CN114897797A - Method, device and equipment for detecting defects of printed circuit board and storage medium - Google Patents

Abstract

The invention belongs to the technical field of detection, and discloses a method, a device, equipment and a storage medium for detecting defects of a printed circuit board. The method comprises the following steps: acquiring a sample training image of a sample printed circuit board and an image to be detected of a printed circuit board to be detected; performing feature calculation according to the sample training image to obtain sample training features corresponding to the sample training image; performing model training according to the sample training characteristics and the sample training images to obtain a preset defect detection model; and carrying out defect detection on the image to be detected according to a preset defect detection model to obtain a defect detection result of the printed circuit board. By the method, the preset defect detection model capable of accurately identifying whether the printed circuit board has defects in the corresponding image is obtained, so that the image to be detected is subjected to defect detection based on the preset defect detection model, the defect detection precision and accuracy during PCB online detection are improved, and the requirement of PCB online defect detection is met.

Description

Defect detection method, device, equipment and storage medium of printed circuit board

technical field

The present invention relates to the technical field of detection, and in particular, to a method, device, equipment and storage medium for defect detection of printed circuit boards.

Background technique

In the automatic production process of vehicle screen, PCB (Printed Circuit Board, printed circuit board) is an important part, but in the process of PCB production, defects such as missing, offset and dislocation of PCB components often occur. Traditional manual inspection cannot meet the requirements of online inspection, so machine vision inspection of PCB defects has become the main inspection method at present, but the inspection accuracy in the existing technology is low.

SUMMARY OF THE INVENTION

The main purpose of the present invention is to provide a method, device, equipment and storage medium for defect detection of a printed circuit board, aiming to solve the technical problem of low detection accuracy of PCB defects in the prior art.

In order to achieve the above object, the present invention provides a method for detecting defects of printed circuit boards, the method for detecting defects of printed circuit boards includes:

Obtain the sample training image of the sample printed circuit board and the to-be-tested image of the to-be-tested printed circuit board;

Perform feature calculation according to the sample training image to obtain sample training features corresponding to the sample training image;

Perform model training according to the sample training features and the sample training images to obtain a preset defect detection model;

Perform defect detection on the to-be-detected image according to the preset defect detection model to obtain a defect detection result of the printed circuit board.

Optionally, the sample training features include target size profile features, image variance features, directional gradient histogram features, edge density features, average gradient strength features, accumulated gradient value features, Fourier spectrum features, and correlation surface features. at least one;

The feature calculation is performed according to the sample training image to obtain the sample training feature corresponding to the sample training image, including:

Perform target size contour feature calculation according to the sample training image to obtain target size contour features;

Perform image variance feature calculation according to the sample training image to obtain image variance feature;

Cell division is performed according to the sample training image to obtain division cells;

Perform direction histogram feature calculation according to the dividing unit to obtain the direction histogram feature;

Determine edge pixel values, row image elements and column image elements according to the sample training image;

Perform edge density feature calculation according to the edge pixel value, row image element and column image element to obtain edge density feature;

Calculate the average gradient intensity according to the sample training image to obtain the average gradient intensity feature;

Determine the gradient of each pixel coordinate direction in the sample training image according to the sample training image;

Calculate the cumulative gradient value according to the gradient in the coordinate direction of each pixel point to obtain the cumulative gradient value feature;

Determine the image size according to the sample training image;

Perform Fourier spectral feature calculation according to the image size to obtain Fourier spectral features;

The relevant surface feature is calculated according to the sample training image to obtain the relevant surface feature.

Optionally, performing target size contour feature calculation according to the sample training image to obtain target size contour features, including:

Determine the gradient of each pixel coordinate direction in the sample training image according to the sample training image;

Determine the gradient angle of each pixel point according to the gradient of the coordinate direction of each pixel point;

Determine the gradient direction entropy according to the gradient angle and the preset angle interval;

The target size contour feature is determined according to the gradient direction entropy and the corresponding quantity of the preset angle interval.

Optionally, the image variance feature calculation is performed according to the sample training image to obtain the image variance feature, including:

Determine the gray value, row image element and column image element corresponding to each pixel point according to the sample training image;

Determine the average grayscale value of the image according to the grayscale value corresponding to each pixel point;

The image variance feature is determined according to the gray value corresponding to each pixel point, the line image element, the column image element and the average value of the image gray level.

Optionally, the average gradient intensity calculation is performed according to the sample training image to obtain the average gradient intensity feature, including:

Determine the gradient of each pixel coordinate direction in the sample training image according to the sample training image;

Determine the gradient intensity value of each pixel point according to the gradient in the coordinate direction of each pixel point;

Determine the gradient intensity map of the sample training image according to the gradient intensity value;

Perform an average filtering operation according to the gradient intensity map to obtain an average gradient intensity map;

An average gradient intensity feature is determined from the average gradient intensity map.

