A PCB board detection method and system based on visual neural network algorithm

Through the PCB board detection method based on visual neural network algorithm, a neural network model is constructed, and the problem of low efficiency in detecting different PCB boards in the existing technology is solved, achieving rapid adaptability and efficient conduction detection.

CN117893513BActive Publication Date: 2024-11-12GUILIN SHIYU ELECTRONIC TECH CO LTD
View PDF 10 Cites 0 Cited by

Patent Information

Application Number
CN202410082965.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-01-19
Publication Date
2024-11-12
Estimated Expiration
2044-01-19

AI Technical Summary

Technical Problem

The existing needle bed testers and fly needle testers are less efficient when detecting different types of PCB boards, and require different fixtures to be replaced, which cannot quickly adapt to the detection needs of different PCB boards.

Method used

Using the PCB board detection method based on visual neural network algorithm, the initial neural network model is constructed, and multiple different PCB board unit images are used for training to generate the target neural network model, which is used to identify and detect the conduction paths on the PCB board, and realize the Rapid detection of different PCB boards.

Benefits of technology

It improves the efficiency and accuracy of PCB board detection, can adapt to different types of PCB boards, realizes fast conduction detection and outputs detection results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117893513B_ABST
    Figure CN117893513B_ABST
Patent Text Reader

Abstract

The present application provides a PCB board detection method based on a visual neural network algorithm, the method comprising: acquiring multiple different PCB board unit images, acquiring multiple first position data of multiple training identification point images based on the PCB board unit images, second position data of the training path image, and third position data representing the edge contour of the PCB board unit. Obtain a target neural network model. Acquire and detect a PCB board unit image, acquire multiple identification points based on the detected PCB board unit image, acquire multiple fourth position data of multiple identification point images, and fifth position data representing the edge contour of the detected PCB board unit, import the fourth position data and the fifth position data into the target neural network model, and obtain sixth position data. Acquire a target path image based on the sixth position data, determine the target point position based on the target path image, perform conduction detection on the target point position, and output the detection result. The present application also provides a PCB board detection system based on a visual neural network algorithm.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of detection technology, and in particular to a PCB board detection method and system based on a visual neural network algorithm. Background Art

[0002] PCB boards are the foundation of the industrial electronics field. In the process of producing PCB boards, there are many ways to detect PCB boards, such as using bed of nails testers and flying probe testers to identify PCB manufacturing defects through electrical performance testing. The main advantages are low testing costs, intelligent ability to quickly check short circuits, open circuits and component values, and provide designers with fast feedback in the prototype stage.

[0003] In the related art, when using a bed of needles tester and a flying probe tester for testing, due to the use of a fixed fixture, only one type of PCB board with fixed pins can be tested. When different types of PCB boards need to be tested, different fixtures need to be used respectively, and the testing efficiency is low. Summary of the invention

[0004] The embodiment of the present application provides a PCB board detection method system based on a visual neural network algorithm to improve the above-mentioned problems.

[0005] In order to achieve the above objectives, this application adopts the following technical solutions:

[0006] In the first aspect, the present application proposes a PCB board detection method based on a visual neural network algorithm, the method comprising:

[0007] A plurality of different PCB board unit images are obtained, wherein the edge of the PCB board unit has a plurality of training identification points, one of the plurality of training identification points is connected with one or more other training identification points to form a training path, and the PCB board unit image includes a plurality of training identification point images and a plurality of training path images; based on the PCB board unit image, a plurality of first position data of the plurality of training identification point images, a second position data of the training path image, and a third position data representing the edge contour of the PCB board unit are obtained; an initial neural network training model is constructed and iterative calculations are performed, and when the initial neural network model meets a preset condition, the training is stopped and a target neural network model is obtained, and the target neural network training model is used to generate the second position data according to the first position data and the third position data; a detection PCB board unit image is obtained, a plurality of identification points are obtained based on the detection PCB board unit image, a plurality of fourth position data of the plurality of identification point images, and a fifth position data representing the edge contour of the detection PCB board unit are obtained, and the fourth position data and the fifth position data are imported into the target neural network model to obtain the sixth position data; a target path image is obtained based on the sixth position data, and a target point position is determined based on the target path image, wherein the target point position is a point position on the detection PCB board unit, a conduction detection is performed on the target point position, and a detection result is output.

