An AI-based PCBA fault detection method and system

Abstract

The application relates to the technical field of fault detection, and discloses an AI-based PCBA fault detection method and system, which comprises a PCBA welding information processing module, a PCBA welding fault detection module and a PCBA welding fault detection result feedback module. The image of each welding spot in the PCBA is intelligently collected by marking the characteristic image of the welding spot in the standard PCBA welding spot marking image according to the standard PCBA welding spot marking image information and combining the image cutout convenient software, the welding spot in the PCBA is intelligently collected, the welding defect of the welding spot in the PCBA is intelligently detected according to the PCBA marking welding spot characteristic image information, the AI intelligent recognition algorithm and the different type PCBA welding spot defect standard image information based on big data storage, the point-to-point fine detection of the PCBA welding defect based on AI is realized, and the precision and quality of the PCBA welding fault detection are improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of fault detection, and particularly relates to an AI-based PCBA fault detection method and system. BACKGROUND

[0002] PCBA is also called printed circuit board or printed wiring board, and PCBA refers to a finished circuit board on which electronic components are welded and assembled; the core function is to integrate discrete components into PCBA through surface mount technology or plug-in technology process to form an electronic system with specific functions. The core process flow of PCBA substrate manufacturing is to form a conductive circuit on an insulating substrate through chemical etching to provide mechanical support and electrical connection basis for components; SMT process: a high-speed chip mounter accurately mounts micro components to the PCBA pad; DIP process: plug-in machine or manual installation of through-hole components. Welding and detection: electrical connection is realized by reflow soldering or wave soldering, and multiple detection means such as solder paste detection, automatic optical detection and X-RAY are used to ensure the welding quality; however, the existing PCBA welding fault detection cannot realize intelligent identification of welding point welding faults, nor can it realize visual marking output of welding point welding faults, which reduces the precision and intuitiveness of PCBA welding fault detection.

[0003] A Chinese invention patent with publication number CN114416450B discloses a PCBA production test management method, which first generates a logical addressing code stored in a database and writes it into a production test wireless tracking unit to establish a tracking matching relationship of the corresponding PCBA circuit board; reads the production test wireless tracking unit data to scan the PCBA circuit board through a microscope, determines and identifies the corresponding production test hardware model, and calls the corresponding production test software module; records the increase and decrease of hardware components and the software running state of each process, and performs PCBA production test process display and running monitoring and tracking management; however, the technical solution cannot realize intelligent quality inspection of the welding quality of the PCBA circuit board. SUMMARY

[0004] (I) Technical problems to be solved

[0005] To solve the above problems that the existing PCBA welding fault detection cannot realize intelligent identification of welding point welding faults, nor can it realize visual marking output of welding point welding faults, which reduces the precision and intuitiveness of PCBA welding fault detection, the purposes of scientifically constructing a PCBA welding point marking image, accurately collecting a PCBA marking welding point feature image, intelligently judging a PCBA marking welding point welding defect, and visually feeding back a PCBA welding point welding fault detection result are achieved.

[0006] (II) Technical solutions

[0007] The application is implemented by the following technical solutions: An AI-based PCBA fault detection method, comprising the following steps:

[0008] S1, collect PCBA feature image data;

[0009] S2, perform image preprocessing based on the PCBA feature image data to generate standard PCBA feature image data;

[0010] S3, perform standard PCBA feature image welding point identification and welding point marking processing according to the standard PCBA feature image data and PCBA welding point standard image data, and generate standard PCBA welding point marking image data;

[0011] S4, perform feature image collection processing of the marked welding points in the standard PCBA welding point marking image data based on the standard PCBA welding point marking image data, and construct PCBA marked welding point feature image data;

[0012] S5, perform PCBA marked welding point welding defect detection processing according to the PCBA marked welding point feature image data and different types of PCBA welding point defect standard image data, and generate PCBA marked welding point welding defect detection data;

[0013] S6, perform PCBA marked welding point welding defect detection result text information display rendering processing based on the PCBA marked welding point welding defect detection data, generate PCBA marked welding point welding fault detection result rendering text data, and perform graphic-text combination processing of the PCBA marked welding point welding defect detection result text information and the standard PCBA welding point marking image based on the standard PCBA welding point marking image data, and construct PCBA marked welding point welding fault detection result display image data;

[0014] S7, execute PCBA welding fault detection result pushing output job according to the PCBA marked welding point welding fault detection result display image data.

[0015] Preferably, the operation steps of collecting PCBA feature image data are as follows:

[0016] S11, complete the online collection of the overall structure feature image information of the front or back of the PCBA after tin soldering welding operation through a cloud lens, and generate PCBA feature image data .

[0017] Preferably, the operation steps of performing image preprocessing based on the PCBA feature image data to generate standard PCBA feature image data are as follows:

[0018] S21. Obtain the PCBA feature image data. ;

[0019] S22. Apply a Gaussian filter to the PCBA feature image data. Perform image denoising preprocessing on PCBAs to generate standard PCBA feature image data. .

[0020] Preferably, the steps for identifying solder joints in the standard PCBA feature image based on the standard PCBA feature image data and the standard PCBA solder joint image data, and for processing solder joint markings to generate standard PCBA solder joint marking image data are as follows:

[0021] S31. Establish a standard image data set for PCBA solder joints. , ;in Indicates the first Standard image data of PCBA solder joints This represents the maximum number of standard images of PCBA solder joints; the PCBA solder joint standard image data represents the optimal image information set for solder joints in the PCBA.

