Intelligent big data monitoring system and method for visual inspection

Abstract

The application relates to the technical field of visual detection, in particular to an intelligent big data supervision system and method for visual detection, which comprises a product data acquisition module, a database, a detection data analysis module, a visual detection supervision module and a product detection module, historical detection data of products subjected to visual detection are collected through the product data acquisition module, all the collected data are transmitted to the database, all the collected data are stored through the database, the historical detection data are analyzed through the detection data analysis module, a necessary degree reference value for pre-processing of product images is set, the necessary degree of pre-processing of current product images is analyzed through the visual detection supervision module, target products are selected out, and the current product is subjected to visual detection through the product detection module, so that the feature extraction speed during visual detection of the product is accelerated, and the visual detection efficiency is further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of visual inspection, in particular to an intelligent big data supervision system and method for visual inspection. BACKGROUND

[0002] Visual inspection refers to converting a target to be taken into an image by a machine vision product, i.e., an image taking device, transmitting the image to a dedicated image processing system, and performing various operations on the image processing system to extract features of the target, and then controlling the action of the equipment on site according to the result of the judgment. In the application of detecting defects, the product is detected for defects by comparing the features of the target to be taken with the features of the target with defects, which has immeasurable value in preventing defective products from being delivered to consumers.

[0003] However, when performing visual inspection on products, there are still some problems: the product has a specific defect detection area, and after collecting the product image, the existing method is to directly transmit the collected product image to the image processing system for defect detection. There may be a large area of unnecessary detection area in the directly transmitted image. Directly transmitting the image without pre-removing the unnecessary detection area, especially in the face of a large number of product defect detection work, cannot speed up the feature extraction speed to further improve the visual inspection efficiency.

[0004] Therefore, people need an intelligent big data supervision system and method for visual inspection to solve the above problems. SUMMARY

[0005] The present application aims to provide an intelligent big data supervision system and method for visual inspection to solve the problems raised in the background.

[0006] In order to solve the above technical problems, the present application provides the following technical scheme: an intelligent big data supervision system for visual inspection, the system comprising: a product data acquisition module, a database, a detection data analysis module, a visual inspection supervision module and a product detection module;

[0007] The output end of the product data acquisition module is connected to the input end of the database, the output end of the database is connected to the input end of the detection data analysis module, the output end of the detection data analysis module is connected to the input end of the visual inspection supervision module, and the output end of the visual inspection supervision module is connected to the input end of the product detection module;

[0008] The product data acquisition module is used to collect the historical detection data of the product after visual inspection, and transmit all the collected data to the database;

[0009] The database is used to store all the collected data;

[0010] The detection data analysis module is configured to analyze historical detection data and set a necessary degree reference value for pre-processing of product images;

[0011] The visual detection supervision module is configured to analyze the necessary degree of pre-processing of a current product image and select a target product;

[0012] The product detection module is configured to perform visual detection on a current product.

[0013] Further, the product data acquisition module comprises a detection data acquisition unit, a detection point position acquisition unit and a product quantity acquisition unit;

[0014] The output ends of the detection data acquisition unit, the detection point position acquisition unit and the product quantity acquisition unit are connected to the input end of the database;

[0015] The detection data acquisition unit is configured to acquire detection point quantity data of products that have been subjected to visual detection in the past and image data of different types of products that have been photographed;

[0016] The detection point position acquisition unit is configured to establish a two-dimensional coordinate system with the center of a photographed image as the origin and acquire position coordinate information of all detection points;

[0017] The product quantity acquisition unit is configured to acquire quantity information of different types of products that have been detected in the past, and the detection point quantity and detection point position of the same type of product are the same.

[0018] Further, the detection data analysis module comprises a historical data analysis unit and a reference data setting unit;

[0019] The input end of the historical data analysis unit is connected to the output end of the database, and the output end of the historical data analysis unit is connected to the input end of the reference data setting unit;

[0020] The historical data analysis unit is configured to retrieve product image data and detection point position coordinate information that have been photographed during visual detection in the past, delimit a minimum effective detection area of different types of products, cover all detection points of the corresponding product in the minimum effective detection area, and estimate the area of an ineffective detection area of different types of products;

[0021] The reference data setting unit is configured to set a necessary degree reference value for pre-processing of an image after photographing a product image according to the area of the ineffective detection area and the necessary degree of pre-processing of the image after photographing the product image according to historical detection data analysis.

