A visual inspection system for power circuit boards

Abstract

The present invention relates to a visual inspection system, and in particular to a visual inspection system suitable for a power supply circuit board. A detection area segmentation model can be used to effectively extract a corresponding detection area in a processed circuit board image according to a detection task. A defect detection module compares the extracted detection area with a standardized circuit board image to perform defect detection, thereby accurately extracting the detection area according to the detection task and ensuring the detection accuracy. A to-be-identified area extraction model can be used to mark the to-be-identified area in a nameplate image. A character recognition module performs character recognition on the marked to-be-identified area to obtain the power supply model and power supply number of the to-be-identified power supply, thereby effectively performing text recognition on the to-be-identified power supply nameplate. The technical solution provided by the present invention can effectively overcome the defects of the prior art that the detection accuracy is poor and the text recognition on the power supply nameplate cannot be performed.

Description

Technical Field

[0001] The invention relates to a visual inspection system, in particular to a visual inspection system suitable for a power circuit board. Background Art

[0002] The power supply comprehensive tester can complete the power supply machine assembly and electrical equipment inspection work in the continuous production mode. At present, the power supply comprehensive tester is mainly composed of machine vision recognition module, insulation performance test module, and electrical performance test module. These modules can complete the inspection tasks required for daily production.

[0003] Circuit boards, also known as printed circuit boards or printed circuit boards, play an important role in modern electronic devices. Circuit board defect detection is a key means of evaluating the quality of circuit boards, which directly affects the performance and safety of power supplies. Traditionally, circuit board defect detection is generally done manually, but this part of the work is boring and tedious, and people have limited energy, which makes it easy to miss and misdetect, and the detection efficiency is not high. At the same time, as Internet technology accelerates its transformation from a production tool to a production factor, its integration with traditional industries is becoming increasingly close. The industrial aspect is mainly reflected in the combination of the Internet and manufacturing, that is, "Internet + Manufacturing". Automatic detection of circuit board defects through machine vision has become an urgent need in the industry and an inevitable development trend.

[0004] However, the existing visual inspection system for circuit boards has poor detection accuracy and poor detection effect. There are cases of missed detection. In addition, it is unable to recognize the text on the power nameplate and cannot determine the power model and power number of the power supply to be inspected, making detection inconvenient. Summary of the invention

[0005] 1. Technical issues to be solved

[0006] In view of the above-mentioned shortcomings of the prior art, the present invention provides a visual inspection system suitable for power supply circuit boards, which can effectively overcome the defects of the prior art such as poor detection accuracy and inability to perform text recognition on power supply nameplates.

[0007] (II) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] A visual inspection system for a power circuit board, comprising a controller and

[0010] Standard information input module, used to input standardized circuit board images and power supply models to be tested;

[0011] An image processing module, used for performing image processing on the circuit board image acquired by the first detection image acquisition module;

[0012] A test project selection module is used to select circuit board test tasks;

[0013] The detection area segmentation model effectively extracts the corresponding detection area in the processed circuit board image according to the detection task;

[0014] The defect detection module compares the extracted inspection area with the standardized circuit board image to find defects;

[0015] A manual labeling module is used to manually label the area to be identified in the standardized nameplate image collected by the training image collection module;

[0016] Extracting a model for the area to be identified, training the model by manually marking a standardized nameplate image of the area to be identified, and marking the area to be identified in the nameplate image collected by the second detection image acquisition module;

[0017] A character recognition module, used for performing character recognition on the area to be recognized marked by the extraction model of the area to be recognized;

[0018] The data comparison module compares the character recognition result obtained by the character recognition module with the model of the power supply to be tested.

[0019] Preferably, the detection area segmentation model effectively extracts the corresponding detection area in the processed circuit board image through model training, including:

[0020] Collect multiple standardized circuit board images, and manually mark the corresponding inspection area in the circuit board image with a rectangular frame according to different circuit board inspection tasks to establish a training data set;

[0021] Inputting the training data set into the detection region segmentation model for model training to obtain a trained detection region segmentation model;

[0022] The processed circuit board image is input into the detection area segmentation model, and the detection area segmentation model marks the corresponding detection area in the processed circuit board image according to the detection task.