Optionally, calculating the relevant surface features according to the sample training image to obtain the relevant surface features, including:

Cell division is performed according to the sample training image to obtain division cells;

Perform gradient intensity calculation according to the division unit image corresponding to the division unit and the sample training image, to obtain a first gradient intensity corresponding to the sample training image and a second gradient intensity corresponding to the division unit image;

The correlation coefficient is calculated according to the first gradient strength, the second gradient strength, the sample training image and the divided unit image, so as to obtain the correlation surface feature.

Optionally, performing model training according to the sample training features and the sample training images to obtain a preset defect detection model, including:

Obtain preset evaluation indicators;

Perform model training according to the sample training features and the sample training images, to obtain the initial training model and the confusion matrix index of the initial training model;

Perform model evaluation according to the preset evaluation index and the confusion matrix index to obtain an evaluation result;

When the evaluation result is a preset result, a preset defect detection model is obtained according to the initial training model.

In addition, in order to achieve the above object, the present invention also provides a defect detection device for a printed circuit board, and the defect detection device for a printed circuit board includes:

an acquisition module for acquiring the sample training image of the sample printed circuit board and the to-be-detected image of the to-be-detected printed circuit board;

a computing module, configured to perform feature calculation according to the sample training image to obtain sample training features corresponding to the sample training image;

a training module, configured to perform model training according to the sample training features and the sample training images to obtain a preset defect detection model;

The detection module is configured to perform defect detection on the to-be-detected image according to the preset defect detection model to obtain a defect detection result of the printed circuit board.

In addition, in order to achieve the above object, the present invention also provides a defect detection device for a printed circuit board. The defect detection device for a printed circuit board includes: a memory, a processor, and a device stored in the memory and available in the processor. A printed circuit board defect detection program running on the printed circuit board defect detection program configured to implement the printed circuit board defect detection method as described above.

In addition, in order to achieve the above object, the present invention also provides a storage medium, on which a defect detection program of a printed circuit board is stored, and the defect detection program of the printed circuit board is executed by a processor to achieve the above-mentioned A method for defect detection of printed circuit boards.

The present invention obtains the sample training image of the sample printed circuit board and the to-be-detected image of the printed circuit board to be detected; performs feature calculation according to the sample training image to obtain the sample training feature corresponding to the sample training image; Perform model training on the features and the sample training image to obtain a preset defect detection model; perform defect detection on the to-be-detected image according to the preset defect detection model to obtain a defect detection result of the printed circuit board. Through the above method, model training is performed based on the sample training features corresponding to the sample training images of the sample printed circuit board, and a preset defect detection model that can accurately identify whether there is a defect in the corresponding image of the printed circuit board is obtained. Detecting images for defect detection improves the accuracy and accuracy of defect detection during PCB online inspection, and meets the needs of PCB online defect detection.

Description of drawings

1 is a schematic structural diagram of a defect detection device for a printed circuit board in a hardware operating environment according to an embodiment of the present invention;

FIG. 2 is a schematic flowchart of a first embodiment of a method for detecting defects in a printed circuit board according to the present invention;

3 is a schematic flowchart of a second embodiment of a method for detecting defects in a printed circuit board according to the present invention;

FIG. 4 is a structural block diagram of a first embodiment of a defect detection device for a printed circuit board according to the present invention.

The realization, functional characteristics and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments.

Detailed ways

It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

Referring to FIG. 1 , FIG. 1 is a schematic structural diagram of a defect detection device for a printed circuit board in a hardware operating environment according to an embodiment of the present invention.

As shown in FIG. 1 , the defect detection device of the printed circuit board may include: a processor 1001 , such as a central processing unit (Central Processing Unit, CPU), a communication bus 1002 , a user interface 1003 , a network interface 1004 , and a memory 1005 . Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. Optionally, the network interface 1004 may include a standard wired interface and a wireless interface (eg, a wireless-fidelity (Wireless-Fidelity, Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM) memory, or may be a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk memory. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001 .

Those skilled in the art can understand that the structure shown in FIG. 1 does not constitute a limitation on the defect detection equipment of the printed circuit board, and may include more or less components than the one shown, or combine some components, or different Component placement.

As shown in FIG. 1 , the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and a defect detection program for a printed circuit board.