[0008] In combination with the first aspect, in some feasible implementations, multiple first position data of multiple training identification point images, second position data of the training path image, and third position data representing the edge contour of the PCB board unit are obtained based on the PCB board unit image, including: obtaining training visual path data based on the PCB board unit image, wherein the training visual path is a conductive path that can be directly obtained on the PCB board unit, and the training visual path data is the position data of the conductive path; and using the training visual path data as the second position data.

[0009] In combination with the first aspect, in some feasible implementations, an initial neural network training model is constructed and iterative calculations are performed. When the initial neural network model meets preset conditions, the training is stopped and a target neural network model is obtained. The target neural network training model is used to generate second position data based on the first position data and the third position data, including: obtaining predicted second position data based on the initial neural network model; comparing the predicted second position data with the second position data and obtaining a correction value; stopping training when the correction value is less than a preset threshold, and using the initial neural network training model at this time as the target neural network training model.

[0010] In combination with the first aspect, in some feasible implementations, the predicted second position data is compared with the second position data, and a correction value is obtained, including: obtaining the correction value, wherein the relationship between the correction value and the predicted second position data and the second position data satisfies:

[0011] t=|(|y1-yt|)y1|

[0012] Among them, the error value is t, the predicted second position data is yt, and the second position data is y1.

[0013] In combination with the first aspect, in some feasible implementations, a detection PCB board unit image is taken, multiple identification points are obtained based on the detection PCB board unit image, multiple fourth position data of the multiple identification point images are obtained, and fifth position data representing the edge contour of the detection PCB board unit are obtained, and the fourth position data and the fifth position data are imported into the target neural network model to obtain sixth position data, including: obtaining detection visual path data based on the PCB board unit image, wherein the detection visual path is a conductive path that can be directly obtained on the detection PCB board unit, and the detection visual path data is the position data of the conductive path; and using the detection visual path data as the sixth position data.

[0014] In combination with the first aspect, in some feasible implementations, an initial neural network training model is constructed and iterative calculations are performed. When the initial neural network model meets preset conditions, training is stopped and a target neural network model is obtained. The target neural network training model is used to generate second position data based on the first position data and the third position data. It also includes: using the first position data and the third position data as different dimensions to generate a feature vector of the second position data; generating the initial neural network training model based on the feature vector of the second position data.

[0015] In combination with the first aspect, in some feasible implementations, a detection PCB board unit image is obtained, multiple identification points are obtained based on the detection PCB board unit image, multiple fourth position data of the multiple identification point images are obtained, and fifth position data representing the edge contour of the detection PCB board unit are obtained, the fourth position data and the fifth position data are imported into the target neural network model to obtain sixth position data, and also includes: importing the detection visual path data as the sixth position data into the target neural network model, wherein the feature vector of the sixth data is composed of the fourth data and the fifth data; and obtaining all the sixth position data based on the target neural network model.

[0016] In combination with the first aspect, in some feasible implementations, the first position data and the third position data are taken as different dimensions, and before the feature vector of the second position data is generated, it also includes: normalizing each dimension so that each dimension is within the range of [0,1]; performing noise reduction on each dimension; and performing data cleaning on each dimension.

[0017] In combination with the first aspect, in some feasible implementations, a target path image is obtained based on the sixth position data, and a target point is determined based on the target path image, wherein the target point is a point on a detection PCB board unit, and a continuity detection is performed on the target point, and a detection result is output, including: generating a mapping path image based on the sixth position data, and determining the target path image based on the mapping path image; determining the target path based on the target path image, wherein the target path is a conduction path on the detection PCB board unit; determining the target point based on the target path, wherein the target point is a conduction pin point on the detection PCB board.

[0018] In a second aspect, an embodiment of the present application proposes a PCB board detection system based on a visual neural network algorithm, and the system is configured to: obtain a plurality of different PCB board unit images, wherein the edge of the PCB board unit has a plurality of training identification points, one of the plurality of training identification points is connected with one or more other training identification points to form a training path, and the PCB board unit image includes a plurality of training identification point images and a plurality of training path images; based on the PCB board unit image, obtain a plurality of first position data of the plurality of training identification point images, a second position data of the training path image, and a third position data representing the edge contour of the PCB board unit; construct an initial neural network training model and perform iterative calculations, when the initial neural network When the model meets the preset conditions, the training is stopped and the target neural network model is obtained, and the target neural network training model is used to generate the second position data according to the first position data and the third position data; the detection PCB board unit image is obtained, and multiple identification points are obtained based on the detection PCB board unit image, and multiple fourth position data of the multiple identification point images and fifth position data representing the edge contour of the detection PCB board unit are obtained, and the fourth position data and the fifth position data are imported into the target neural network model to obtain the sixth position data; the target path image is obtained based on the sixth position data, and the target point position is determined based on the target path image, wherein the target point position is a point position on the detection PCB board unit, the target point position is subjected to conduction detection, and the detection result is output.