[0023] Preferably, the image processing module performs image enhancement, image noise reduction and image extraction on the circuit board image, wherein

[0024] Image enhancement: Select the grayscale transformation area, modify the grayscale value of the image through linear operation, and perform grayscale transformation on the grayscale transformation area;

[0025] Image denoising: Select a denoising window with the target pixel as the center, calculate the average value of all pixels in the denoising window to replace the original pixel value, maintain the shape and size of the window, and replace it with the average value with each pixel in the image as the center;

[0026] Image extraction: The image edge is determined by detecting the position where the grayscale value of the image changes step by step, and the processed circuit board image is obtained according to the image edge extraction.

[0027] Preferably, the inspection item selection module is used to select one or more inspection tasks from screws, cables, inductors, fuses, electrolytic capacitors, and PCB boards.

[0028] Preferably, the defect detection module performs defect detection by comparing the circuit board image of the marked detection area with the corresponding area in the standardized circuit board image pixel by pixel.

[0029] Preferably, it also includes:

[0030] The detection area customization module is used to customize the detection area in the standardized circuit board image;

[0031] A detection information statistics module is used to collect statistics on the defect detection results of the defect detection module;

[0032] The detection result display module is used to display the relevant detection results of the circuit board detection task and the statistical results of the detection information statistics module.

[0033] Preferably, the character recognition module performs character recognition on the area to be recognized marked by the area to be recognized extraction model, including:

[0034] Obtain a nameplate image with the area to be recognized and perform OCR recognition, and input the OCR recognition results into the language model one by one to obtain an OCR output sequence set;

[0035] The output sequences in the OCR output sequence set are converted into digital vectors one by one, and the dimension reduction process is performed. The digital vectors after dimension reduction are input into the recurrent neural network one by one to obtain the text sequence.

[0036] Preferably, the to-be-recognized area extraction model is trained by manually annotating standardized nameplate images of the to-be-recognized area, including:

[0037] Manually annotate the areas to be identified in the standardized nameplate images at different angles and lighting conditions, including the areas where the power model and power number are located, and establish a training data set;

[0038] The training data set is input into the to-be-identified region extraction model for model training to obtain a trained to-be-identified region extraction model.

[0039] Preferably, it also includes a robot arm control module, which is used to control the robot arm according to the defect detection result of the defect detection module and the comparison result of the data comparison module;

[0040] When the defect detection result of the defect detection module is lower than the detection result of the corresponding detection task of the good product, or the data comparison module compares and judges that the power supply model recognized by the character recognition module is inconsistent with the power supply model to be tested, the controller controls the robotic arm through the robotic arm control module to clamp the corresponding circuit board to be inspected to the unqualified product position.

[0041] (III) Beneficial effects

[0042] Compared with the prior art, the visual inspection system for power supply circuit boards provided by the present invention has the following advantages:

[0043] 1) The detection area segmentation model can effectively extract the corresponding detection area in the processed circuit board image according to the detection task. The defect detection module compares the extracted detection area with the standardized circuit board image to perform defect detection, so that the detection area can be accurately extracted according to the detection task to ensure the detection accuracy;

[0044] 2) The to-be-recognized area extraction model can mark the to-be-recognized area in the nameplate image, and the character recognition module performs character recognition on the marked to-be-recognized area to obtain the power model and power number of the power supply to be tested, thereby enabling effective text recognition on the nameplate of the power supply to be tested. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0046] Figure 1 It is a system schematic diagram of the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] A visual inspection system for power circuit boards, such as Figure 1 As shown, including a controller, and

[0049] Standard information input module, used to input standardized circuit board images and power supply models to be tested;

[0050] An image processing module, used for performing image processing on the circuit board image acquired by the first detection image acquisition module;

[0051] A test project selection module is used to select circuit board test tasks;

[0052] The detection area segmentation model effectively extracts the corresponding detection area in the processed circuit board image according to the detection task;

[0053] The defect detection module compares the extracted inspection area with the standardized circuit board image to find defects;

[0054] A manual labeling module is used to manually label the area to be identified in the standardized nameplate image collected by the training image collection module;

[0055] Extracting a model for the area to be identified, training the model by manually marking a standardized nameplate image of the area to be identified, and marking the area to be identified in the nameplate image collected by the second detection image acquisition module;

[0056] A character recognition module, used for performing character recognition on the area to be recognized marked by the extraction model of the area to be recognized;

[0057] The data comparison module compares the character recognition result obtained by the character recognition module with the model of the power supply to be tested.

