grayscale deviation threshold and set it as the Mth grayscale deviation threshold label of the Mth cluster quality indicator.

[0064] The first product image is grayscaled to obtain a first product grayscale image.

[0065] Defect identification is performed on the first product grayscale image based on the grayscale deviation threshold to obtain a suspicious abnormal area.

[0066] When the number of the suspected abnormal areas is not equal to 0, the first product image is sent to a cloud server.

[0067] When the number of the suspected abnormal areas is equal to 0, the industrial product to be inspected is marked as having no abnormality in the industrial product surface quality index.

[0068] After acquiring the first product image, the first product image is grayed to obtain the first product gray image. Subsequently, the first product gray image is defect identified based on the gray deviation threshold, and it is determined whether there is an area greater than or equal to the gray deviation threshold in the gray image, and the position of the corresponding area is obtained to obtain the suspicious abnormal area. When the number of the suspicious abnormal areas is not equal to 0, that is, the suspicious abnormal area exists, and it can be judged that the corresponding industrial product may have defects, the first product image is sent to the cloud server, and the product image is further defect-detected by the cloud server, and the final defect specific detection result is determined by the cloud server. When the number of the suspicious abnormal areas is equal to 0, the corresponding industrial product to be inspected does not have a suspicious abnormal area, and there is no defect, and the industrial product surface quality index of the industrial product to be inspected is marked as abnormal. The technical problem that the industrial product quality detection method in the prior art has a large computing pressure on the cloud large model and the overall detection efficiency and accuracy are difficult to improve is solved. By deploying an edge server locally to perform preliminary processing and screening on the gray image, some images are sent to the cloud large model for processing, thereby effectively reducing the computing pressure of the cloud large model and improving the overall detection efficiency and accuracy.

[0069] The method provided in the embodiment of the present application also includes:

[0070] A healthy grayscale image of the product is obtained, wherein the healthy grayscale image of the product meets the environmental constraint brightness and the camera acquisition constraint parameters.

[0071] The healthy grayscale image of the product is compared with the first product grayscale image to obtain grayscale deviation distribution information.

[0072] Feature pixel point distribution information satisfying the grayscale deviation threshold is extracted from the grayscale deviation distribution information.

[0073] Perform neighborhood hierarchical clustering on the characteristic pixel distribution information to obtain the suspicious abnormal area.

[0074] For the first product grayscale image, defect identification is performed based on the grayscale deviation threshold to obtain a suspicious abnormal area, including: obtaining a healthy product grayscale image, wherein the healthy product grayscale image satisfies the environmental constraint brightness and the camera acquisition constraint parameters. Grayscale values ​​of the healthy product grayscale image and the first product grayscale image are compared to obtain grayscale deviation distribution information. Feature pixel point distribution information that satisfies the grayscale deviation threshold is extracted from the grayscale deviation distribution information, that is, feature pixel point distribution information whose grayscale value deviation is greater than or equal to the grayscale deviation threshold is extracted from the grayscale deviation distribution information. Neighborhood hierarchical clustering is performed on the feature pixel point distribution information to obtain the suspicious abnormal area.

[0075] The method provided in the embodiment of the present application also includes:

[0076] The first characteristic pixel coordinates and the second characteristic pixel coordinates are obtained.

[0077] When the pixel coordinate distance between the first feature pixel coordinate and the second feature pixel coordinate is less than or equal to a pixel coordinate distance threshold, the first feature pixel coordinate and the second feature pixel coordinate are connected by a straight line to obtain straight line associated pixel coordinates.

[0078] The pixel coordinates of the straight line are associated and added into the first suspicious abnormal region, and the first suspicious abnormal region is added into the suspicious abnormal region.

[0079] The clustering is repeated, and when the pixel coordinate distance between any two feature pixel coordinates in different suspicious abnormal regions is greater than the pixel coordinate distance threshold, the suspicious abnormal region is output.