[0019] In conjunction with the second aspect, in some embodiments, the system is configured as follows:

[0020] The training visual path data is acquired based on the PCB board unit image, wherein the training visual path is a conductive path that can be directly acquired on the PCB board unit, and the training visual path data is position data of the conductive path; and the training visual path data is used as the second position data.

[0021] In conjunction with the second aspect, in some embodiments, the system is configured as follows:

[0022] Acquire predicted second position data based on the initial neural network model; compare the predicted second position data with the second position data and obtain a correction value; stop training when the correction value is less than a preset threshold, and use the initial neural network training model at this time as the target neural network training model.

[0023] In conjunction with the second aspect, in some embodiments, the system is configured as follows:

[0024] Obtain a correction value, and the relationship between the correction value and the predicted second position data and the second position data satisfies:

[0025] t=|(|y1-yt|)y1|

[0026] Among them, the error value is t, the predicted second position data is yt, and the second position data is y1.

[0027] In conjunction with the second aspect, in some embodiments, the system is configured as follows:

[0028] The detection visual path data is acquired based on the PCB board unit image, wherein the detection visual path is a conductive path that can be directly acquired on the PCB board unit, and the detection visual path data is position data of the conductive path; the detection visual path data is used as the sixth position data.

[0029] In conjunction with the second aspect, in some embodiments, the system is configured as follows:

[0030] The first position data and the third position data are used as different dimensions to generate a feature vector of the second position data; and an initial neural network training model is generated according to the feature vector of the second position data.

[0031] In conjunction with the second aspect, in some embodiments, the system is configured as follows:

[0032] The detected visual path data is imported into the target neural network model as the sixth position data, wherein the feature vector of the sixth data is composed of the fourth data and the fifth data; and all the sixth position data are obtained based on the target neural network model.

[0033] In conjunction with the second aspect, in some embodiments, the system is configured as follows:

[0034] Each dimension is normalized so that each dimension is in the range of [0,1]; each dimension is denoised; and each dimension is cleaned.

[0035] In conjunction with the second aspect, in some embodiments, the system is configured as follows:

[0036] A mapping path image is generated based on the sixth position data, and a target path image is determined based on the mapping path image; a target path is determined based on the target path image, and the target path is a conductive path on a detection PCB board unit; a target point is determined based on the target path, and the target point is a conductive pin point on the detection PCB board.

[0037] The third aspect of the present application proposes an electronic device, which includes: a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; the memory is used to store computer programs; when the processor is used to execute the program stored in the memory, the method steps of the first aspect are implemented.

[0038] In summary, the above method and device have the following technical effects:

[0039] The present application proposes a PCB board detection method based on a visual neural network algorithm. First, an initial neural network training model is constructed by using multiple different PCB board unit images as training sets and iterative calculations are performed. When the initial neural network model meets the preset conditions, the training is stopped and a target neural network model is obtained. The target neural network training model is used to generate second position data according to first position data and third position data. Then, the detection PCB board unit image is obtained, multiple identification points are obtained based on the detection PCB board unit image, multiple fourth position data of the multiple identification point images, and fifth position data representing the edge contour of the detection PCB board unit are obtained. The fourth position data and the fifth position data are imported into the target neural network model to obtain sixth position data. Finally, a target path image is obtained based on the sixth position data, and a target point position is determined based on the target path image, wherein the target point position is a point position on the detection PCB board unit, a conduction detection is performed on the target point position, and a detection result is output. The present application proposes a PCB board detection method based on a visual neural network algorithm. Through training, a neural network model that iteratively calculates data of the PCB board edge, training identification points, and training path images can be obtained. During detection, by inputting data in the detection PCB board unit image into the neural network model, the conductive path on the detection PCB board can be obtained, and by performing a conductive detection on the conductive path, the detection result of the detection PCB board can be obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A schematic diagram of the flow of a PCB board detection method based on a visual neural network algorithm proposed in an embodiment of the present application. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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.