[0022] S32. The standard PCBA feature image data is processed using the SIFT image feature search algorithm. Standard image data set of PCBA solder joints Standard image data of PCBA solder joints described in the document Perform image feature matching to search and identify the standard PCBA feature image data. All PCBA solder joints are analyzed, and image annotation and editing software is used simultaneously to process the standard PCBA feature image data. The PCBA solder joints retrieved from the search are digitally marked to construct standard PCBA solder joint marking image data. The image labeling and editing software includes any one of the following: XunJie Image Converter, Photoshop, and LabelImg.

[0023] Preferably, the steps for acquiring and processing the feature images of marked solder joints in the standard PCBA solder joint marking images based on the standard PCBA solder joint marking image data to construct PCBA marked solder joint feature image data are as follows:

[0024] S41. Obtain the standard PCBA solder joint marking image data. ;

[0025] S42, adopting an image matting convenient software to mark the image data of the standard PCBA welding spot The welding spot feature image in the corresponding standard PCBA welding spot marking image which is subjected to the digital marking processing in step S32 is sequentially subjected to matting collection processing according to the digital marking sequence, and a PCBA marking welding spot feature image data set is constructed , ; wherein denotes the PCBA marking welding spot feature image data corresponding to the i-th PCBA marking welding spot, denotes the maximum value of the number of PCBA marking welding spots; the PCBA marking welding spot feature image data denotes the feature image information of a single marking welding spot collected through image matting processing, and the image matting convenient software includes any one of Remove.bg, Fococlipping and Cool AI intelligent matting.

[0026] Preferably, the operation steps of welding defect detection processing of the PCBA marking welding spot according to the PCBA marking welding spot feature image data and different type PCBA welding spot defect standard image data to generate PCBA marking welding spot welding defect detection data are as follows:

[0027] S51, establishing a different type PCBA welding spot defect standard image data set , ; wherein denotes the different type PCBA welding spot defect standard image data corresponding to the j-th PCBA welding spot defect type, denotes the maximum value of the number of PCBA welding spot defect types; the PCBA welding spot defect type includes virtual welding, cold welding, continuous welding, welding cracking, solder accumulation and welding bridging; the different type PCBA welding spot defect standard image data represents the welding spot welding defect image information set based on different types of PCBA welding spot welding defect standards;

[0028] S52, the PCBA marking welding spot feature image data in the PCBA marking welding spot feature image data set is sequentially matched with the different type PCBA welding spot defect standard image data in the different type PCBA welding spot defect standard image data set according to the PCBA marking welding spot number . ​​​Perform image feature matching, and generate PCBA marked solder joint welding defect detection data set according to image feature matching result ; and perform specific operation steps of generating the PCBA marked solder joint welding defect detection data set as follows:

[0029] S521, initialization, the solder joint defect identification cheetah updates the position in the search space of the different type PCBA solder joint defect standard image data set with the space dimension of The initialization position update formula of the solder joint defect identification cheetah is as follows: , wherein represents the position of the solder joint defect identification cheetah individual i in the search space of the different type PCBA solder joint defect standard image data set with the space dimension of , and , , , respectively represent the upper limit value and the lower limit value of the search space of the different type PCBA solder joint defect standard image data set with the space dimension of , rand is a random number between 0 and 1, and T represents the maximum number of iterations;

[0030] S522, complete initialization and perform search strategy search prey, the solder joint defect identification cheetah performs full range scanning or active search in its territory or surrounding area to find prey; the solder joint defect identification cheetah searches out the different type PCBA solder joint defect standard image data matching the PCBA marked solder joint feature image data in the search space of the different type PCBA solder joint defect standard image data set ; the mathematical description of the search strategy is as follows: , wherein represents the position of the solder joint defect identification cheetah individual i after the (t+1) th iteration in the search space of the different type PCBA solder joint defect standard image data set with the space dimension of , represents the position of the solder joint defect identification cheetah individual i after the t th iteration in the search space of the different type PCBA solder joint defect standard image data set with the space dimension of , represents the position of the solder joint defect identification cheetah individual i after the t th iteration in the search space of the different type PCBA solder joint defect standard image data set with the space dimension of , represents the position of the solder joint defect identification cheetah individual i after the t th iteration in the search space of the different type PCBA solder joint defect standard image data set with the space dimension of , ​a random number normally distributed in the search space of the different types of PCBA solder joint defect standard image data set represents the search space of the different types of PCBA solder joint defect standard image data set after the tth iteration of the individual of the solder joint defect recognition cheetah in the search space of the different types of PCBA solder joint defect standard image data set ;

[0031] S523, search prey using the wait-and-see strategy, in the search mode, the prey is exposed to the field of view of the solder joint defect recognition cheetah, and the different types of PCBA solder joint defect standard image data set in the search space is searched to find the different types of PCBA solder joint defect standard image data matching the PCBA marked solder joint feature image data , each action of the solder joint defect recognition cheetah will cause the prey to escape, the solder joint defect recognition cheetah adopts the wait-and-ambush strategy to approach the prey, the wait-and-ambush strategy includes lying on the ground or hiding in the bushes, and the strategy is mathematically described as ;