[0080] Performing neighborhood hierarchical clustering on the characteristic pixel distribution information to obtain the suspected abnormal area includes: obtaining a first characteristic pixel coordinate and a second characteristic pixel coordinate, wherein the first characteristic pixel coordinate and the second characteristic pixel coordinate are any two characteristic pixel coordinates in the characteristic pixel distribution information. When the pixel coordinate distance between the first characteristic pixel coordinate and the second characteristic pixel coordinate is less than or equal to the pixel coordinate distance threshold, connecting the first characteristic pixel coordinate and the second characteristic pixel coordinate by a straight line to obtain straight line-associated pixel coordinates, wherein the straight line-associated pixel coordinates are the pixel positions covered by the connecting straight line. Adding the straight line-associated pixel coordinates to the first suspected abnormal area, and adding the first suspected abnormal area to the suspected abnormal area. Repeating clustering, when the pixel coordinate distance between any two characteristic pixel coordinates in different suspected abnormal areas is greater than the pixel coordinate distance threshold, outputting the suspected abnormal area.

[0081] In the above, refer to Figure 1The industrial product quality detection method for artificial intelligence according to an embodiment of the present invention is described in detail. Figure 2 An industrial product quality inspection system for artificial intelligence according to an embodiment of the present invention is described.

[0082] The industrial product quality inspection system for artificial intelligence according to the embodiment of the present invention solves the technical problem that the industrial product quality inspection method in the prior art has high computational pressure of large cloud models and is difficult to improve the overall detection efficiency and accuracy. By deploying edge servers locally to perform preliminary processing and screening on grayscale images, some images are sent to the large cloud model for processing, thereby effectively reducing the computational pressure of the large cloud model and improving the overall detection efficiency and accuracy. The industrial product quality inspection system for artificial intelligence includes: a quality indicator acquisition module 11, a constraint parameter acquisition module 12, a product image acquisition module 13, a grayscale image acquisition module 14, an abnormal area acquisition module 15, an abnormality processing module 16, and a non-abnormality processing module 17.

[0083] The quality index acquisition module 11 is used to obtain the surface quality index of the industrial product.

[0084] The constraint parameter acquisition module 12 is used to perform grayscale recognition feature analysis on the surface quality index of the industrial product, generate environmental constraint brightness, and collect constraint parameters and grayscale deviation thresholds by a camera.

[0085] The product image acquisition module 13 is used to control the camera to shoot according to the environmental constraint brightness and the camera acquisition constraint parameters to obtain a first product image when the industrial product to be inspected enters the preset area of ​​the carrier.

[0086] The grayscale image acquisition module 14 is used to perform grayscale processing on the first product image to obtain a first product grayscale image.

[0087] The abnormal region acquisition module 15 is used to perform defect identification on the first product grayscale image based on the grayscale deviation threshold to obtain a suspicious abnormal region.

[0088] The exception handling module 16 is configured to send the first product image to a cloud server when the number of the suspected abnormal areas is not equal to 0.

[0089] The no-abnormality processing module 17 is used to mark the industrial product surface quality index as no-abnormality for the industrial product to be inspected when the number of the suspected abnormality areas is equal to 0.

[0090] The specific configuration of the constraint parameter acquisition module 12 will be described in detail below. The grayscale recognition feature analysis is performed on the surface quality index of the industrial product to generate the environmental constraint brightness, the camera acquisition constraint parameter and the grayscale deviation threshold, including: the surface quality index of the industrial product includes the first surface quality index until the Nth surface quality index. The first surface quality index is traversed until the Nth surface quality index is subjected to grayscale recognition feature analysis to obtain the first environmental constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold until the Nth environmental constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold. According to the environmental constraint brightness and the camera acquisition constraint parameter, the first environmental constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold until the Nth environmental constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold are clustered to generate the first cluster quality index until the Mth cluster quality index. The first cluster quality indicator has a first environment brightness tag, a first camera acquisition constraint parameter tag, and a first grayscale deviation threshold tag, and the Mth cluster quality indicator has an Mth environment brightness tag, an Mth camera acquisition constraint parameter tag, and an Mth grayscale deviation threshold tag. According to the first environment brightness tag, the first camera acquisition constraint parameter tag, and the first grayscale deviation threshold tag, until the Mth environment brightness tag, the Mth camera acquisition constraint parameter tag, and the Mth grayscale deviation threshold tag, the environment constraint brightness, the camera acquisition constraint parameter, and the grayscale deviation threshold are configured.