[0022] Further, the visual detection supervision module comprises a product image photographing unit, a processing necessity analysis unit and a target product selection unit;

[0023] An output end of the product image shooting unit is connected to an input end of the processing necessity analysis unit, and an output end of the processing necessity analysis unit and the reference data setting unit is connected to an input end of the target product selection unit;

[0024] The product image shooting unit is configured to shoot an image of a current product;

[0025] The processing necessity analysis unit is configured to analyze a necessity degree of pre-processing of the image of the current product;

[0026] The target product selection unit is configured to compare the necessity degree of pre-processing of the image of the current product with a necessity degree reference value, and if the necessity degree exceeds the reference value, the current product is selected as the target product; otherwise, the image of the current product is directly subjected to feature extraction, matched with features of a defective product, and the current product is subjected to defect detection.

[0027] Further, the product detection module comprises an image pre-processing unit and a visual detection unit;

[0028] An output end of the target product selection unit is connected to an input end of the image pre-processing unit, and an output end of the image pre-processing unit is connected to an input end of the visual detection unit;

[0029] The image pre-processing unit is configured to intercept a minimum effective detection area in the image of the target product;

[0030] The visual detection unit is configured to extract features of the intercepted image by using an image processing system, match the features with features of a defective product, and detect defects of the target product.

[0031] An intelligent big data supervision method for visual detection, comprising the following steps:

[0032] S1: collecting historical detection data of products subjected to visual detection;

[0033] S2: estimating areas of invalid detection regions of different types of products;

[0034] S3: setting a necessity degree reference value of pre-processing of product images according to the areas of the invalid detection regions and the historical detection data;

[0035] S4: analyzing a necessity degree of pre-processing of an image of a current product and selecting a target product;

[0036] S5: visual detection of the current product.

[0037] Further, in step S1: a set of detection point quantities of different types of products that have been visually detected in the past is collected as E={E1, E2, …, E m} and the detection point quantities are all greater than 1, wherein a total set of quantities of each type of product that needs to be detected when the corresponding type of product is detected in the past is collected as B={B1, B2, …, B m}, product images are collected when the m types of products are visually detected, a two-dimensional coordinate system is established with the center of the captured image as the origin, and the position coordinate information of all detection points is collected.

[0038] Further, in step S2: the product images captured when the m types of products are visually detected and the position coordinate information of the detection points are retrieved, the area of the captured image is obtained as X, and the minimum effective detection area covering all detection points of a random type of product is obtained by using the random increment method as: a circular area with (a, b) as the center and r i as the radius, and a set of radii of the minimum effective detection areas of the m types of products is obtained as r={r1, r2, …, r i , …, r m}, according to the formula Z i =X-π*r i 2 The area Z i of the invalid detection area of a random type of product is estimated, wherein X>π*r i 2 , and a set of areas of the invalid detection areas of the m types of products is obtained as Z={Z1, Z2, …, Z i , …, Z m} by using the same calculation method;

[0039] The random increment method here is an algorithm for finding the smallest circle that can cover a group of points on a plane, and the random increment algorithm is an important algorithm in computational geometry. The essence is to transform a problem into a sub-problem that is just one layer smaller in size, which has the effect of low time complexity of the algorithm. The random increment method is applied to solve the minimum circle covering problem, which can effectively reduce the time complexity of finding the smallest circle that covers all detection points;

[0040] The detection point information of different products that have been visually detected is collected by using big data technology, and the purpose of dividing the smallest circle covering all detection points of the product is to judge the area that does not need to be detected when the corresponding product is detected, i.e. the area of the invalid detection area. The area of the invalid detection area is used as one of the judgment bases for setting the pre-processing of the current product image, which is conducive to selecting the appropriate product and pre-processing the product image. Only the effective detection area image is transmitted, which is conducive to speeding up the feature extraction speed of the image.