[0058] The image processing module performs image enhancement, image noise reduction, and image extraction on the circuit board image.

[0059] Image enhancement: Select the grayscale transformation area, modify the grayscale value of the image through linear operation, and perform grayscale transformation on the grayscale transformation area;

[0060] Image denoising: Select a denoising window with the target pixel as the center, calculate the average value of all pixels in the denoising window to replace the original pixel value, maintain the shape and size of the window, and replace it with the average value with each pixel in the image as the center;

[0061] Image extraction: The image edge is determined by detecting the position where the grayscale value of the image changes step by step, and the processed circuit board image is obtained according to the image edge extraction.

[0062] In the technical solution of the present application, the inspection item selection module is used to select one or more inspection tasks from screws, cables, inductors, fuses, electrolytic capacitors, and PCB boards.

[0063] The detection area segmentation model effectively extracts the corresponding detection area in the processed circuit board image through model training, including:

[0064] Collect multiple standardized circuit board images, and manually mark the corresponding inspection area in the circuit board image with a rectangular frame according to different circuit board inspection tasks to establish a training data set;

[0065] Inputting the training data set into the detection region segmentation model for model training to obtain a trained detection region segmentation model;

[0066] The processed circuit board image is input into the detection area segmentation model, and the detection area segmentation model marks the corresponding detection area in the processed circuit board image according to the detection task.

[0067] The defect detection module performs defect detection by comparing the circuit board image of the marked detection area with the corresponding area in the standardized circuit board image pixel by pixel.

[0068] The technical solution of this application also includes:

[0069] The detection area customization module is used to customize the detection area in the standardized circuit board image;

[0070] A detection information statistics module is used to collect statistics on the defect detection results of the defect detection module;

[0071] The detection result display module is used to display the relevant detection results of the circuit board detection task and the statistical results of the detection information statistics module.

[0072] When the user customizes the inspection area in the standardized circuit board image through the inspection area customization module, the inspection area segmentation model directly extracts the corresponding inspection area in the processed circuit board image based on the customized annotated inspection area, and the defect detection module performs defect detection through pixel-by-pixel comparison, making the setting of the inspection area more flexible.

[0073] The model for extracting the area to be identified is trained by manually annotating standardized nameplate images of the area to be identified, including:

[0074] Manually annotate the areas to be identified in the standardized nameplate images at different angles and lighting conditions, including the areas where the power model and power number are located, and establish a training data set;

[0075] The training data set is input into the to-be-identified region extraction model for model training to obtain a trained to-be-identified region extraction model.

[0076] The character recognition module performs character recognition on the area to be recognized marked by the area extraction model, including:

[0077] Obtain a nameplate image with the area to be recognized and perform OCR recognition, and input the OCR recognition results into the language model one by one to obtain an OCR output sequence set;

[0078] The output sequences in the OCR output sequence set are converted into digital vectors one by one, and the dimension reduction process is performed. The digital vectors after dimension reduction are input into the recurrent neural network one by one to obtain the text sequence.

[0079] Among them, the reduced-dimensional digital vectors are input into the recurrent neural network one by one to obtain the text sequence, including:

[0080] The reduced digital vector is input into the Bi-LSTM encoder to generate a feature vector, and the feature vector is input into the Bi-LSTM decoder to obtain an output vector;

[0081] The output vector is input into the Softmax algorithm module to obtain the word ID, and the word ID is converted into a text sequence according to the corresponding relationship of the dictionary.

[0082] The technical solution of the present application also includes a robot arm control module, which is used to control the robot arm according to the defect detection results of the defect detection module and the comparison results of the data comparison module.

[0083] When the defect detection result of the defect detection module is lower than the detection result of the corresponding detection task of the good product, or the data comparison module compares and judges that the power supply model recognized by the character recognition module is inconsistent with the power supply model to be tested, the controller controls the robotic arm through the robotic arm control module to clamp the corresponding circuit board to be inspected to the unqualified product position.