In the printed circuit board defect detection device shown in FIG. 1, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; in the printed circuit board defect detection device of the present invention The processor 1001 and the memory 1005 can be set in the defect detection device of the printed circuit board, and the defect detection device of the printed circuit board calls the defect detection program of the printed circuit board stored in the memory 1005 through the processor 1001, and executes the present invention The embodiment provides a method for detecting defects of a printed circuit board.

An embodiment of the present invention provides a method for detecting defects in a printed circuit board. Referring to FIG. 2 , FIG. 2 is a schematic flowchart of a first embodiment of a method for detecting defects in a printed circuit board according to the present invention.

The defect detection method of printed circuit board includes the following steps:

Step S10: Obtain a sample training image of a sample printed circuit board and an image to be detected of the printed circuit board to be detected.

It should be noted that the execution body of this embodiment is a terminal device, and a printed circuit board defect detection system is installed on the terminal device. When the terminal device receives a detection instruction, it forwards the detection instruction to the printed circuit board defect detection system. The defect detection system of the printed circuit board obtains the sample training image and the image to be inspected according to the inspection instruction. Perform feature calculation according to the sample training images to obtain sample training features corresponding to each sample training image, and perform model training according to the sample training features and the sample training images to obtain a preset defect detection model. Based on the obtained preset defect detection model, the image to be inspected is inspected for defects, and the defect inspection result of the printed circuit board is obtained.

It can be understood that the sample training images of a large number of sample printed circuit boards are obtained from the preset database, and the sample training images may be defective sample training images or non-defective sample training images. The sample training images corresponding to the sample printed circuit boards are manually discriminated. The label corresponding to the defective sample training image is set to 1, and the label corresponding to the non-defective sample training image is set to 0. Finally, all the sample printed circuit boards are set to The sample training images and corresponding labels are stored to obtain the sample training image dataset.

In a specific implementation, the to-be-detected image is an image corresponding to the to-be-detected printed circuit board.

Step S20: Perform feature calculation according to the sample training images to obtain sample training features corresponding to the sample training images.

It should be noted that the sample training feature refers to the image feature value obtained after feature calculation based on the sample training image, and the sample training feature includes the target size contour feature, image variance feature, directional gradient histogram feature, edge density feature, and average gradient. At least one of an intensity feature, an accumulated gradient value feature, a Fourier spectrum feature, and a correlation surface feature.

Step S30: Perform model training according to the sample training features and the sample training images to obtain a preset defect detection model.

It should be noted that the sample training features and corresponding sample training images are split according to a preset ratio to obtain a test set and a training set. GBDT (Gradient Boosting Decision Tree, gradient descent decision tree) is used to train and fit the model according to the training set, and the initial training model after model training is tested according to the test set. Preset defect detection models for defects.

It can be understood that, in order to ensure the accuracy of model training and improve the accuracy of subsequent defect detection, further, the model training is performed according to the sample training features and the sample training images to obtain a preset defect detection model, Including: obtaining a preset evaluation index; performing model training according to the sample training feature and the sample training image to obtain an initial training model and a confusion matrix index of the initial training model; according to the preset evaluation index and the confusion The matrix index is used for model evaluation to obtain an evaluation result; when the evaluation result is a preset result, a preset defect detection model is obtained according to the initial training model.

In a specific implementation, the preset evaluation index refers to a preset index by which the model can stop training. According to the sample training features and sample training images, the GBDT classification algorithm is used for model training, and the initial training model and the confusion matrix indicators of the initial training model are obtained. Perform model evaluation on the preset evaluation index and the confusion matrix index, and obtain the evaluation result of whether the confusion matrix index meets the requirements. When the evaluation result is the preset result that the confusion matrix index meets the requirements, the initial training model can be used as the preset defect detection model. When the evaluation result is that the confusion matrix index does not meet the requirements, it is necessary to adjust the training parameters during model training according to the preset evaluation index, such as step size, learning rate, number of iterations, and the number of nodes for model classification, and then re-train the model.

Step S40: Perform defect detection on the to-be-detected image according to the preset defect detection model to obtain a defect detection result of the printed circuit board.

It should be noted that, after obtaining the preset defect detection model, input the image to be detected into the preset defect detection model to perform defect detection on the image to be detected, and obtain whether the printed circuit board has defects such as missing components, offsets and dislocations. result.

In this embodiment, a sample training image of a sample printed circuit board and a to-be-detected image of a printed circuit board to be detected are obtained; feature calculation is performed according to the sample training image to obtain sample training features corresponding to the sample training image; Perform model training on the training feature and the sample training image to obtain a preset defect detection model; perform defect detection on the to-be-detected image according to the preset defect detection model to obtain a defect detection result of the printed circuit board. Through the above method, model training is performed based on the sample training features corresponding to the sample training images of the sample printed circuit board, and a preset defect detection model that can accurately identify whether there is a defect in the corresponding image of the printed circuit board is obtained. Detecting images for defect detection improves the accuracy and accuracy of defect detection during PCB online inspection, and meets the needs of PCB online defect detection.