[0032] S524, execute the wait-and-see strategy to approach the prey, attack the prey according to the attack strategy, in the algorithm, each solder joint defect recognition cheetah adjusts its position according to the position of the escaping prey, the leading solder joint defect recognition cheetah or the nearby solder joint defect recognition cheetah to obtain the best attack position, and the different types of PCBA solder joint defect standard image data set in the search space is searched to find the different types of PCBA solder joint defect standard image data most matching the PCBA marked solder joint feature image data , the attack strategy is mathematically described as , wherein represents the best position in the search space of the different types of PCBA solder joint defect standard image data set after the tth iteration of the individual of the solder joint defect recognition cheetah in the search space of the different types of PCBA solder joint defect standard image data set ; is the turning factor of the individual i of the solder joint defect recognition cheetah in the search space of the different types of PCBA solder joint defect standard image data set ; represents the interaction factor of the individual i of the solder joint defect recognition cheetah in the search space of the different types of PCBA solder joint defect standard image data set ;

[0033] ​​S525, repeat S522, S523, S524 until the maximum number of iterations is met, output the PCBA mark welding spot feature image data with the different types of PCBA welding spot defect standard image data image feature matching results;

[0034] S526, according to the CBA mark welding spot feature image data with the different types of PCBA welding spot defect standard image data Image feature matching results generate PCBA mark welding spot welding defect detection data set , wherein The PCBA mark welding spot welding defect detection data corresponding to the PCBA mark welding spot is represented by the PCBA mark welding spot welding defect detection data corresponding to the PCBA mark welding spot.

[0035] When and Image feature matching is successful, indicating that the PCBA mark welding spot exists in the PCBA mark welding spot, and the PCBA mark welding spot welding defect detection data is output Defects exist, and the PCBA mark welding spot defect type text information is output When and

[0036] Image feature matching is not successful, indicating that the PCBA mark welding spot does not exist welding spot defect, and the PCBA mark welding spot welding defect detection data is output Defects do not exist.

[0037] Preferably, based on the PCBA mark welding spot welding defect detection data, the welding defect detection result text information display rendering process of the PCBA mark welding spot is carried out, the PCBA mark welding spot welding fault detection result rendering text data is generated, and the PCBA mark welding spot welding defect detection result text information and the standard PCBA mark welding spot image are combined with the image of the PCBA mark welding spot welding defect detection result text information and the standard PCBA mark welding spot image. The operation steps of constructing the PCBA mark welding spot welding fault detection result display image data are as follows:

[0038] S61, using text editing software to respectively edit the PCBA mark welding spot welding defect detection data set The PCBA mark welding spot welding defect detection data in which the PCBA mark welding spot welding defect detection result is defect exists​​​​​ The rendering process uses a red background for highlighting, and the PCBA marking solder joint welding defect detection data shows that no defects are found. The rendering process uses a green background with highlighted highlights, and generates a set of rendered text data showing the detection results of solder joint defects on PCBA. ,in Indicates the first The rendered text data of the welding fault detection results of the PCBA marked solder joints is represented by the color-coded text information of the different welding fault detection results for the PCBA marked solder joints; the text editing software includes any one of Vim, EditPlus, and Word.

[0039] S62. Render the text data set of the CBA-marked welding weld point welding fault detection results. The rendered text data of the PCBA marking solder joint soldering fault detection results described in the article Image data of solder joint markings on the standard PCBA In the process of identifying and matching the text information of the PCBA solder joints marked with their numerical labels, the images of the marked PCBA solder joints are mapped and matched to create an image data display of the solder joint failure detection results. The image data for displaying the welding fault detection results of PCBA marked solder joints represents the text information of the welding fault detection results of PCBA marked solder joints that has been rendered with different background colors and highlighted, and then mapped onto the standard PCBA solder joint marked image to form the image information for displaying the welding fault detection results of PCBA solder joints.

[0040] Preferably, the operation steps for pushing and outputting PCBA welding fault detection results based on the displayed image data of the PCBA marked solder joint welding fault detection results are as follows:

[0041] S71. Display the image data of the PCBA marked solder joint welding fault detection results. The PCBA welding fault detection results are transmitted to the PCBA quality inspection platform via the Internet of Things (IoT) communication network and then pushed out.

[0042] An AI-based PCBA fault detection system is provided to implement the AI-based PCBA fault detection method. The system includes a PCBA soldering information processing module, a PCBA soldering fault detection module, and a PCBA soldering fault detection result feedback module.

[0043] The PCBA welding information processing module includes a PCBA feature image acquisition unit, a PCBA feature image preprocessing unit, a PCBA welding welding point standard image storage unit, and a PCBA welding welding point marked image generation unit.

[0044] The PCBA feature image acquisition unit acquires PCBA feature image data through a cloud lens. The PCBA feature image preprocessing unit performs image preprocessing based on the PCBA feature image data to generate standard PCBA feature image data. The PCBA welding welding point standard image storage unit is used to store PCBA welding welding point standard image data. The PCBA welding welding point marked image generation unit identifies welding points in the standard PCBA feature image according to the standard PCBA feature image data, PCBA welding welding point standard image data, and image marking editing software, and performs welding point marking processing to generate standard PCBA welding welding point marked image data.

[0045] The PCBA welding fault detection module includes a PCBA marked welding point image acquisition unit, a different type PCBA welding point defect standard image storage unit, and a PCBA marked welding point welding defect detection unit.