[0041] Furthermore, in step S3: the necessity W for preprocessing a random product image is calculated according to the following formula. i :

[0042]

[0043] Among them, E i B represents the number of testing points for a random product. i This represents the total number of corresponding products that need to be detected when detecting a random product in the past. The set of necessary preprocessing degrees for images of m products is then obtained as W = {W1, W2, ..., W...}. i ,…,W m Arrange m types of products in ascending order of necessity. Divide the arranged products into k groups. Preprocess the images of products in the first group so that their necessity is less than that of the next group. The set of average necessity values ​​for the k groups of products after random grouping is M = {M1, M2, ..., M}. e M k According to the formula Calculate the goodness F of a random grouping method. j , of which M e Let F represent the mean necessity of the products in group e after being grouped according to a random grouping method, where e = 1, 2, ..., k, j = 1, 2, ..., g. There are g grouping methods in total. The set of goodness of the g grouping methods obtained by the same calculation method is F = {F1, F2, ..., F...}. j F g To obtain the grouping results with the highest quality, the set of the average necessity values ​​for each product group in the corresponding grouping results is obtained as M' = {M1', M2', ..., M}. k Set the reference value for the necessity of preprocessing product images to M. k ';

[0044] The historical data is used as the reference data for judging the pre-processing of the current product image, in order to screen the data with reference value from the collected historical data, the necessary degree of pre-processing of a random product image detected in the past is calculated, in the calculation process, the necessary degree is analyzed in combination with the area of invalid detection region, the number of products to be detected at that time and the number of detection points, the purpose of adding the product number parameter analysis is to eliminate the invalid detection region, which is not obvious for accelerating the visual detection speed of all products, which means that the necessity of pre-processing the image is low, the reference value of the set necessary degree reference value is improved, the product with the highest necessary degree is selected by grouping, the grouping result corresponding to the grouping mode is obtained, and the reference value is found from the grouping result, which is beneficial to further improve the reference value of the set necessary degree reference value.

[0045] Further, in steps S4-S5: the number of products to be detected of the current product is c, the image of the current product is collected, the number of detection points of the current product is H, the minimum effective detection region radius of the current product is R, and the necessary degree w of pre-processing of the image of the current product is calculated according to the formula w = (H / R) * c k ’ : if w<=M k ’ , the image of the current product is directly subjected to feature extraction, matched with the features of defective products, and the current product is subjected to defect detection; if w>M k ’ , the current product is selected as a target product, the image of the target product is pre-processed: the minimum effective detection region in the image of the target product is intercepted, the intercepted image is transmitted to an image processing system, the image processing system is used to extract features from the intercepted image, the features are matched with the features of defective products, and the target product is subjected to defect detection.

[0046] Data supervision and analysis are performed during visual detection of the current product, the necessary degree of pre-processing of the image of the current product is analyzed, and the necessary degree is compared with a reference value, whether the image of the current product needs to be pre-processed is judged, which is beneficial to pre-eliminate unnecessary detection regions when necessary, accelerate the feature extraction speed, and further improve the visual detection efficiency.

[0047] Compared with the prior art, the present application has the following advantages:

[0048] The application can effectively reduce the time complexity of finding the minimum circle covering all detection points by collecting the detection point information of different products that have completed visual detection through big data technology, dividing the minimum circle covering all detection points of the product, judging the unnecessary detection area, i.e. the invalid detection area, when detecting the corresponding product, applying the random increment method to solve the minimum circle covering problem, taking the invalid detection area as one of the judgment bases for setting the pretreatment of the current product image, which is conducive to selecting the appropriate product and pretreating the product image, only transmitting the effective detection area image, and speeding up the image feature extraction speed.

[0049] The historical data is used as the reference data for judging the pretreatment of the current product image. In order to filter out the data with reference value from the collected historical data, the necessity of pretreating a random product image that has been detected in the past is calculated. In the calculation process, the necessity is analyzed in combination with the invalid detection area, the number of products that need to be detected at that time, and the number of detection points. The number of products that need to be detected is added to the analysis to improve the reference value of the set necessity reference value. The product with the highest overall necessity is selected by grouping, the grouping result corresponding to the grouping mode is obtained, and the reference value is found therefrom, which is conducive to further improving the reference value of the set necessity reference value.