Referring to FIG. 3 , FIG. 3 is a schematic flowchart of a second embodiment of a method for detecting defects of a printed circuit board according to the present invention.

Based on the above-mentioned first embodiment, the sample training features include target size profile features, image variance features, directional gradient histogram features, edge density features, average gradient strength features, accumulated gradient value features, Fourier spectral features, and correlation surfaces At least one of the features, the step S20 in the method for detecting defects of a printed circuit board of this embodiment, includes:

Step S21: Perform target size contour feature calculation according to the sample training image to obtain target size contour features.

It should be noted that the target size contour feature is the large-scale contour feature H _nor corresponding to each sample training image.

It can be understood that, in order to obtain the accurate target size contour feature H _nor , further, performing the target size contour feature calculation according to the sample training image to obtain the target size contour feature includes: determining the target size contour feature according to the sample training image. the gradient of each pixel coordinate direction in the sample training image; the gradient angle of each pixel point is determined according to the gradient of the coordinate direction of each pixel point; the gradient direction entropy is determined according to the gradient angle and the preset angle interval; The gradient direction entropy and the corresponding quantity of the preset angle interval determine the contour feature of the target size.

In a specific implementation, the gradient in the coordinate direction of each pixel point refers to the gradient G _x (x, y) in the x direction of each pixel point in the sample training image and the gradient G _y (x, y) in the y direction. The gradient angle of each pixel refers to the gradient angle of each pixel

在得到梯度方向熵

后，根据熵的定义可知，当每个方向的梯度出现概率都相等的情况下，熵取最大值，得到归一化梯度方向熵

其中，

对归一化后的熵值进行统计，对H中像素点进行由大到小排序，将排序结果中占像素数的前30％的点进行加权，即行程最终目标区域的值

w和h分别是目标区域的宽和高，H_nor即为目标尺寸轮廓特征It should be noted that the preset angle interval refers to n angle intervals obtained by dividing the gradient direction of 0 to 360 degrees into n intervals on average. n is the corresponding number of preset angle intervals, divide the sample pixel image into cells to obtain a plurality of division units, the size is about w*h=10*10, calculate the gradient of each division unit in each sample pixel image The pixel probability p _i of the angle within each preset angle interval. Calculate the probability of each direction in the gradient direction histogram of the division unit separately, so as to obtain the gradient direction entropy of each division unit

In getting the gradient direction entropy

Then, according to the definition of entropy, when the probability of the gradient in each direction is equal, the entropy takes the maximum value, and the normalized gradient direction entropy is obtained.

in,

Count the normalized entropy values, sort the pixels in H from large to small, and weight the top 30% of the pixels in the sorting result, that is, the value of the final destination area of the trip

w and h are the width and height of the target area, respectively, and H _nor is the target size contour feature

Step S22: Perform image variance feature calculation according to the sample training image to obtain image variance feature.

It should be noted that the image variance feature refers to the grayscale variance feature of the image, which reflects the fluctuation degree of the overall grayscale of the image. The image variance feature is calculated according to the sample training image, and the image variance feature V _ar is obtained.

It can be understood that, in order to obtain the accurate image variance feature, further, performing the image variance feature calculation according to the sample training image to obtain the image variance feature includes: determining the grayscale corresponding to each pixel point according to the sample training image. degree value, line image element and column image element; determine the average gray level of the image according to the gray value corresponding to each pixel point; according to the gray value, line image element, column image element and image corresponding to each pixel point The gray mean value determines the image variance characteristics.

In the specific implementation, the gray value X(i,j) corresponding to each pixel is determined according to the sample training image, where (i,j) refers to the coordinates of the pixel, the row image element m and the column image element of the sample training image n. According to the gray level corresponding to each pixel point, the average value E(X(i,j)) of the sample training image gray level can be determined. value to determine image variance characteristics

Step S23: Perform cell division according to the sample training image to obtain division cells.

It should be noted that the sample training images are divided into cells to obtain divided cells of the same size.

Step S24: Perform direction histogram feature calculation according to the dividing unit to obtain the direction histogram feature.

It should be noted that the sum of the entropy values of all division units is calculated to obtain the Hog entropy of the direction histogram feature.

Step S25: Determine edge pixel values, line image elements and column image elements according to the sample training image.

It should be noted that the edge pixel value is obtained through the edge operator according to the sample training image, and the row image element m and the column image element n of the sample training image are determined.