[0042] When using a bed of nails tester or a flying probe tester to test a PCB board, it is first necessary to determine the conductive path before testing. Since some paths on the PCB board are hidden under the isolation layer or blocked by some components, they cannot be directly obtained through conventional visual matching.

[0043] With the popularization of neural network algorithms, since the wiring on PCB boards of different shapes has certain rules, the inventors thought of using machine learning to calculate the wiring method on the PCB board from the image information that can be directly obtained, and control the probes of the bed of nails tester and flying probe tester to detect the wiring path.

[0044] Therefore, the present application embodiment proposes a PCB board detection method based on a visual neural network algorithm, please refer to Figure 1 , including the following steps:

[0045] S101: Acquire multiple different PCB board unit images, wherein the edge of the PCB board unit has multiple training identification points, one of the multiple training identification points is connected with one or more other training identification points to form a training path, and the PCB board unit image includes multiple training identification point images and multiple training path images.

[0046] It can be understood that the PCB board unit in this step is used to train the neural network model, and the data obtained from the PCB board unit image constitutes the training set of the initial neural network model. In this embodiment, the training identification point is the pin point directly connected to the bed of needles tester and the flying probe tester on the PCB board unit, and the pin point to be detected is generally set at the edge position of the PCB board unit. Of course, in other embodiments, the pin point to be detected can also be set at other positions on the PCB board, which is not limited here.

[0047] S102: Acquire a plurality of first position data of a plurality of training recognition point images, second position data of a training path image, and third position data representing an edge contour of the PCB unit based on the PCB unit image.

[0048] It is understandable that the first position data may be a quantitative representation of the position data in the PCB unit image, for example, in the position (x, y), the position data may be represented by N=(x+y)xy, etc. The specific quantization method is not limited in this embodiment.

[0049] The training path image is the path marked in advance on the PCB unit. The specific marking method can be manual marking, and then the marked path is also processed by data. It can be understood that the marked path here is a hidden path that cannot be directly obtained through image recognition.

[0050] For a portion of the path that can be directly seen, optionally, step S102 may further include:

[0051] Acquire training visual path data based on the PCB board unit image, wherein the training visual path is a conductive path that can be directly acquired on the PCB board unit, and the training visual path data is position data of the conductive path;

[0052] The training visual path data is used as the second position data.

[0053] It is understandable that the conductive paths that can be directly obtained on some PCB board structures can be directly or indirectly marked through image recognition methods such as feature recognition. The specific recognition method has been disclosed in the relevant technology and is not limited here.

[0054] S103: construct an initial neural network training model and perform iterative calculations, stop training when the initial neural network model meets preset conditions and obtain a target neural network model, the target neural network training model is used to generate second position data based on the first position data and the third position data.

[0055] Specifically, as an implementation method, the first position data and the third position data can be used as different dimensions to generate a feature vector of the second position data, and then an initial neural network training model is generated based on the feature vector of the second position data. The feature vector contains information of multiple dimensions and can be used as an input feature of the neural network training model for model training and prediction.

[0056] Optionally, before taking the first position data and the third position data as different dimensions and generating the feature vector of the second position data, a data processing step is also included. Specifically, the data processing may include the following steps:

[0057] Each dimension is normalized so that it lies in the range [0,1].

[0058] It is understandable that normalizing each dimension can make data of different dimensions comparable and avoid different effects of data of different scales on model training. Specifically, let x be the original data of a certain dimension, xmin and xmax be the minimum and maximum values ​​of the dimension respectively, then the normalized value of the dimension is:

[0059] y=(x-xmin) / (xmax-xmin)

[0060] Among them, y is the normalized value, and its value range is [0,1]. This method can retain the relative size relationship of the data and make the data comparable. In addition, for some special cases, such as data with outliers or abnormal values, you can consider using the RobustScaler normalization method based on the median and quartiles. This method can robustly handle outliers and avoid the influence of outliers on the normalization results.

[0061] Perform noise reduction on each dimension.

[0062] It can be understood that the moving average filter can average the data within a certain time window to remove the impact of instantaneous abnormal data, thereby retaining the trend and regularity of the data. Specifically, for a data sequence x(1), x(2), ..., x(N) of length N, a window size w can be defined, and the window is placed on the data sequence each time, and the average value of the data in the window is calculated as the estimated value of the current data point.

[0063] Data cleaning is performed for each dimension.