[0091] The specific configuration of the constraint parameter acquisition module 12 will be described in detail below. The constraint parameter acquisition module 12 further includes: traversing the first surface quality index until the Nth surface quality index to perform grayscale recognition feature analysis, obtain the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold, including: setting the first environment test brightness and the first camera acquisition test parameter. According to the first environment test brightness and the first camera acquisition test parameter, a plurality of first surface quality index defect samples are subjected to defect area grayscale acquisition to obtain a comparison grayscale set. According to the first environment test brightness and the first camera acquisition test parameter, a plurality of first surface quality index healthy samples are subjected to preset area grayscale acquisition to obtain a reference grayscale set, wherein the coordinates of the preset area and the defect area are the same. Compare the grayscale deviation of the comparison grayscale set with the reference grayscale set to obtain a grayscale deviation set. Perform minimum value sorting on the grayscale deviation set to obtain the minimum grayscale deviation. When the minimum grayscale deviation is less than or equal to the recognizable grayscale accuracy threshold, update the first environment test brightness and the first camera acquisition test parameter to repeat the test. When the minimum grayscale deviation is greater than the recognizable grayscale accuracy threshold, the first environment test brightness is set to the first environment constraint brightness of the first surface quality index, the first camera acquisition test parameter is set to the first camera acquisition constraint parameter of the first surface quality index, and the minimum grayscale deviation is set to the first grayscale deviation threshold of the first surface quality index. Traverse and analyze until the Nth surface quality index, obtain the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold.

[0092] Below, the specific configuration of the constraint parameter acquisition module 12 will be described in detail. The constraint parameter acquisition module 12 may further include: performing minimum value sorting on the grayscale deviation set to obtain the minimum grayscale deviation, including: sorting the grayscale deviation set in order from small to large, extracting the one-quarter sequence grayscale deviation and the three-quarter sequence grayscale deviation. Subtracting the one-quarter sequence grayscale from the three-quarter sequence grayscale deviation to obtain the first range. Subtracting 1.5 times the first range from the one-quarter sequence grayscale deviation to obtain the grayscale deviation lower limit. Extract the minimum value of the grayscale deviation set that is greater than or equal to the grayscale deviation lower limit, and set it as the grayscale deviation minimum value.

[0093] The specific configuration of the constraint parameter acquisition module 12 will be described in detail below. According to the environment constraint brightness and the camera acquisition constraint parameter, the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold are clustered until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold, and a first cluster quality indicator is generated until the Mth cluster quality indicator, including: according to the environment constraint brightness and the camera acquisition constraint parameter, the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold are clustered until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold are clustered to obtain a first cluster quality indicator and a first cluster grayscale deviation threshold. The first cluster grayscale deviation threshold is sorted for the minimum value and set as the first grayscale deviation threshold label of the first cluster quality indicator. Until the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold are clustered according to the environment constraint brightness and the camera acquisition constraint parameter, until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold are clustered to obtain the Mth cluster quality index and the Mth cluster grayscale deviation threshold. The Mth cluster grayscale deviation threshold is sorted for the minimum value and is set as the Mth grayscale deviation threshold label of the Mth cluster quality index.

[0094] The specific configuration of the abnormal area acquisition module 15 will be described in detail below. For the first product grayscale image, defect identification is performed based on the grayscale deviation threshold to obtain a suspicious abnormal area, including: obtaining a healthy product grayscale image, wherein the healthy product grayscale image satisfies the environmental constraint brightness and the camera acquisition constraint parameters. The healthy product grayscale image is compared with the first product grayscale image to obtain grayscale deviation distribution information. Feature pixel point distribution information that meets the grayscale deviation threshold is extracted from the grayscale deviation distribution information. The feature pixel point distribution information is subjected to neighborhood hierarchical clustering to obtain the suspicious abnormal area.

[0095] Below, the specific configuration of the abnormal area acquisition module 15 will be described in detail. The characteristic pixel distribution information is subjected to neighborhood hierarchical clustering to obtain the suspected abnormal area, including: obtaining a first characteristic pixel coordinate and a second characteristic pixel coordinate. When the pixel coordinate distance between the first characteristic pixel coordinate and the second characteristic pixel coordinate is less than or equal to a pixel coordinate distance threshold, the first characteristic pixel coordinate and the second characteristic pixel coordinate are connected by a straight line to obtain straight line associated pixel coordinates. The straight line associated pixel coordinates are added to the first suspected abnormal area, and the first suspected abnormal area is added to the suspected abnormal area. Repeat clustering, and when the pixel coordinate distance between any two characteristic pixel coordinates in different suspected abnormal areas is greater than the pixel coordinate distance threshold, the suspected abnormal area is output.