Step S26: Perform edge density feature calculation according to the edge pixel values, line image elements and column image elements to obtain edge density features.

其中，edge是边缘像素值的绝对值。It should be noted that the edge density feature is used to determine whether the edge feature distribution in the sample training image is dense, and is an index reflecting the amount of image information. The edge density feature can be obtained by calculating the edge density feature according to the edge pixel value, line image element and column image element.

where edge is the absolute value of the edge pixel value.

Step S27: Calculate the average gradient intensity according to the sample training image to obtain the average gradient intensity feature.

It should be noted that the average gradient intensity can be calculated according to the sample training image to obtain the average gradient intensity feature.

It can be understood that, in order to obtain an accurate average gradient strength, further, performing the average gradient strength calculation according to the sample training image to obtain the average gradient strength feature includes: determining the sample training image according to the sample training image. According to the gradient of the coordinate direction of each pixel point, the gradient intensity value of each pixel point is determined; the gradient intensity map of the sample training image is determined according to the gradient intensity value; according to the gradient intensity A mean value filtering operation is performed on the graph to obtain an average gradient intensity map; the average gradient intensity feature is determined according to the average gradient intensity map.

从而得到样本训练图像的梯度强度图。并对样本训练图像进行单元格划分，得到划分单元，以划分单元的大小为模板进行均值滤波操作，得到划分单元的平均梯度强度图。根据平均梯度强度图排名，取前30％的梯度强度值的平均值为样本训练图像的平均梯度强度值。In the specific implementation, the gradient in the coordinate direction of each pixel point refers to the gradient G _x (x, y) in the x direction of each pixel point in the sample training image and the gradient G _y (x, y) in the y direction, and then each pixel is obtained. Gradient strength value of point

Thereby, the gradient intensity map of the sample training image is obtained. The sample training image is divided into cells to obtain the divided cells, and the mean value filtering operation is performed with the size of the divided cells as a template to obtain the average gradient intensity map of the divided cells. According to the average gradient intensity map ranking, the average gradient intensity value of the top 30% is taken as the average gradient intensity value of the sample training images.

Step S28: Determine the gradient of each pixel coordinate direction in the sample training image according to the sample training image.

It should be noted that the gradient in the coordinate direction of each pixel point refers to the gradient G _x (x, y) in the x direction of each pixel point in the sample training image and the gradient G _y (x, y) in the y direction.

Step S29: Calculate the cumulative gradient value according to the gradient in the coordinate direction of each pixel point to obtain the cumulative gradient value feature.

最终

It should be noted that the cumulative gradient value feature describes the sum of the gradient strengths in a region, and the stronger the cumulative number of edges in a region, the more obvious the gradient feature is. Calculate the cumulative gradient value according to the gradient in the coordinate direction of each pixel point, and accumulate the gradient value characteristics

finally

Step S210: Determine the image size according to the sample training image.

It should be noted that the original image size I of the sample training image is determined according to the sample training image.

Step S211: Perform Fourier spectral feature calculation according to the image size to obtain Fourier spectral features.

It should be noted that the Fourier spectral feature describes the texture feature of the sample training image. After the image size is obtained, the Fourier transform energy map FI of the sample training image is obtained, FI=|F(I)|, F(I ) is the two-dimensional Fourier transform of I, the size of FI is also m×n, E=sum(FI)=sum[|F(I)|], E is the Fourier energy sum, then

Among them, BW2, BW4, BW8, and BW16 are respectively referred to as 1/2 radius energy sum, 1/4 radius energy sum, 1/8 radius energy sum, and 1/16 radius energy sum.

Step S212: Calculate the relevant surface features according to the sample training image to obtain the relevant surface features.

It should be noted that the correlation surface feature refers to the correlation coefficient between the sample training image and the divided unit image.

It can be understood that, in order to obtain accurate relevant surface features, further, performing the relevant surface feature calculation according to the sample training image to obtain the relevant surface features includes: performing cell division according to the sample training image to obtain a division. unit; carry out gradient intensity calculation according to the division unit image corresponding to the division unit and the sample training image, and obtain the first gradient intensity corresponding to the sample training image and the second gradient intensity corresponding to the division unit image; The first gradient strength, the second gradient strength, the sample training image and the division unit image are used to calculate the correlation coefficient to obtain the correlation surface feature.