[0064] It is understandable that data cleaning may include checking whether the data is complete, specifically, by checking whether there are missing values ​​or abnormal values ​​in the data. If there are, the missing values ​​need to be filled or deleted, and the abnormal values ​​need to be corrected or deleted. In some other embodiments, it may also include the detection of data format or repeatability, and the specific content of data cleaning is not limited here.

[0065] Optionally, step S102 may further include:

[0066] Acquire predicted second position data based on the initial neural network model;

[0067] Comparing the predicted second position data with the second position data, and obtaining a correction value;

[0068] When the correction value is less than the preset threshold, the training is stopped, and the initial neural network training model at this time is used as the target neural network training model.

[0069] Specifically, in this step, a fully connected neural network training model commonly used in deep learning can be used. This model can model the nonlinear relationship between input features and output values ​​through the calculation of multiple layers of neurons. It can be understood that in this embodiment, the feature vector is set to x = [y1, x2], where y1 and x2 represent the normalized first position data and third position data respectively, and the neural network training model is set to f (x; θ), where θ represents the model parameters, so that the model output value f (x; θ) can be close to the true value y of the second position data. Then the training process of the neural network training model can be achieved by minimizing the loss function, that is:

[0070] L(θ)=1 / N*Σ(f(x_i;θ)-y_i)^2

[0071] Where N represents the number of training samples, xi and yi represent the input features and output values ​​of the i-th sample respectively. This loss function can measure the error between the model prediction value and the true value. Through optimization algorithms such as back propagation algorithm and gradient descent, the model parameters θ can be updated to improve the accuracy of the model.

[0072] Specifically, a correction value is obtained, and the relationship between the correction value and the predicted second position data and the second position data satisfies:

[0073] t=|(|y1-yt|)y1|

[0074] Among them, the error value is t, the predicted second position data is yt, and the second position data is y1.

[0075] It is understandable that when the error value is less than the preset value, the training process of the neural network training model is stopped. This is because when the error value is less than a certain threshold, the model has predicted the data accurately enough, and continuing to train the model will not be effective and may cause problems such as overfitting.

[0076] S104: Acquire a detected PCB board unit image, acquire multiple identification points based on the detected PCB board unit image, acquire multiple fourth position data of the multiple identification point images, and fifth position data representing the edge contour of the detected PCB board unit, import the fourth position data and the fifth position data into the target neural network model to obtain sixth position data.

[0077] It can be understood that the prediction process of the target path can be realized through this step.

[0078] Optionally, this step may also include:

[0079] Acquire detection visual path data based on the PCB board unit image, wherein the detection visual path is a conductive path that can be directly obtained on the PCB board unit, and the detection visual path data is position data of the conductive path;

[0080] The detected visual path data is used as the sixth position data.

[0081] It is understandable that, since the detected visual path can be directly obtained, in order to improve the accuracy, the detected visual path data can be added as a sample to the target neural network model to obtain a more accurate result value.

[0082] Specifically, first, the detected visual path data is imported into the target neural network model as the sixth position data, wherein the feature vector of the sixth data is composed of the fourth data and the fifth data, and then all the sixth position data are obtained based on the target neural network model.

[0083] S105: acquiring a target path image based on the sixth position data, determining a target point based on the target path image, wherein the target point is a point on a detection PCB board unit, performing a continuity detection on the target point, and outputting a detection result.

[0084] Specifically, first, a mapping path image can be generated based on the sixth position data, a target path image can be determined based on the mapping path image, then a target path can be determined based on the target path image, the target path is a conductive path for detecting a PCB board unit, and finally, a target point can be determined based on the target path, the target point is a conductive pin point for detecting a PCB board. The state of the PCB board can be detected by measuring the current and voltage between the conductive pin points. The specific detection method will not be described in detail here.

[0085] The present application proposes a PCB board detection method based on a visual neural network algorithm. First, an initial neural network training model is constructed by using multiple different PCB board unit images as training sets and iterative calculations are performed. When the initial neural network model meets the preset conditions, the training is stopped and a target neural network model is obtained. The target neural network training model is used to generate second position data according to first position data and third position data. Then, the detection PCB board unit image is obtained, multiple identification points are obtained based on the detection PCB board unit image, multiple fourth position data of the multiple identification point images, and fifth position data representing the edge contour of the detection PCB board unit are obtained. The fourth position data and the fifth position data are imported into the target neural network model to obtain sixth position data. Finally, a target path image is obtained based on the sixth position data, and a target point position is determined based on the target path image, wherein the target point position is a point position on the detection PCB board unit, a conduction detection is performed on the target point position, and a detection result is output. The present application proposes a PCB board detection method based on a visual neural network algorithm. Through training, a neural network model that iteratively calculates data of the PCB board edge, training identification points, and training path images can be obtained. During detection, by inputting data in the detection PCB board unit image into the neural network model, the conductive path on the detection PCB board can be obtained, and by performing a conductive detection on the conductive path, the detection result of the detection PCB board can be obtained.