[0046] The PCBA marked welding point image acquisition unit acquires feature images of marked welding points in the standard PCBA welding welding point marked image according to the standard PCBA welding welding point marked image data and image matting software, and constructs PCBA marked welding point feature image data. The different type PCBA welding point defect standard image storage unit is used to store different type PCBA welding point defect standard image data. The PCBA marked welding point welding defect detection unit detects welding defects of the PCBA marked welding point according to the PCBA marked welding point feature image data and different type PCBA welding point defect standard image data, and generates PCBA marked welding point welding defect detection data.

[0047] The PCBA welding fault detection result feedback module includes a PCBA marked welding point welding fault detection result rendering text construction unit, a PCBA marked welding point welding fault detection result image generation unit, and a PCBA welding fault detection result pushing unit.

[0048] The PCBA marked welding point welding fault detection result rendering text construction unit performs PCBA marked welding point welding defect detection result text information display rendering processing based on the PCBA marked welding point welding defect detection data in combination with a text editing software, and generates PCBA marked welding point welding fault detection result rendering text data; the PCBA marked welding point welding fault detection result image generation unit performs graphic-text combination processing of PCBA marked welding point welding defect detection result text information and standard PCBA welding point mark image according to the PCBA marked welding point welding fault detection result rendering text data and the standard PCBA welding point mark image data, and constructs PCBA marked welding point welding fault detection result display image data; and the PCBA welding fault detection result pushing unit performs PCBA welding fault detection result pushing output work according to the PCBA marked welding point welding fault detection result display image data in combination with a PCBA quality inspection platform.

[0049] (Three) beneficial effects

[0050] The application provides an AI-based PCBA fault detection method and system. The following beneficial effects are achieved:

[0051] I. The through-cloud lens efficiently acquires PCBA feature images and performs image noise reduction preprocessing in combination with an image noise reduction algorithm, realizes accurate acquisition of PCBA feature image information, and provides reliable data support for PCBA fault detection; according to standard PCBA feature image information, an intelligent search algorithm, an image mark editing software, and scientifically preset PCBA welding point standard image information are combined to perform intelligent identification and marking of welding points in the standard PCBA feature image, realize accurate and efficient capture of PCBA welding point targets, and improve the accuracy and efficiency of PCBA welding fault detection.

[0052] II. According to the standard PCBA welding point mark image information, the feature images of the marked welding points in the standard PCBA welding point mark image are accurately collected in combination with an image matting convenient software, the image of each welding point in the PCBA is intelligently collected, according to the PCBA marked welding point feature image information, an AI intelligent recognition algorithm, and different types of PCBA welding point defect standard image information based on big data storage, the welding defects of the welding points in the PCBA are intelligently detected, the point-to-point fine detection of the AI-based PCBA welding defects is realized, and the precision and quality of the PCBA welding fault detection are improved.

[0053] III. Different welding defect detection result differential output display of the PCBA welding solder joint is realized through text highlight rendering processing; based on data processing, the text rendering result of the welding defect detection result of the PCBA welding solder joint and the PCBA welding solder joint marking image are matched and combined according to the welding solder joint, the welding fault detection result of the PCBA welding solder joint is visually displayed based on the image, and the readability of the PCBA welding fault detection result is improved; the PCBA welding fault detection result pushing unit combines the PCBA marking welding solder joint welding fault detection result display image data with the PCBA quality inspection platform to timely execute the PCBA welding fault detection result pushing output task, realizes efficient feedback of the PCBA welding fault detection result, and improves the applicability of the PCBA welding fault detection. BRIEF DESCRIPTION OF DRAWINGS

[0054] Fig. 1 A module schematic diagram of an AI-based PCBA fault detection system provided by the present application is shown in the figure.

[0055] Fig. 2 A flowchart of an AI-based PCBA fault detection method provided by the present application is shown in the figure. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0057] The embodiments of the AI-based PCBA fault detection method and system are as follows:

[0058] Embodiment 1

[0059] Please refer to Figs. 1-2 An AI-based PCBA fault detection method, the method comprising the following steps:

[0060] S1, collecting PCBA feature image data;

[0061] S2, performing image preprocessing based on the PCBA feature image data to generate standard PCBA feature image data;

[0062] S3, performing welding solder joint recognition in the standard PCBA feature image and welding solder joint marking processing according to the standard PCBA feature image data and the PCBA welding solder joint standard image data to generate standard PCBA welding solder joint marking image data;

[0063] S4, mark the feature image of the soldering joint in the standard PCBA soldering joint mark image data according to the standard PCBA soldering joint mark image data, and construct the PCBA mark soldering joint feature image data;

[0064] S5, perform soldering defect detection processing of the PCBA mark soldering joint according to the PCBA mark soldering joint feature image data and the different types of PCBA soldering joint defect standard image data, and generate PCBA mark soldering joint soldering defect detection data;

[0065] S6, based on the PCBA mark soldering joint soldering defect detection data, perform PCBA mark soldering joint soldering defect detection result text information display rendering processing, generate PCBA mark soldering joint soldering fault detection result rendering text data, and perform PCBA mark soldering joint soldering defect detection result text information and standard PCBA soldering joint mark image graphic text combination processing with the standard PCBA soldering joint mark image data, and construct the PCBA mark soldering joint soldering fault detection result display image data;

[0066] S7, execute PCBA soldering fault detection result pushing output job according to the PCBA mark soldering joint soldering fault detection result display image data.