[0096] The industrial product quality detection system for artificial intelligence provided by the embodiment of the present invention can execute the industrial product quality detection method for artificial intelligence provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0097] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or the server, and the various units and modules included are only divided according to functional logic, but are not limited to the above divisions, as long as the corresponding functions can be achieved. In addition, the specific names of the functional units are only for the convenience of distinguishing from each other, and are not used to limit the scope of protection of the present invention.

[0098] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. An industrial product quality detection method for artificial intelligence, characterized in that: Applied to edge servers, including: Obtain surface quality indicators of industrial products; Performing grayscale recognition feature analysis on the surface quality index of the industrial product, generating environmental constraint brightness, camera acquisition constraint parameters and grayscale deviation threshold; When the industrial product to be inspected enters the preset area of ​​the carrier, the camera is controlled to shoot according to the environmental constraint brightness and the camera acquisition constraint parameters to obtain a first product image; Performing grayscale processing on the first product image to obtain a first product grayscale image; Performing defect identification on the first product grayscale image based on the grayscale deviation threshold to obtain a suspicious abnormal area; When the number of the suspected abnormal areas is not equal to 0, sending the first product image to a cloud server; When the number of the suspected abnormal areas is equal to 0, the industrial product to be inspected is marked as having no abnormality in the industrial product surface quality index; The grayscale recognition feature analysis is performed on the surface quality index of the industrial product to generate the environmental constraint brightness, camera acquisition constraint parameters and grayscale deviation threshold, including: The industrial product surface quality index includes a first surface quality index to an Nth surface quality index; Traversing the first surface quality index until the Nth surface quality index performs grayscale recognition feature analysis, and obtains a first environmental constraint brightness, a first camera acquisition constraint parameter, and a first grayscale deviation threshold, until an Nth environmental constraint brightness, an Nth camera acquisition constraint parameter, and an Nth grayscale deviation threshold; According to the environment constraint brightness and the camera acquisition constraint parameter, clustering the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold, up to the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold, to generate a first cluster quality indicator up to an Mth cluster quality indicator; The first cluster quality indicator has a first environment brightness label, a first camera acquisition constraint parameter label and a first grayscale deviation threshold label, and the Mth cluster quality indicator has an Mth environment brightness label, an Mth camera acquisition constraint parameter label and an Mth grayscale deviation threshold label; According to the first environment brightness tag, the first camera acquisition constraint parameter tag and the first grayscale deviation threshold tag, up to the Mth environment brightness tag, the Mth camera acquisition constraint parameter tag and the Mth grayscale deviation threshold tag, configure the environment constraint brightness, the camera acquisition constraint parameter and the grayscale deviation threshold; Traversing the first surface quality index until the Nth surface quality index performs grayscale recognition feature analysis, obtains the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold, including: Setting the first environment test brightness and the first camera acquisition test parameters; Perform defect area grayscale acquisition on a plurality of first surface quality index defect samples according to the first environment test brightness and the first camera acquisition test parameter to obtain a comparison grayscale set; According to the first environment test brightness and the first camera acquisition test parameter, grayscale acquisition is performed on a plurality of healthy samples of the first surface quality index to obtain a reference grayscale set, wherein the coordinates of the preset area and the defective area are the same; Comparing the grayscale deviations of the comparison grayscale set and the reference grayscale set to obtain a grayscale deviation set; Performing minimum value sorting on the grayscale deviation set to obtain the minimum grayscale deviation; When the grayscale deviation minimum value is less than or equal to the recognizable grayscale accuracy threshold, updating the first environment test brightness and the first camera acquisition test parameters to repeat the test; When the minimum grayscale deviation is greater than the recognizable grayscale accuracy threshold, the first environment test brightness is set to the first environment constraint brightness of the first surface quality index, the first camera acquisition test parameter is set to the first camera acquisition constraint parameter of the first surface quality index, and the minimum grayscale deviation is set to the first grayscale deviation threshold of the first surface quality index; Traversing and analyzing until the Nth surface quality index is obtained, the Nth environmental constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold are obtained; Performing defect identification on the first product grayscale image based on the grayscale deviation threshold to obtain a suspicious abnormal area includes: Obtaining a healthy grayscale image of the product, wherein the healthy grayscale image of the product satisfies the environmental constraint brightness and the camera acquisition constraint parameters; Comparing the healthy grayscale image of the product with the grayscale image of the first product to obtain grayscale deviation distribution information; Extracting characteristic pixel point distribution information satisfying the grayscale deviation threshold from the grayscale deviation distribution information; Performing neighborhood hierarchical clustering on the characteristic pixel distribution information to obtain the suspected abnormal area; Performing neighborhood hierarchical clustering on the characteristic pixel distribution information to obtain the suspicious abnormal area includes: Obtaining first characteristic pixel coordinates and second characteristic pixel coordinates; When the pixel coordinate distance between the first characteristic pixel coordinate and the second characteristic pixel coordinate is less than or equal to a pixel coordinate distance threshold, connecting the first characteristic pixel coordinate and the second characteristic pixel coordinate by a straight line to obtain straight line associated pixel coordinates; Associating the pixel coordinates of the straight line into the first suspected abnormal area, and adding the first suspected abnormal area into the suspected abnormal area; The clustering is repeated, and when the pixel coordinate distance between any two feature pixel coordinates in different suspicious abnormal regions is greater than the pixel coordinate distance threshold, the suspicious abnormal region is output.