[0086] Based on the same inventive concept, the embodiment of the present application also proposes a PCB board detection system based on a visual neural network algorithm, and the system is configured as follows:

[0087] Acquire a plurality of different PCB board unit images, wherein the edge of the PCB board unit has a plurality of training identification points, one of the plurality of training identification points is connected with one or more other training identification points to form a training path, and the PCB board unit image includes a plurality of training identification point images and a plurality of training path images; acquire a plurality of first position data of the plurality of training identification point images, a second position data of the training path image, and a third position data representing the edge contour of the PCB board unit based on the PCB board unit image; construct an initial neural network training model and perform iterative calculations, stop training when the initial neural network model meets a preset condition and obtain a target neural network model, the target neural network training model is used to generate second position data according to the first position data and the third position data;

[0088] Acquire a detection PCB board unit image, acquire multiple identification points based on the detection PCB board unit image, acquire multiple fourth position data of the multiple identification point images, and fifth position data representing the edge contour of the detection PCB board unit, import the fourth position data and the fifth position data into the target neural network model to obtain sixth position data; acquire a target path image based on the sixth position data, determine the target point position based on the target path image, wherein the target point position is a point position on the detection PCB board unit, perform conduction detection on the target point position, and output the detection result.

[0089] In some embodiments, the system is configured to:

[0090] The training visual path data is acquired based on the PCB board unit image, wherein the training visual path is a conductive path that can be directly acquired on the PCB board unit, and the training visual path data is position data of the conductive path; and the training visual path data is used as the second position data.

[0091] In some embodiments, the system is configured to:

[0092] Acquire predicted second position data based on the initial neural network model; compare the predicted second position data with the second position data and obtain a correction value; stop training when the correction value is less than a preset threshold, and use the initial neural network training model at this time as the target neural network training model.

[0093] In some embodiments, the system is configured to:

[0094] Obtain a correction value, and the relationship between the correction value and the predicted second position data and the second position data satisfies:

[0095] t=|(|y1-yt|)y1|

[0096] Among them, the error value is t, the predicted second position data is yt, and the second position data is y1.

[0097] In some embodiments, the system is configured to:

[0098] The detection visual path data is acquired based on the PCB board unit image, wherein the detection visual path is a conductive path that can be directly acquired on the PCB board unit, and the detection visual path data is position data of the conductive path; the detection visual path data is used as the sixth position data.

[0099] In some embodiments, the system is configured to:

[0100] The first position data and the third position data are used as different dimensions to generate a feature vector of the second position data; and an initial neural network training model is generated according to the feature vector of the second position data.

[0101] In some embodiments, the system is configured to:

[0102] The detected visual path data is imported into the target neural network model as the sixth position data, wherein the feature vector of the sixth data is composed of the fourth data and the fifth data; and all the sixth position data are obtained based on the target neural network model.

[0103] In some embodiments, the system is configured to:

[0104] Each dimension is normalized so that each dimension is in the range of [0,1]; each dimension is denoised; and each dimension is cleaned.

[0105] In some embodiments, the system is configured to:

[0106] A mapping path image is generated based on the sixth position data, and a target path image is determined based on the mapping path image; a target path is determined based on the target path image, and the target path is a conductive path on a detection PCB board unit; a target point is determined based on the target path, and the target point is a conductive pin point on the detection PCB board.

[0107] The present application proposes a PCB board detection system based on a visual neural network algorithm. First, an initial neural network training model is constructed by using multiple different PCB board unit images as training sets and iterative calculations are performed. When the initial neural network model meets the preset conditions, the training is stopped and a target neural network model is obtained. The target neural network training model is used to generate second position data according to first position data and third position data. Then, the detection PCB board unit image is obtained, multiple identification points are obtained based on the detection PCB board unit image, multiple fourth position data of the multiple identification point images, and fifth position data representing the edge contour of the detection PCB board unit are obtained. The fourth position data and the fifth position data are imported into the target neural network model to obtain sixth position data. Finally, a target path image is obtained based on the sixth position data, and a target point position is determined based on the target path image, wherein the target point position is a point position on the detection PCB board unit, a conduction detection is performed on the target point position, and a detection result is output. The present application proposes a PCB board detection system based on a visual neural network algorithm. Through training, a neural network model can be obtained which is iteratively calculated by data of PCB board edges, training identification points and training path images. During detection, by inputting data in the detection PCB board unit image into the neural network model, the conductive path on the detection PCB board can be obtained, and by performing a conductive detection on the conductive path, the detection result of the detection PCB board can be obtained.