[0084] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A visual inspection system for power circuit boards, characterized in that: Includes controller, and Standard information input module, used to input standardized circuit board images and power supply models to be tested; An image processing module, used for performing image processing on the circuit board image acquired by the first detection image acquisition module; A test project selection module is used to select circuit board test tasks; The detection area segmentation model effectively extracts the corresponding detection area in the processed circuit board image according to the detection task; The defect detection module compares the extracted inspection area with the standardized circuit board image to find defects; A manual labeling module is used to manually label the area to be identified in the standardized nameplate image collected by the training image collection module; Extracting a model for the area to be identified, training the model by manually marking a standardized nameplate image of the area to be identified, and marking the area to be identified in the nameplate image collected by the second detection image acquisition module; A character recognition module, used for performing character recognition on the area to be recognized marked by the extraction model of the area to be recognized; A data comparison module compares the character recognition result obtained by the character recognition module with the model of the power supply to be tested; The to-be-recognized area extraction model is trained by manually annotating standardized nameplate images of the to-be-recognized area, including: Manually annotate the areas to be identified in the standardized nameplate images at different angles and lighting conditions, including the areas where the power model and power number are located, and establish a training data set; Inputting the training data set into the to-be-recognized region extraction model for model training to obtain a trained to-be-recognized region extraction model; The character recognition module performs character recognition on the area to be recognized marked by the area to be recognized extraction model, including: Obtain a nameplate image with the area to be recognized and perform OCR recognition, and input the OCR recognition results into the language model one by one to obtain an OCR output sequence set; The output sequences in the OCR output sequence set are converted into digital vectors one by one, and the dimension reduction processing is performed, and the digital vectors after dimension reduction are input into the recurrent neural network one by one to obtain the text sequence; The step of inputting the dimension-reduced digital vectors into the recurrent neural network one by one to obtain a text sequence includes: The reduced digital vector is input into the Bi-LSTM encoder to generate a feature vector, and the feature vector is input into the Bi-LSTM decoder to obtain an output vector; The output vector is input into the Softmax algorithm module to obtain the word ID, and the word ID is converted into a text sequence according to the corresponding relationship of the dictionary.

2. The visual inspection system for power circuit boards according to claim 1, characterized in that: The detection area segmentation model effectively extracts the corresponding detection area in the processed circuit board image through model training, including: Collect multiple standardized circuit board images, and manually mark the corresponding inspection area in the circuit board image with a rectangular frame according to different circuit board inspection tasks to establish a training data set; Inputting the training data set into the detection region segmentation model for model training to obtain a trained detection region segmentation model; The processed circuit board image is input into the detection area segmentation model, and the detection area segmentation model marks the corresponding detection area in the processed circuit board image according to the detection task.

3. The visual inspection system for power circuit boards according to claim 2, characterized in that: The image processing module performs image enhancement, image noise reduction, and image extraction on the circuit board image, wherein Image enhancement: Select the grayscale transformation area, modify the grayscale value of the image through linear operation, and perform grayscale transformation on the grayscale transformation area; Image denoising: Select a denoising window with the target pixel as the center, calculate the average value of all pixels in the denoising window to replace the original pixel value, maintain the shape and size of the window, and replace it with the average value with each pixel in the image as the center; Image extraction: The image edge is determined by detecting the position where the grayscale value of the image changes step by step, and the processed circuit board image is obtained according to the image edge extraction.

4. The visual inspection system for power circuit boards according to claim 2, characterized in that: The detection item selection module is used to select one or more detection tasks from screws, cables, inductors, fuses, electrolytic capacitors, and PCB boards.

5. The visual inspection system for power circuit boards according to claim 3, characterized in that: The defect detection module performs defect detection by comparing the circuit board image with the marked detection area with the corresponding area in the standardized circuit board image pixel by pixel.

6. The visual inspection system for power circuit boards according to claim 5, characterized in that: Also includes: The detection area customization module is used to customize the detection area in the standardized circuit board image; A detection information statistics module is used to collect statistics on the defect detection results of the defect detection module; The detection result display module is used to display the relevant detection results of the circuit board detection task and the statistical results of the detection information statistics module.

7. The visual inspection system for power circuit boards according to claim 1, characterized in that: It also includes a robot arm control module, which is used to control the robot arm according to the defect detection result of the defect detection module and the comparison result of the data comparison module; When the defect detection result of the defect detection module is lower than the detection result of the corresponding detection task of the good product, or the data comparison module compares and judges that the power supply model recognized by the character recognition module is inconsistent with the power supply model to be tested, the controller controls the robotic arm through the robotic arm control module to clamp the corresponding circuit board to be inspected to the unqualified product position.

Images