第二梯度强度指的是划分单元图像在各像素点的梯度强度

In a specific implementation, the cells are divided according to the sample training images to obtain division units of the same size, and the division unit images corresponding to the division units are obtained. The first gradient strength refers to the gradient strength of the sample training image at each pixel

The second gradient strength refers to the gradient strength of the divided unit image at each pixel point

其中，E(ai)是划分单元图像梯度强度的均值，E(AI)是样本训练图像对应区域梯度强度的均值。(x,y)是划分单元图像图左上角坐标在样本训练图像中。It should be noted that, after obtaining the first gradient strength, the second gradient strength, the sample training image and the divided unit image, the correlation coefficient can be calculated to obtain the correlation surface feature. The divided unit image is traversed on the sample training image with pixels as the cloth length. The size of the correlation surface coefficient S is (W-w+1)×(H-h+1), and each pixel value is the dividing unit image and sample training. The correlation coefficient value of the corresponding area of the image. The local maximum value in the correlation surface is the peak value, the size of the local range is set by the outside world, and the final correlation surface features

Among them, E(ai) is the mean value of the gradient strength of the divided unit image, and E(AI) is the mean value of the gradient strength of the corresponding region of the sample training image. (x,y) is the coordinate of the upper left corner of the divided unit image graph in the sample training image.

In this embodiment, the target size contour feature is obtained by calculating the target size contour feature according to the sample training image; the image variance feature is obtained by calculating the image variance feature according to the sample training image; and the cell division is performed according to the sample training image. , obtain a division unit; perform direction histogram feature calculation according to the division unit, obtain the direction histogram feature; determine edge pixel values, line image elements and column image elements according to the sample training image; Perform edge density feature calculation on image elements and column image elements to obtain edge density features; perform average gradient intensity calculation according to the sample training image to obtain average gradient intensity features; determine each pixel in the sample training image according to the sample training image The gradient of the point coordinate direction; calculate the accumulated gradient value according to the gradient of the coordinate direction of each pixel point, and obtain the accumulated gradient value feature; determine the image size according to the sample training image; calculate the Fourier spectrum feature according to the image size , to obtain the Fourier spectral feature; and to calculate the correlation surface feature according to the sample training image to obtain the correlation surface feature. Thus, accurate sample training features can be obtained, and accurate training data can be provided for subsequent model training.

In addition, referring to FIG. 4 , an embodiment of the present invention further provides a defect detection device for a printed circuit board, and the defect detection device for a printed circuit board includes:

The acquiring module 10 is used for acquiring the sample training image of the sample printed circuit board and the to-be-detected image of the to-be-detected printed circuit board.

The calculation module 20 is configured to perform feature calculation according to the sample training image to obtain the sample training feature corresponding to the sample training image.

The training module 30 is configured to perform model training according to the sample training features and the sample training images to obtain a preset defect detection model.

The detection module 40 is configured to perform defect detection on the to-be-detected image according to the preset defect detection model to obtain a defect detection result of the printed circuit board.

In this embodiment, a sample training image of a sample printed circuit board and a to-be-detected image of a printed circuit board to be detected are obtained; feature calculation is performed according to the sample training image to obtain sample training features corresponding to the sample training image; Perform model training on the training feature and the sample training image to obtain a preset defect detection model; perform defect detection on the to-be-detected image according to the preset defect detection model to obtain a defect detection result of the printed circuit board. Through the above method, model training is performed based on the sample training features corresponding to the sample training images of the sample printed circuit board, and a preset defect detection model that can accurately identify whether there is a defect in the corresponding image of the printed circuit board is obtained. Detecting images for defect detection improves the accuracy and accuracy of defect detection during PCB online inspection, and meets the needs of PCB online defect detection.

In one embodiment, the computing module 20 is further configured to perform target size contour feature calculation according to the sample training image to obtain target size contour features;

Perform image variance feature calculation according to the sample training image to obtain image variance feature;

Cell division is performed according to the sample training image to obtain division cells;

Perform direction histogram feature calculation according to the dividing unit to obtain the direction histogram feature;

Determine edge pixel values, row image elements and column image elements according to the sample training image;

Perform edge density feature calculation according to the edge pixel value, row image element and column image element to obtain edge density feature;

Calculate the average gradient intensity according to the sample training image to obtain the average gradient intensity feature;

Determine the gradient of each pixel coordinate direction in the sample training image according to the sample training image;

Calculate the cumulative gradient value according to the gradient in the coordinate direction of each pixel point to obtain the cumulative gradient value feature;

Determine the image size according to the sample training image;

Perform Fourier spectral feature calculation according to the image size to obtain Fourier spectral features;

The relevant surface feature is calculated according to the sample training image to obtain the relevant surface feature.

In one embodiment, the computing module 20 is further configured to determine the gradient of each pixel coordinate direction in the sample training image according to the sample training image;

Determine the gradient angle of each pixel point according to the gradient of the coordinate direction of each pixel point;

Determine the gradient direction entropy according to the gradient angle and the preset angle interval;

The target size contour feature is determined according to the gradient direction entropy and the corresponding quantity of the preset angle interval.