[0050] The data is supervised and analyzed when the current product is visually detected. The necessity of pretreating the current product image is analyzed and compared with the reference value to determine whether the current product image needs to be pretreated, which is conducive to pre-eliminating unnecessary detection areas when necessary, speeding up the feature extraction speed, and further improving the visual detection efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0051] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, which together with the embodiments of the application, serve to explain the application, and do not constitute a limitation on the application. In the drawings:

[0052] Fig. 1 is a structural diagram of an intelligent big data supervision system for visual detection according to the application;

[0053] Fig. 2 is a flowchart of an intelligent big data supervision method for visual detection according to the application. DETAILED DESCRIPTION

[0054] The preferred embodiments of the application are described below in conjunction with the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the application, and do not limit the application.

[0055] The preferred embodiments of the application are described below in conjunction with the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the application, and do not limit the application. Figs. 1-2The present application is further illustrated by the specific embodiments.

[0056] Embodiment one:

[0057] As shown in the figure, the embodiment provides an intelligent big data supervision system for visual inspection, which comprises a product data acquisition module, a database, a detection data analysis module, a visual inspection supervision module and a product inspection module. Fig. 1 The output end of the product data acquisition module is connected to the input end of the database, the output end of the database is connected to the input end of the detection data analysis module, the output end of the detection data analysis module is connected to the input end of the visual inspection supervision module, and the output end of the visual inspection supervision module is connected to the input end of the product inspection module.

[0058] The product data acquisition module is used for collecting historical detection data of products that have been subjected to visual inspection and transmitting all the collected data to the database.

[0059] The database is used for storing all the collected data.

[0060] The detection data analysis module is used for analyzing the historical detection data and setting a necessary degree reference value for pre-processing of product images.

[0061] The visual inspection supervision module is used for analyzing the necessary degree of pre-processing of current product images and selecting target products.

[0062] The product inspection module is used for visual inspection of current products.

[0063] The product data acquisition module comprises a detection data acquisition unit, a detection point position acquisition unit and a product quantity acquisition unit.

[0064] The output ends of the detection data acquisition unit, the detection point position acquisition unit and the product quantity acquisition unit are connected to the input end of the database.

[0065] The detection data acquisition unit is used for collecting detection point quantity data of products that have been subjected to visual inspection in the past and image data of different types of products that have been photographed. Due to different detection contents, there can be multiple detection points when visual inspection is performed on products, for example, when visual inspection is performed on a PCB bare board, it is necessary to detect whether multiple element positions are correct, whether a circuit is open or closed, whether an element shape is incorrect, etc.

[0066] The detection point position acquisition unit is used for establishing a two-dimensional coordinate system with the center of the photographed image as the origin and collecting position coordinate information of all detection points.

[0067] The detection point position acquisition unit is used for establishing a two-dimensional coordinate system with the center of the photographed image as the origin and collecting position coordinate information of all detection points.

[0068] The product quantity acquisition unit is configured to acquire quantity information of different types of products that have been detected in the past, and the number and positions of detection points for the same type of product are the same.

[0069] The detection data analysis module includes a historical data analysis unit and a reference data setting unit.

[0070] The output end of the database is connected to the input end of the historical data analysis unit, and the output end of the historical data analysis unit is connected to the input end of the reference data setting unit.

[0071] The historical data analysis unit is configured to retrieve product image data and detection point position coordinate information that have been used for visual detection in the past, to demarcate minimum effective detection regions for different types of products, to cover all detection points of the corresponding products with the minimum effective detection regions, and to estimate areas of invalid detection regions for different types of products.

[0072] The reference data setting unit is configured to set a necessary degree reference value for image preprocessing after product image shooting according to the areas of invalid detection regions and the necessary degree of image preprocessing after product image shooting according to historical detection data analysis.

[0073] The visual detection supervision module includes a product image shooting unit, a processing necessity analysis unit, and a target product selection unit.

[0074] The output end of the product image shooting unit is connected to the input end of the processing necessity analysis unit, and the output ends of the processing necessity analysis unit and the reference data setting unit are connected to the input end of the target product selection unit.

[0075] The product image shooting unit is configured to shoot images of the current product.

[0076] The processing necessity analysis unit is configured to analyze the necessary degree of preprocessing of the images of the current product.

[0077] The target product selection unit is configured to compare the necessary degree of preprocessing of the images of the current product with the necessary degree reference value, and if the necessary degree exceeds the reference value, the current product is selected as the target product; otherwise, the images of the current product are directly subjected to feature extraction, matched with the features of defective products, and the current product is subjected to defect detection.

[0078] The product detection module includes an image preprocessing unit and a visual detection unit.