[0108] Based on the same inventive concept, an embodiment of the present application further proposes an electronic device, the electronic device comprising:

[0109] At least one processor; and a memory in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the PCB board detection method based on the visual neural network algorithm of the embodiment of the present application.

[0110] In addition, to achieve the above-mentioned purpose, an embodiment of the present application also proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the PCB board detection method based on the visual neural network algorithm of the embodiment of the present application is implemented.

[0111] The following is a detailed introduction to the various components of electronic equipment:

[0112] The processor is the control center of the electronic device, which can be a processor or a general term for multiple processing elements. For example, the processor is one or more central processing units (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention, such as one or more microprocessors (digital signal processors, DSP), or one or more field programmable gate arrays (field programmable gate arrays, FPGA).

[0113] Optionally, the processor can perform various functions of the electronic device by running or executing a software program stored in the memory, and calling data stored in the memory.

[0114] The memory is used to store the software program for executing the solution of the present invention, and the execution is controlled by the processor. The specific implementation method can refer to the above method embodiment and will not be repeated here.

[0115] Optionally, the memory may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory may be integrated with the processor, or may exist independently and be coupled to the processor through an interface circuit of the electronic device, and the embodiments of the present invention do not specifically limit this.

[0116] A transceiver is used to communicate with a network device or a terminal device.

[0117] Optionally, the transceiver may include a receiver and a transmitter, wherein the receiver is used to implement a receiving function, and the transmitter is used to implement a sending function.

[0118] Optionally, the transceiver may be integrated with the processor, or may exist independently and be coupled to the processor via an interface circuit of the router, which is not specifically limited in the embodiment of the present invention.

[0119] In addition, the technical effects of the electronic device can refer to the technical effects of the data transmission method in the above method embodiment, which will not be repeated here.

[0120] It should be understood that the processor in the embodiment of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0121] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct rambus RAM (DR RAM).

[0122] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When a computer instruction or computer program is loaded or executed on a computer, a process or function according to an embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0123] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0124] In the present invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0125] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0126] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

Claims

1. A PCB board detection method based on visual neural network algorithm, It is characterized in that The method comprises: Acquire a plurality of different PCB board unit images, wherein the edge of the PCB board unit has a plurality of training identification points, one of the plurality of training identification points is connected with one or more other training identification points to form a training path, and the PCB board unit image includes a plurality of training identification point images and a plurality of training path images; Acquire a plurality of first position data of a plurality of the training recognition point images, a second position data of the training path image, and a third position data representing an edge contour of the PCB board unit based on the PCB board unit image; wherein, based on the PCB board unit image, the training visual path data is acquired, wherein the training visual path is a conductive path that can be directly acquired on the PCB board unit, and the training visual path data is position data of the conductive path; Using the training visual path data as the second position data; Constructing an initial neural network training model and performing iterative calculations, stopping training when the initial neural network model meets a preset condition and obtaining a target neural network model, wherein the target neural network training model is used to generate the second position data according to the first position data and the third position data; Acquire a detection PCB board unit image, acquire multiple identification points based on the detection PCB board unit image, acquire multiple fourth position data of the multiple identification point images, and fifth position data representing the edge contour of the detection PCB board unit, import the fourth position data and the fifth position data into the target neural network model, and obtain sixth position data; A target path image is acquired based on the sixth position data, and a target point is determined based on the target path image, wherein the target point is a point on the detection PCB board unit, a conduction detection is performed on the target point, and a detection result is output.

2. A PCB board detection method based on a visual neural network algorithm according to claim 1, It is characterized in that The constructing of an initial neural network training model and performing iterative calculations, stopping training when the initial neural network model meets a preset condition and obtaining a target neural network model, wherein the target neural network training model is used to generate the second position data according to the first position data and the third position data, comprises: Acquire predicted second position data based on the initial neural network model; Comparing the predicted second position data with the second position data, and obtaining a correction value; When the correction value is less than a preset threshold, the training is stopped, and the initial neural network training model at this time is used as the target neural network training model.