In one embodiment, the computing module 20 is further configured to determine the gray value, the row image element and the column image element corresponding to each pixel point according to the sample training image;

Determine the average grayscale value of the image according to the grayscale value corresponding to each pixel point;

The image variance feature is determined according to the gray value corresponding to each pixel point, the line image element, the column image element and the average value of the image gray level.

In one embodiment, the computing module 20 is further configured to determine the gradient of each pixel coordinate direction in the sample training image according to the sample training image;

Determine the gradient intensity value of each pixel point according to the gradient in the coordinate direction of each pixel point;

Determine the gradient intensity map of the sample training image according to the gradient intensity value;

Perform an average filtering operation according to the gradient intensity map to obtain an average gradient intensity map;

An average gradient intensity feature is determined from the average gradient intensity map.

In one embodiment, the computing module 20 is further configured to perform cell division according to the sample training image to obtain division cells;

Perform gradient intensity calculation according to the division unit image corresponding to the division unit and the sample training image, to obtain a first gradient intensity corresponding to the sample training image and a second gradient intensity corresponding to the division unit image;

The correlation coefficient is calculated according to the first gradient strength, the second gradient strength, the sample training image and the divided unit image, so as to obtain the correlation surface feature.

In one embodiment, the training module 30 is further configured to obtain a preset evaluation index;

Perform model training according to the sample training features and the sample training images, to obtain the initial training model and the confusion matrix index of the initial training model;

Perform model evaluation according to the preset evaluation index and the confusion matrix index to obtain an evaluation result;

When the evaluation result is a preset result, a preset defect detection model is obtained according to the initial training model.

Since the device adopts all the technical solutions of all the above-mentioned embodiments, it has at least all the beneficial effects brought by the technical solutions of the above-mentioned embodiments, which will not be repeated here.

In addition, an embodiment of the present invention also provides a storage medium, where a defect detection program of a printed circuit board is stored on the storage medium, and the printed circuit board as described above is implemented when the defect detection program of the printed circuit board is executed by a processor. Steps of a method for defect detection of a board.

Since the storage medium adopts all the technical solutions of all the above-mentioned embodiments, it has at least all the beneficial effects brought by the technical solutions of the above-mentioned embodiments, which will not be repeated here.

It should be noted that the above-described workflow is only illustrative, and does not limit the protection scope of the present invention. In practical applications, those skilled in the art can select some or all of them to implement according to actual needs. The purpose of the solution in this embodiment is not limited here.

In addition, for technical details that are not described in detail in this embodiment, reference may be made to the defect detection method for a printed circuit board provided by any embodiment of the present invention, and details are not repeated here.

Furthermore, it should be noted that, herein, the terms "comprising", "comprising" or any other variation thereof are intended to encompass non-exclusive inclusion, such that a process, method, article or system comprising a series of elements includes not only those elements, but also other elements not expressly listed or inherent to such a process, method, article or system. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, method, article or system that includes the element.

The above-mentioned serial numbers of the embodiments of the present invention are only for description, and do not represent the advantages or disadvantages of the embodiments.

From the description of the above embodiments, those skilled in the art can clearly understand that the method of the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is better implementation. Based on such understanding, the technical solutions of the present invention can be embodied in the form of software products in essence or the parts that make contributions to the prior art, and the computer software products are stored in a storage medium (such as a read-only memory). , ROM)/RAM, magnetic disk, optical disk), including several instructions to make a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) execute the methods described in the various embodiments of the present invention.

The above are only preferred embodiments of the present invention, and are not intended to limit the scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the description and drawings of the present invention, or directly or indirectly applied in other related technical fields , are similarly included in the scope of patent protection of the present invention.

Claims (10)