[0067] Further, please refer to Figs. 1-2 , the operation steps of collecting PCBA feature image data are as follows:

[0068] S11, collect the overall structure feature image information of the front or back of the PCBA after completing the soldering joint operation through the cloud lens online, and generate PCBA feature image data .

[0069] The operation steps of performing image preprocessing based on the PCBA feature image data to generate standard PCBA feature image data are as follows:

[0070] S21, acquire PCBA feature image data ;

[0071] S22, perform PCBA image noise reduction preprocessing on the PCBA feature image data using a Gaussian filter to generate standard PCBA feature image data .

[0072] The operation steps of performing standard PCBA feature image soldering joint recognition and soldering joint marking processing according to the standard PCBA feature image data and the PCBA soldering joint standard image data to generate standard PCBA soldering joint mark image data are as follows:

[0073] S31, establish a PCBA welding joint standard image data set , ; wherein represents the PCBA welding joint standard image data, represents the maximum value of the number of PCBA welding joint standard images; PCBA welding joint standard image data represents the optimal image information set for the welding joint in the PCBA;

[0074] S32, using the SIFT image feature search algorithm to perform image feature matching between the standard PCBA feature image data and the PCBA welding joint standard image data set PCBA welding joint standard image data in the PCBA welding joint standard image data set , search and identify all PCBA welding joints in the standard PCBA feature image data , and simultaneously perform digital marking processing on the searched PCBA welding joints in the standard PCBA feature image data using image marking editing software to construct standard PCBA welding joint marking image data , the image marking editing software includes any one of‌Quick Picture Converter,‌Photoshop, and‌LabelImg.

[0075] Through the cooperation of the PCBA feature image acquisition unit and the PCBA feature image preprocessing unit, the cloud lens is used to efficiently acquire the PCBA feature image, and the image noise reduction algorithm is used for image noise reduction preprocessing, realizing accurate acquisition of PCBA feature image information and providing reliable data support for PCBA fault detection; the PCBA welding joint marking image generation unit, according to the standard PCBA feature image information, combines the intelligent search algorithm and the image marking editing software with the scientifically preset PCBA welding joint standard image information to intelligently identify and mark the welding joints in the standard PCBA feature image, realizing accurate and efficient capture of the PCBA welding joint target, and improving the accuracy and efficiency of the PCBA welding fault detection.

[0076] Figs. 1-2 Further, according to the standard PCBA welding joint marking image data, the operation steps of constructing the PCBA marking welding joint feature image data are as follows:

[0077] S41, acquire standard PCBA welding joint marking image data ;

[0078] S42, use image cutout software to process the standard PCBA welding joint marking image data The solder joint feature images of the corresponding standard PCBA solder joint marking images, which underwent digital marking processing in step S32, are then systematically extracted and processed according to the order of the digital markings to construct a PCBA marked solder joint feature image data set. , ;in Indicates the first PCBA marked solder joint feature image data corresponding to each PCBA marked solder joint. This indicates the maximum number of marked solder joints on the PCBA; the PCBA marked solder joint feature image data represents the data from standard PCBA solder joint marked image data. The image matting process captures the feature image information of a single marked welding weld point. Convenient image matting software includes Remove.bg, Fococlipping, and any one of CoolSparrow AI Intelligent Matting.

[0079] The steps for generating PCBA marked solder joint defect detection data by processing the characteristic image data of PCBA solder joints and standard image data of different types of PCBA solder joint defects are as follows:

[0080] S51. Establish a standard image data set of different types of PCBA solder joint defects. , ;in Indicates the first Standard image data of different types of PCBA solder joint defects corresponding to various PCBA solder joint defect types. This represents the maximum number of PCBA solder joint defect types. PCBA solder joint defect types include cold solder joints, bridging, solder joint cracks, solder buildup, and solder bridging. The standard image data for different types of PCBA solder joint defects represents the image information of solder joint defects based on the standard settings for different types of PCBA solder joint defects.

[0081] S52, Set up PCBA marked solder joint feature image data PCBA marked solder joint feature image data A collection of standard image data on solder joint defects of different types of PCBAs, ordered by PCBA solder joint numbering. Standard image data of solder joint defects of different types of PCBA Image feature matching is performed, and a set of PCBA marked solder joint welding defect detection data is generated based on the image feature matching results. ; Execute to generate a set of PCBA marked solder joint welding defect detection data The specific operation steps are as follows:

[0082] S521, initialization, the position of the solder defect recognition cheetah in the search space of the different type PCBA solder joint defect standard image data set is updated according to the formula: , wherein represents the position of the solder defect recognition cheetah individual i in the search space of the different type PCBA solder joint defect standard image data set with the spatial dimension , , , respectively represent the upper limit value and the lower limit value of the search space of the different type PCBA solder joint defect standard image data set with the spatial dimension , rand is a random number between 0 and 1, and T represents the maximum number of iterations;

[0083] S522, complete initialization and search strategy search prey, the solder defect recognition cheetah performs full range scanning or active search in its territory or surrounding area to find prey; the solder defect recognition cheetah searches the different type PCBA solder joint defect standard image data that matches the PCBA marked solder joint feature image data in the search space of the different type PCBA solder joint defect standard image data set ; the mathematical description of the search strategy is