2. The industrial product quality detection method for artificial intelligence according to claim 1, characterized in that: Performing minimum value sorting on the grayscale deviation set to obtain the minimum grayscale deviation, including: The grayscale deviation set is sorted in order from small to large, and a grayscale deviation with a one-quarter sequence number and a grayscale deviation with a three-quarter sequence number are extracted; Subtract one quarter of the serial number grayscale from the three quarters of the serial number grayscale deviation to obtain a first range; Subtract 1.5 times the first range from the one-quarter grayscale deviation to obtain a grayscale deviation lower limit; A minimum value of the grayscale deviation set that is greater than or equal to the grayscale deviation lower limit is extracted and set as the grayscale deviation minimum value.

3. The industrial product quality detection method for artificial intelligence according to claim 1, characterized in that: According to the environment constraint brightness and the camera acquisition constraint parameter, clustering the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold, and generating a first cluster quality indicator until an Mth cluster quality indicator, including: According to the environment constraint brightness and the camera acquisition constraint parameter, clustering the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold, until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold, to obtain a first cluster quality indicator and a first cluster grayscale deviation threshold; performing minimum value sorting on the grayscale deviation threshold of the first cluster, and setting the minimum value as the first grayscale deviation threshold label of the first cluster quality indicator; Until the first environment constraint brightness, the first camera acquisition constraint parameter and the first grayscale deviation threshold are clustered according to the environment constraint brightness and the camera acquisition constraint parameter, until the Nth environment constraint brightness, the Nth camera acquisition constraint parameter and the Nth grayscale deviation threshold are clustered to obtain an Mth cluster quality indicator and an Mth cluster grayscale deviation threshold; The grayscale deviation threshold of the Mth cluster is sorted for minimum value, and is set as the Mth grayscale deviation threshold label of the Mth cluster quality indicator.

4. An industrial product quality inspection system for artificial intelligence, characterized in that: The system is used to execute the method according to any one of claims 1 to 3, comprising: A quality index acquisition module is used to obtain the surface quality index of industrial products; A constraint parameter acquisition module is used to perform grayscale recognition feature analysis on the surface quality index of the industrial product, generate environmental constraint brightness, and collect constraint parameters and grayscale deviation thresholds by a camera; A product image acquisition module, used for controlling the camera to shoot according to the environmental constraint brightness and the camera acquisition constraint parameters to obtain a first product image when the industrial product to be inspected enters the preset area of ​​the carrier; A grayscale image acquisition module, used for performing grayscale processing on the first product image to obtain a first product grayscale image; an abnormal region acquisition module, used to perform defect identification on the first product grayscale image based on the grayscale deviation threshold to obtain a suspicious abnormal region; an exception handling module, configured to send the first product image to a cloud server when the number of the suspected abnormal areas is not equal to 0; The no-abnormality processing module is used to mark the industrial product surface quality index as no-abnormality for the industrial product to be inspected when the number of the suspected abnormality areas is equal to 0.

Images