[0079] The output end of the target product selection unit is connected to the input end of the image preprocessing unit, and the output end of the image preprocessing unit is connected to the input end of the visual detection unit.

[0080] The image preprocessing unit is configured to crop the minimum effective detection region in the images of the target product.

[0081] The visual inspection unit is used to extract features from the captured image using an image processing system, match them with the features of the defective product, and perform defect detection on the target product.

[0082] Example 2:

[0083] like Fig. 2 As shown, this embodiment provides an intelligent big data supervision method for visual inspection, which is implemented based on the data supervision system in this embodiment, and specifically includes the following steps:

[0084] S1: Collect historical inspection data of products that have undergone visual inspection. The set of inspection points for different types of products that have undergone visual inspection in the past is E = {E1, E2, ..., E...} m The number of inspection points is greater than 1. Data for m different types of products were collected. The set of the total number of each type of product that needed to be inspected in previous inspections of the corresponding product types is B = {B1, B2, ..., B...}. m} Collect product images captured during visual inspection of m types of products, establish a two-dimensional coordinate system with the center of the captured image as the origin, and collect the position coordinate information of all inspection points;

[0085] For example: Data for 7 different types of products were collected. The set of inspection points for the 7 products that had been visually inspected in the past was E = {E1, E2, E3, E4, E5, E6, E7} = {2, 5, 4, 3, 7, 8, 5}. The set of the total number of each product that needed to be inspected when inspecting the 7 products in the past was B = {B1, B2, B3, B4, B5, B6, B7} = {100, 200, 50, 70, 36, 90, 120}. Product images were collected when visually inspecting the 7 products.

[0086] S2: Estimate the area of ​​invalid detection regions for different types of products. Retrieve product images and the coordinates of detection points taken during visual inspection of 7 products. Obtain the image area as X = 1200 (square millimeters). Using the random increment method, obtain the minimum effective detection region covering all detection points of a random product as: a circle centered at (a, b) = (10, 10), with r... i A circular area with a radius of 10 is used to obtain the set of minimum effective detection area radii for the 7 products as r = {r1, r2, r3, r4, r5, r6, r7} = {10, 15, 18, 9, 12, 7, 19}, in square millimeters. According to formula Z... i =X-π*r i 2 Estimate the area Z of invalid detection region for a random product. i Where X > π*ri 2 The invalid detection area set of m products is Z = {Z1, Z2, Z3, Z4, Z5, Z6, Z7} = {886, 493, 182, 946, 748, 1046, 66} by the same calculation method.

[0087] S3: Set the necessary degree reference value of pre-processing the product image according to the invalid detection area and historical detection data, and calculate the necessary degree W of pre-processing a random product image according to the formula i , wherein E i represents the number of detection points of a random product, B i represents the total number of corresponding types of products that need to be detected when detecting a random product in the past, and the necessary degree set of m product images is W = {W1, W2, W3, W4, W5, W6, W7} = {7.8, 21.8, 1.6, 8.8, 8.3, 33.3, 1.7}. Arrange m products in order from small to large necessary degree, and arrange the arranged products into k = 3 groups. The necessary degree of pre-processing the image of the previous group product is less than that of the next group. After grouping according to a random one, the mean value set of the necessary degree corresponding to the k groups of products is M = {M1, M2, M3} = {2.45, 8.3, 27.55}. According to the formula , calculate the degree of excellence F of the random one grouping method j = 10.7, wherein M e represents the mean value of the necessary degree corresponding to the e-th group of products after grouping according to a random one, e = 1, 2, …, k, j = 1, 2, …, g, and there are g grouping methods. The degree of excellence set of g grouping methods is F = {F1, F2, …, F j , …, F g} by the same calculation method. The grouping result of the grouping method with the highest degree of excellence is {1.6, 1.7}, {7.8, 8.3, 8.8}, {21.8, 33.3}. The mean value set of the necessary degree corresponding to each group of products in the corresponding grouping result is M ’ = {M1 ’ , M2 ’ , M3 ’} = {2.45, 8.3, 27.55}. Set the necessary degree reference value of pre-processing the product image as M3 ’ = 27.55.

[0088] S4: Analyze the necessary degree of pre-processing the current product image and select the target product.