1. A defect detection method for a printed circuit board, wherein the defect detection method for the printed circuit board comprises: Obtain the sample training image of the sample printed circuit board and the to-be-tested image of the to-be-tested printed circuit board; Perform feature calculation according to the sample training image to obtain sample training features corresponding to the sample training image; Perform model training according to the sample training features and the sample training images to obtain a preset defect detection model; Perform defect detection on the to-be-detected image according to the preset defect detection model to obtain a defect detection result of the printed circuit board. 2. The defect detection method of a printed circuit board according to claim 1, wherein the sample training features include target size contour features, image variance features, directional gradient histogram features, edge density features, and average gradient strength features , at least one of cumulative gradient value features, Fourier spectral features, and correlation surface features; The feature calculation is performed according to the sample training image to obtain the sample training feature corresponding to the sample training image, including: Perform target size contour feature calculation according to the sample training image to obtain target size contour features; Perform image variance feature calculation according to the sample training image to obtain image variance feature; Cell division is performed according to the sample training image to obtain division cells; Perform direction histogram feature calculation according to the dividing unit to obtain the direction histogram feature; Determine edge pixel values, row image elements and column image elements according to the sample training image; Perform edge density feature calculation according to the edge pixel value, row image element and column image element to obtain edge density feature; Calculate the average gradient intensity according to the sample training image to obtain the average gradient intensity feature; Determine the gradient of each pixel coordinate direction in the sample training image according to the sample training image; Calculate the cumulative gradient value according to the gradient in the coordinate direction of each pixel point to obtain the cumulative gradient value feature; Determine the image size according to the sample training image; Perform Fourier spectral feature calculation according to the image size to obtain Fourier spectral features; The relevant surface feature is calculated according to the sample training image to obtain the relevant surface feature. 3 . The defect detection method of a printed circuit board according to claim 2 , wherein the calculation of the target size contour feature according to the sample training image to obtain the target size contour feature comprises: 3 . Determine the gradient of each pixel coordinate direction in the sample training image according to the sample training image; Determine the gradient angle of each pixel point according to the gradient of the coordinate direction of each pixel point; Determine the gradient direction entropy according to the gradient angle and the preset angle interval; The target size contour feature is determined according to the gradient direction entropy and the corresponding quantity of the preset angle interval. 4. The method for detecting defects of a printed circuit board according to claim 2, wherein the calculation of the image variance feature according to the sample training image to obtain the image variance feature comprises: Determine the gray value, row image element and column image element corresponding to each pixel point according to the sample training image; Determine the average grayscale value of the image according to the grayscale value corresponding to each pixel point; The image variance feature is determined according to the gray value corresponding to each pixel point, the line image element, the column image element and the average value of the image gray level. 5 . The defect detection method of a printed circuit board according to claim 2 , wherein the average gradient intensity calculation is performed according to the sample training image to obtain the average gradient intensity feature, comprising: 5 . Determine the gradient of each pixel coordinate direction in the sample training image according to the sample training image; Determine the gradient intensity value of each pixel point according to the gradient in the coordinate direction of each pixel point; Determine the gradient intensity map of the sample training image according to the gradient intensity value; Perform an average filtering operation according to the gradient intensity map to obtain an average gradient intensity map; An average gradient intensity feature is determined from the average gradient intensity map. 6 . The defect detection method of a printed circuit board according to claim 2 , wherein the calculation of the relevant surface features according to the sample training image to obtain the relevant surface features comprises: 6 . Cell division is performed according to the sample training image to obtain division cells; Perform gradient intensity calculation according to the division unit image corresponding to the division unit and the sample training image, to obtain a first gradient intensity corresponding to the sample training image and a second gradient intensity corresponding to the division unit image; The correlation coefficient is calculated according to the first gradient strength, the second gradient strength, the sample training image and the divided unit image, so as to obtain the correlation surface feature. 7 . The defect detection method of a printed circuit board according to claim 1 , wherein the performing model training according to the sample training features and the sample training images to obtain a preset defect detection model, comprising: 8 . Obtain preset evaluation indicators; Perform model training according to the sample training features and the sample training images, to obtain the initial training model and the confusion matrix index of the initial training model; Perform model evaluation according to the preset evaluation index and the confusion matrix index to obtain an evaluation result; When the evaluation result is a preset result, a preset defect detection model is obtained according to the initial training model. 8. A defect detection device for a printed circuit board, wherein the defect detection device for the printed circuit board comprises: an acquisition module for acquiring the sample training image of the sample printed circuit board and the to-be-detected image of the to-be-detected printed circuit board; a computing module, configured to perform feature calculation according to the sample training image to obtain sample training features corresponding to the sample training image; a training module, configured to perform model training according to the sample training features and the sample training images to obtain a preset defect detection model; The detection module is configured to perform defect detection on the to-be-detected image according to the preset defect detection model to obtain a defect detection result of the printed circuit board. 9. A defect detection device for a printed circuit board, characterized in that the device comprises: a memory, a processor, and a defect detection program for a printed circuit board stored on the memory and executable on the processor, The defect detection program of the printed circuit board is configured to implement the defect detection method of the printed circuit board according to any one of claims 1 to 7. 10. A storage medium, wherein a defect detection program for a printed circuit board is stored on the storage medium, and when the printed circuit board defect detection program is executed by a processor, any one of claims 1 to 7 is implemented The described method for detecting defects of printed circuit boards.

Images