[0084] , wherein represents the position of the solder defect recognition cheetah individual i after the (t+1) th iteration in the search space of the different type PCBA solder joint defect standard image data set with the spatial dimension , represents the position of the solder defect recognition cheetah individual i after the t th iteration in the search space of the different type PCBA solder joint defect standard image data set with the spatial dimension , represents a random number normally distributed in the search space of the different type PCBA solder joint defect standard image data set with the spatial dimension after the t th iteration of the solder defect recognition cheetah individual i, represents the position of the solder defect recognition cheetah individual i after the t th iteration in the search space of the different type PCBA solder joint defect standard image data set Different types of PCBA solder joint defect standard image data sets Search step in search space

[0085] S523, search prey using a waiting strategy, in search mode, prey is exposed in the field of view of the solder joint defect identification cheetah, in different types of PCBA solder joint defect standard image data sets Search for different types of PCBA solder joint defect standard image data that match PCBA marked solder joint feature image data In the search space of different types of PCBA solder joint defect standard image data Each action of the solder joint defect identification cheetah will cause the prey to escape, and the solder joint defect identification cheetah will take a waiting ambush strategy to approach the prey, which includes lying on the ground or hiding in the bushes, and the strategy is mathematically described as ;

[0086] S524, execute the waiting strategy to approach the prey, and attack the prey according to the attack strategy, in the algorithm, each solder joint defect identification cheetah adjusts its position according to the position of the escaped prey, the leading solder joint defect identification cheetah or the nearby solder joint defect identification cheetah to obtain the best attack position, in different types of PCBA solder joint defect standard image data sets Search for different types of PCBA solder joint defect standard image data that match PCBA marked solder joint feature image data In the search space of different types of PCBA solder joint defect standard image data The attack strategy is mathematically described as , where Indicates the best position of the solder joint defect identification cheetah individual after the tth iteration in the search space of different types of PCBA solder joint defect standard image data sets In the search space of different types of PCBA solder joint defect standard image data sets In the search space of different types of PCBA solder joint defect standard image data sets The turning factor of the solder joint defect identification cheetah individual i in the search space of different types of PCBA solder joint defect standard image data sets In the search space of different types of PCBA solder joint defect standard image data sets In the search space of different types of PCBA solder joint defect standard image data sets Indicates the interaction factor of the solder joint defect identification cheetah individual i after the tth iteration in the search space of different types of PCBA solder joint defect standard image data sets In the search space of different types of PCBA solder joint defect standard image data sets In the search space of different types of PCBA solder joint defect standard image data sets

[0087] S525, repeat S522, S523, S524 until the maximum number of iterations is met, output PCBA marked solder joint feature image data Image feature matching results of different types of PCBA solder joint defect standard image data ​

[0088] S526, generating the PCBA marked solder joint welding defect detection data set according to the image feature matching result of the different types of PCBA solder joint defect standard image data and the PCBA marked solder joint feature image data PCBA marked solder joint welding defect detection data corresponding to the i-th PCBA marked solder joint;

[0089] When the image feature matching of the i-th PCBA marked solder joint and the j-th PCBA solder joint defect type is successful, it indicates that the i-th PCBA marked solder joint has the j-th PCBA solder joint defect type, and the PCBA marked solder joint welding defect detection data

[0090] When the image feature matching of the i-th PCBA marked solder joint and the j-th PCBA solder joint defect type is not successful, it indicates that the i-th PCBA marked solder joint does not have the welding joint defect, and the PCBA marked solder joint welding defect detection data

[0091] Through the PCBA marked solder joint image acquisition unit, the characteristic image of the marked solder joint in the standard PCBA solder joint marked image is autonomously and accurately collected according to the standard PCBA solder joint marked image information combined with the image cutout convenient software, and the image intelligent acquisition of each solder joint in the PCBA is realized. The PCBA marked solder joint welding defect detection unit, according to the PCBA marked solder joint feature image information combined with the AI intelligent recognition algorithm and the different types of PCBA solder joint defect standard image information based on big data storage, the welding joint defect intelligent detection of the welding joint in the PCBA is realized, the point-to-point fine detection of the PCBA welding defect based on AI is realized, and the precision and quality of the PCBA welding fault detection are improved.

[0092] Further, please refer to Figs. 1-2 ​​​​​​​​​​​​​​The steps for displaying the text information of the welding defect detection results of PCBA marked solder joints, based on the welding defect detection data of PCBA marked solder joints, are as follows: The text information of the welding defect detection results of PCBA marked solder joints is combined with standard PCBA solder joint mark image data to create a combined image and text display data of the welding defect detection results of PCBA marked solder joints.

[0093] S61. Use text editing software to process the PCBA marked solder joint welding defect detection data set. The results of the PCBA marking solder joint welding defect detection are the welding defect detection data of PCBA marking solder joints with defects. The rendering process uses a red background for highlighting, and the PCBA marking solder joint defect detection data shows no defects. The rendering process uses a green background with highlighted highlights, and generates a set of rendered text data showing the detection results of solder joint defects on PCBA. ,in Indicates the first The rendered text data represents the welding fault detection results of the PCBA marked solder joints. The rendered text data represents the text information of different welding fault detection results for PCBA marked solder joints, distinguished by color. The text editing software includes any one of Vim, EditPlus, and Word.