[0089] ​S5: visual inspection is carried out on the current product, the number of current products to be detected is obtained as c, the image of the current product is collected, the number of detection points of the current product is obtained as H, the minimum effective detection area radius of the current product is obtained as R, and the necessary degree w of pre-processing of the current product image is calculated according to the formula w and M are compared k ’ : if w≤M k ’ , the image of the current product is directly subjected to feature extraction, matched with the features of the defective product, and the current product is subjected to defect detection; if w>M k ’ , the current product is selected as the target product, the image of the target product is pre-processed: the minimum effective detection area in the image of the target product is intercepted, the intercepted image is transmitted to the image processing system, the image processing system is used to extract features from the intercepted image, matched with the features of the defective product, and the target product is subjected to defect detection.

[0090] For example: the number of current products to be detected is obtained as 178, the image of the current product is collected, the number of detection points of the current product is obtained as 4, the minimum effective detection area radius of the current product is obtained as 5 by using the random increment method, and the necessary degree w of pre-processing of the current product image is obtained as w≈35.26>27.55, w>M3 ’ , the current product is selected as the target product, the image of the target product is pre-processed: the minimum effective detection area in the image of the target product is intercepted, the intercepted image is transmitted to the image processing system, the image processing system is used to extract features from the intercepted image, matched with the features of the defective product, and the target product is subjected to defect detection.

[0091] Finally, it should be noted that the above is only a preferred example of the present application and is not intended to limit the present application, although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent replacements of some technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. An intelligent big data monitoring method for visual inspection, characterized in that: The method comprises the following steps: S1: collecting historical detection data of products that have been visually detected; S2: estimating the area of invalid detection regions of different types of products; S3: setting a necessary degree reference value for pre-processing of product images according to the area of invalid detection regions and the historical detection data; S4: analyzing the necessary degree for pre-processing of a current product image and selecting a target product; S5: visually detecting the current product; In step S1: a set of detection point quantities of different types of products detected in the past is collected as E={E1, E2, …, E m}, and the detection point quantities are all greater than 1, wherein data of m different types of products are collected, a set of total quantities of each type of product that needs to be detected when the corresponding type of product is detected in the past is collected as B={B1, B2, …, B m}, product images are collected when the m products are visually detected, a two-dimensional coordinate system is established with the center of the captured image as the origin, and the position coordinate information of all detection points is collected; In step S2: retrieve the product images and the coordinate information of the detection points captured during visual inspection of m types of products, obtain the area of ​​the captured image as X, and use the random increment method to obtain the minimum effective detection area covering all detection points of a random product as: with (a, b) as the center and r i Given a circular region with radius r, the minimum effective detection region radius set for m products is obtained as r = {r1, r2, ..., r...} i , ..., r m According to formula Z i =X-π*r i 2 Estimate the area Z of invalid detection region for a random product. i Where X > π*r i 2 The set of invalid detection area areas for m products, obtained through the same calculation method, is Z = {Z1, Z2, ..., Z...} i , ..., Z m }; In step S3: the necessary degree of pre-processing W for a random product image is calculated according to the following formula i : wherein, E i represents the number of detection points of a random product, B i represents the total number of corresponding types of products that need to be detected when detecting a random product in the past, and the necessary degree set for pre-processing m product images is W = {W1, W2, …, W i , …, W m}, the m products are arranged in order of the necessary degree from small to large, the arranged products are divided into k groups, the necessary degree of pre-processing the product images of the previous group is less than that of the next group, the average necessary degree set of the k groups of products after grouping in a random manner is M = {M1, M2, …, M e , …, M k}, the degree of optimization F of the random grouping manner is calculated according to the formula j , wherein M e represents the average necessary degree of the e-th group of products after grouping in a random manner, e = 1, 2, …, k, j = 1, 2, …, g, and there are g grouping manners, the degree of optimization set of the g grouping manners is F = {F1, F2, …, F j , …, F g} obtained by the same calculation method, and the grouping result of the grouping manner with the largest degree of optimization is obtained: the average necessary degree set of each group of products corresponding to the grouping result is M ’ = {M1 ’ , M2 ’ , …, M k ’}, and the necessary degree reference value for pre-processing the product images is set as M k ’ ; In steps S4-S5: the number of current products to be detected is obtained as c, the image of the current product is collected, the number of detection points of the current product is obtained as H, the minimum effective detection area radius of the current product is obtained as R, and the necessary degree w of pre-processing of the current product image is calculated according to the formula w = H / R The necessary degree w of pre-processing of the current product image is compared with M k ’ If w≤M k ’ , the image of the current product is directly subjected to feature extraction, matched with the features of the defective products, and the current product is subjected to defect detection; if w>M k ’ , the current product is selected as a target product, the image of the target product is pre-processed, the minimum effective detection area in the image of the target product is intercepted, the intercepted image is transmitted to an image processing system, the image processing system is used to extract features from the intercepted image, the features are matched with the features of the defective products, and the target product is subjected to defect detection.