3. A PCB board detection method based on a visual neural network algorithm according to claim 2, It is characterized in that Comparing the predicted second position data with the second position data and obtaining a correction value includes: The correction value is obtained, and the relationship between the correction value and the predicted second position data and the second position data satisfies: t=|(|y1-yt|)y1| Among them, the correction value is t, the predicted second position data is yt, and the second position data is y1.

4. A PCB board detection method based on a visual neural network algorithm according to claim 3, It is characterized in that Acquire a detection PCB board unit image, acquire multiple identification points based on the detection PCB board unit image, acquire multiple fourth position data of the multiple identification point images, and fifth position data representing the edge contour of the detection PCB board unit, import the fourth position data and the fifth position data into the target neural network model, and obtain sixth position data, including: Acquire detection visual path data based on the PCB board unit image, wherein the detection visual path is a conductive path that can be directly acquired on the detection PCB board unit, and the detection visual path data is position data of the conductive path; Importing the detected visual path data into the target neural network model as the sixth position data, wherein a feature vector of the sixth position data is composed of the fourth position data and the fifth position data; Acquire all the sixth position data based on the target neural network model; use the detected visual path data as the sixth position data; An initial neural network training model is constructed and iterative calculation is performed, and when the initial neural network model meets a preset condition, the training is stopped and a target neural network model is obtained, wherein the target neural network training model is used to generate the second position data according to the first position data and the third position data, and further includes: Taking the first position data and the third position data as different dimensions, generating a feature vector of the second position data; The initial neural network training model is generated according to the feature vector of the second position data.

5. A PCB board detection method based on a visual neural network algorithm according to claim 4, It is characterized in that Before taking the first position data and the third position data as different dimensions to generate a feature vector of the second position data, the method further includes: Normalizing each dimension so that each dimension is within the range of [0, 1]; Performing noise reduction processing on each of the dimensions; Data cleaning is performed for each of the dimensions.

6. A PCB board detection method based on a visual neural network algorithm according to claim 5, It is characterized in that Acquiring a target path image based on the sixth position data, determining a target point based on the target path image, wherein the target point is a point on the detection PCB board unit, performing a conduction detection on the target point, and outputting a detection result, including: generating a mapping path image based on the sixth position data, and determining a target path image based on the mapping path image; Determine a target path based on the target path image, where the target path is a conductive path on the detection PCB board unit; A target point is determined based on the target path, and the target point is a conductive pin point on the detection PCB board.

7. A PCB board detection system based on visual neural network algorithm, It is characterized in that The system is configured to: Acquire a plurality of different PCB board unit images, wherein the edge of the PCB board unit has a plurality of training identification points, one of the plurality of training identification points is connected with one or more other training identification points to form a training path, and the PCB board unit image includes a plurality of training identification point images and a plurality of training path images; Acquire a plurality of first position data of a plurality of the training recognition point images, a second position data of the training path image, and a third position data representing an edge contour of the PCB board unit based on the PCB board unit image; wherein, based on the PCB board unit image, the training visual path data is acquired, wherein the training visual path is a conductive path that can be directly acquired on the PCB board unit, and the training visual path data is position data of the conductive path; Using the training visual path data as the second position data; Constructing an initial neural network training model and performing iterative calculations, stopping training when the initial neural network model meets a preset condition and obtaining a target neural network model, wherein the target neural network training model is used to generate the second position data according to the first position data and the third position data; Acquire a detection PCB board unit image, acquire multiple identification points based on the detection PCB board unit image, acquire multiple fourth position data of the multiple identification point images, and fifth position data representing the edge contour of the detection PCB board unit, import the fourth position data and the fifth position data into the target neural network model, and obtain sixth position data; A target path image is acquired based on the sixth position data, and a target point is determined based on the target path image, wherein the target point is a point on the detection PCB board unit, a conduction detection is performed on the target point, and a detection result is output.

Citation Information

Patent Citations

  • Solder joint defect method for electronic components

    CN107992898A

  • Circuit board defect detection model training method and device and detection method and device

    CN108961236A

  • Image recognition method based on single hidden layer neural network and related equipment

    CN113139466A

  • Circuit board electronic element target detection method based on simplified YOLOv3 network.

    CN113159064A

  • Welding spot detection method based on convolutional neural network

    CN113409250A