[0094] S62. Render the text data set of welding fault detection results marked by CBA welding points. Rendering text data of solder joint failure detection results in PCBA Image data of solder joint markings with standard PCBA In the process of identifying and matching the text information of the PCBA solder joints marked with their numerical labels, the images of the marked PCBA solder joints are mapped and matched to create an image data display of the solder joint failure detection results. The image data for displaying the welding fault detection results of PCBA marked solder joints represents the text information of the welding fault detection results of PCBA marked solder joints that has been highlighted with a different text background color and then mapped onto a standard PCBA solder joint marked image to form the image information for displaying the welding fault detection results of PCBA solder joints.

[0095] The operation steps of executing the PCBA welding fault detection result pushing output job according to the PCBA marked welding spot welding fault detection result display image data are as follows:

[0096] S71, PCBA marked welding spot welding fault detection result display image data PCBA welding fault detection result pushing output job is performed through the Internet of Things communication network to the PCBA quality inspection platform.

[0097] The PCBA marked welding spot welding fault detection result rendering text construction unit is used to realize the differential output display of the welding defect detection results of the PCBA welding spot based on the text highlight rendering processing; the PCBA marked welding spot welding fault detection result image generation unit is used to realize the visual display of the welding fault detection results of the PCBA welding spot based on the image visualization by combining the text rendering results of the welding defect detection results of the PCBA welding spot with the PCBA welding spot marked image according to the welding spot, thereby improving the readability of the PCBA welding fault detection results; the PCBA welding fault detection result pushing unit is used to realize the efficient feedback of the PCBA welding fault detection results and improve the applicability of the PCBA welding fault detection by executing the PCBA welding fault detection result pushing output job in combination with the PCBA quality inspection platform according to the PCBA marked welding spot welding fault detection result display image data.

[0098] Embodiment 2:

[0099] Please refer to Figs. 1-2 An AI-based PCBA fault detection system is used to realize an AI-based PCBA fault detection method. The system includes a PCBA welding information processing module, a PCBA welding fault detection module, and a PCBA welding fault detection result feedback module.

[0100] The PCBA welding information processing module includes a PCBA feature image acquisition unit, a PCBA feature image preprocessing unit, a PCBA welding spot standard image storage unit, and a PCBA welding spot marked image generation unit.

[0101] The PCBA feature image acquisition unit acquires PCBA feature image data through a cloud lens; the PCBA feature image preprocessing unit performs image preprocessing based on the PCBA feature image data to generate standard PCBA feature image data; the PCBA soldering solder joint standard image storage unit is used for storing PCBA soldering solder joint standard image data; the PCBA soldering solder joint marking image generation unit identifies solder joints in the standard PCBA feature image according to the standard PCBA feature image data in combination with an image marking editing software and the PCBA soldering solder joint standard image data, and performs solder joint marking processing to generate standard PCBA soldering solder joint marking image data;

[0102] The PCBA soldering fault detection module includes a PCBA marked solder joint image acquisition unit, a different type PCBA solder joint defect standard image storage unit, and a PCBA marked solder joint soldering defect detection unit.

[0103] The PCBA marked solder joint image acquisition unit performs feature image acquisition processing of the marked solder joint in the standard PCBA soldering solder joint marking image according to the standard PCBA soldering solder joint marking image data in combination with an image matting convenient software, and constructs PCBA marked solder joint feature image data; the different type PCBA solder joint defect standard image storage unit is used for storing different type PCBA solder joint defect standard image data; and the PCBA marked solder joint soldering defect detection unit performs soldering defect detection processing of the PCBA marked solder joint according to the PCBA marked solder joint feature image data and the different type PCBA solder joint defect standard image data to generate PCBA marked solder joint soldering defect detection data.

[0104] The PCBA soldering fault detection result feedback module includes a PCBA marked solder joint soldering fault detection result rendering text construction unit, a PCBA marked solder joint soldering fault detection result image generation unit, and a PCBA soldering fault detection result pushing unit.

[0105] The PCBA-marked solder joint welding fault detection result rendering text construction unit, based on the PCBA-marked solder joint welding defect detection data and combined with text editing software, performs text information rendering processing of the PCBA-marked solder joint welding defect detection result text information to generate PCBA-marked solder joint welding fault detection result rendering text data; the PCBA-marked solder joint welding fault detection result image generation unit, based on the PCBA-marked solder joint welding fault detection result rendering text data and standard PCBA solder joint mark image data, performs image and text combination processing of the PCBA-marked solder joint welding defect detection result text information and standard PCBA solder joint mark image data to construct PCBA-marked solder joint welding fault detection result display image data; the PCBA welding fault detection result push unit, based on the PCBA-marked solder joint welding fault detection result display image data and combined with the PCBA quality inspection platform, performs the PCBA welding fault detection result push output operation.