2. An intelligent big data monitoring system for visual inspection, applied to the intelligent big data monitoring method for visual inspection according to claim 1, characterized in that: The system comprises a product data collection module, a database, a detection data analysis module, a visual detection supervision module, and a product detection module; The output end of the product data collection module is connected to the input end of the database, the output end of the database is connected to the input end of the detection data analysis module, the output end of the detection data analysis module is connected to the input end of the visual detection supervision module, and the output end of the visual detection supervision module is connected to the input end of the product detection module; The product data collection module is used to collect historical detection data of products that have been visually detected and transmit all the collected data to the database; The database is used to store all the collected data; The detection data analysis module is used to analyze the historical detection data and set a necessary degree reference value for pre-processing of product images; The visual detection supervision module is used to analyze the necessary degree for pre-processing of a current product image and select a target product; The product detection module is used to visually detect the current product.

3. The intelligent big data monitoring system for visual inspection of claim 2, wherein: The product data collection module comprises a detection data collection unit, a detection point position collection unit, and a product quantity collection unit; The output ends of the detection data collection unit, the detection point position collection unit, and the product quantity collection unit are connected to the input end of the database; The detection data collection unit is used to collect the number of detection points of products that have been visually detected in the past and image data of different types of products that have been photographed; The detection point position collection unit is used to establish a two-dimensional coordinate system with the center of the photographed image as the origin and collect position coordinate information of all detection points; The product quantity collection unit is used to collect quantity information of different types of products that have been detected in the past, the number of detection points of the same type of product, and the position of the detection points.

4. The intelligent big data monitoring system for visual inspection of claim 2, wherein: The detection data analysis module comprises a historical data analysis unit and a reference data setting unit; The input end of the historical data analysis unit is connected to the output end of the database, and the output end of the historical data analysis unit is connected to the input end of the reference data setting unit; The historical data analysis unit is used to retrieve product image data and detection point position coordinate information that have been photographed for visual detection in the past, delimit the minimum effective detection region of different types of products, cover all detection points of the corresponding product with the minimum effective detection region, and estimate the area of invalid detection regions of different types of products; The reference data setting unit is used to set a necessary degree reference value for pre-processing of images after photographing the product images according to the necessary degree for pre-processing of images after photographing the product images according to the area of invalid detection regions and the historical detection data analysis.

5. The intelligent big data monitoring system for visual inspection as claimed in claim 4 wherein: The visual detection supervision module comprises a product image shooting unit, a processing necessity analysis unit and a target product selection unit; An output end of the product image shooting unit is connected to an input end of the processing necessity analysis unit, and output ends of the processing necessity analysis unit and the reference data setting unit are connected to an input end of the target product selection unit; The product image shooting unit is configured to shoot an image of a current product; The processing necessity analysis unit is configured to analyze a necessity degree of pre-processing of the image of the current product; The target product selection unit is configured to compare the necessity degree of pre-processing of the image of the current product with a reference value of the necessity degree, and if the necessity degree exceeds the reference value, select the current product as a target product; otherwise, directly extract features from the image of the current product, match the features with features of a defective product, and perform defect detection on the current product.

6. The intelligent big data monitoring system for visual inspection of claim 5, wherein: The product detection module comprises an image pre-processing unit and a visual detection unit; An input end of the image pre-processing unit is connected to an output end of the target product selection unit, and an output end of the image pre-processing unit is connected to an input end of the visual detection unit; The image pre-processing unit is configured to intercept a minimum effective detection area in the image of the target product; The visual detection unit is configured to extract features from the intercepted image by using an image processing system, match the features with features of a defective product, and perform defect detection on the target product.

Images