[0106] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An AI-based PCBA fault detection method, characterized in that, The method comprises the following steps: S1, collect PCBA feature image data; S2, perform image preprocessing based on the PCBA feature image data to generate standard PCBA feature image data; S3, perform standard PCBA feature image welding spot recognition and welding spot marking processing to generate standard PCBA welding spot marking image data; S4, perform standard PCBA welding spot marking image feature image acquisition processing of the marked welding spot to construct PCBA marked welding spot feature image data; S5, perform PCBA marked welding spot welding defect detection processing to generate PCBA marked welding spot welding defect detection data; The S5 comprises the following steps: S51, establish different types of PCBA welding spot defect standard image data set , ; wherein represents the first PCBA welding spot defect type corresponding to different types of PCBA welding spot defect standard image data represents the maximum value of the number of PCBA welding spot defect types; S52、the In According to the PCBA mark welding spot number order and the In the Perform image feature matching, and generate a PCBA mark welding spot welding defect detection data set according to the image feature matching result , the specific operation steps of generating the PCBA mark welding spot welding defect detection data set are as follows: wherein represents a PCBA mark solder joint feature image data set, and the PCBA mark solder joint feature image data set comprises a plurality of PCBA mark solder joint feature image data, and the PCBA mark solder joint feature image data set is represented as: comprises wherein represents the first PCBA mark solder joint corresponding to the PCBA mark solder joint feature image data. S521, initialize, the location of the search space in the spatial dimension for the spot defect recognition cheetah update is ​​ S522, completing initialization, performing search strategy, searching for prey, the welding defect identification cheetah performs full range scanning or active search in its territory or surrounding area to find prey; the welding defect identification cheetah searches in the search space of the welding defect identification cheetah to find the welding defect identification prey matching the welding defect identification prey. ;​​ S523、search prey using a wait-and-see strategy, in search mode, the prey is exposed to the weld defect identification cheetah's field of view, in the search space of the searches the matching the ; S524, execute the waiting strategy close to the prey to attack the prey according to the attack strategy, and each weld defect identification cheetah in the algorithm adjusts its position according to the position of the fleeing prey, the leading weld defect identification cheetah or the nearby weld defect identification cheetah to obtain the best attack position, in the search space of , search the most matching ; S525, repeatedly performing S522, S523, S524 until a maximum iteration number is met, and outputting the image feature matching result of the image feature matching result of the S526、According to the With the Image feature matching result of the PCBA mark welding spot Welding spot defect detection data set Wherein Indicates the PCBA mark welding spot welding defect detection data corresponding to the PCBA mark welding spot When With image feature matching is successful, output the defects exist, and output the first PCBA solder joint defect type text information; When With No image feature matching is successful, output the No defects exist; S6, perform PCBA marked welding spot welding defect detection result text information display rendering processing to generate PCBA marked welding spot welding fault detection result rendering text data and perform PCBA marked welding spot welding defect detection result text information and standard PCBA welding spot marking image graphic-text combination processing with the standard PCBA welding spot marking image data, to construct PCBA marked welding spot welding fault detection result display image data; S7, execute PCBA welding fault detection result pushing output job according to the PCBA marked welding spot welding fault detection result display image data.

2. The AI-based PCBA fault detection method of claim 1, wherein: The S1 comprises the following steps: S11, collect the overall structure feature image information of the front or back of the PCBA after completing the soldering welding operation through the cloud lens online, and generate PCBA feature image data .

3. The AI-based PCBA fault detection method of claim 2, wherein: The S2 comprises the following steps: S21, acquiring the ; S22, adopting a Gaussian filter to the standard PCBA feature image data carrying out image noise reduction preprocessing of the PCBA to generate standard PCBA feature image data .

4. The AI-based PCBA fault detection method of claim 3, wherein: The S3 comprises the following steps: S31, establish a PCBA welding solder joint standard image data set , ; wherein represents the PCBA welding solder joint standard image data, represents the maximum value of the number of PCBA welding solder joint standard images; S32. The SIFT image feature search algorithm is used to... With the The above Perform image feature matching to search and identify the... All PCBA solder joints were analyzed, and image marking and editing software was used simultaneously to mark and edit them. The PCBA solder joints searched in the database are digitally marked to construct standard PCBA solder joint marking image data. .

5. The AI-based PCBA fault detection method of claim 4, wherein: The S4 comprises the following steps: S41, acquiring the ; S42, using image matting convenient software to the The corresponding standard PCBA welding spot marking image is sequentially subjected to digital marking processing in step S32, and the welding spot feature image is subjected to matting collection processing in order according to the digital marking sequence, and a PCBA marking welding spot feature image data set is constructed , ; wherein represents the th PCBA marking welding spot corresponding to the PCBA marking welding spot feature image data, represents the maximum value of the number of PCBA marking welding spots.

6. The AI-based PCBA fault detection method of claim 5, wherein: The S6 comprises the following steps: S61, using text editing software respectively to said PCBA marked solder joint solder joint defect detection result is the existence of defects in said Using red background highlight display rendering processing, and PCBA marked solder joint solder joint defect detection result is not the existence of defects in said Using green background highlight display rendering processing, and generating PCBA marked solder joint solder joint fault detection result rendering text data set , wherein represents the PCBA marked solder joint corresponding to the PCBA marked solder joint solder joint fault detection result rendering text data; S62、the PCBA mark welding spot in the PCBA mark welding spot in the PCBA mark welding spot, PCBA mark welding spot digital mark number order PCBA mark welding spot welding defect detection result text information and PCBA welding spot mark image of the image-text mapping matching combination processing, build PCBA mark welding spot welding fault detection result display image data .

7. The AI-based PCBA fault detection method of claim 6, wherein: The S7 comprises the following steps: S71, the above The PCBA welding fault detection results are transmitted to the PCBA quality inspection platform via the Internet of Things (IoT) communication network and then pushed out.

8. An AI-based PCBA fault detection system for implementing the AI-based PCBA fault detection method of any one of claims 1-7. The system comprises a PCBA welding information processing module, a PCBA welding fault detection module, and a PCBA welding fault detection result feedback module.

Images