CN117750643B - Surface processing method of printed circuit board - Google Patents
Surface processing method of printed circuit board Download PDFInfo
- Publication number
- CN117750643B CN117750643B CN202410183031.9A CN202410183031A CN117750643B CN 117750643 B CN117750643 B CN 117750643B CN 202410183031 A CN202410183031 A CN 202410183031A CN 117750643 B CN117750643 B CN 117750643B
- Authority
- CN
- China
- Prior art keywords
- surface pattern
- printed circuit
- circuit board
- training
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 29
- 230000007547 defect Effects 0.000 claims abstract description 40
- 238000012545 processing Methods 0.000 claims abstract description 31
- 229920002120 photoresistant polymer Polymers 0.000 claims abstract description 18
- 238000007689 inspection Methods 0.000 claims abstract description 17
- 238000001259 photo etching Methods 0.000 claims abstract description 12
- 239000011248 coating agent Substances 0.000 claims abstract description 8
- 238000000576 coating method Methods 0.000 claims abstract description 8
- 239000004020 conductor Substances 0.000 claims abstract description 8
- 238000005530 etching Methods 0.000 claims abstract description 8
- 238000011161 development Methods 0.000 claims abstract description 5
- 238000012549 training Methods 0.000 claims description 96
- 230000008447 perception Effects 0.000 claims description 59
- 238000010586 diagram Methods 0.000 claims description 33
- 238000003062 neural network model Methods 0.000 claims description 32
- 239000013598 vector Substances 0.000 claims description 29
- 238000005728 strengthening Methods 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 17
- 238000013527 convolutional neural network Methods 0.000 claims description 15
- 238000000034 method Methods 0.000 claims description 15
- 230000002950 deficient Effects 0.000 claims description 8
- 230000000903 blocking effect Effects 0.000 claims description 7
- 239000011159 matrix material Substances 0.000 claims description 7
- 238000004519 manufacturing process Methods 0.000 abstract description 14
- 230000005518 electrochemistry Effects 0.000 abstract description 2
- 238000001514 detection method Methods 0.000 description 18
- 238000011176 pooling Methods 0.000 description 11
- 238000009826 distribution Methods 0.000 description 7
- 238000000605 extraction Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 239000000306 component Substances 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 230000002787 reinforcement Effects 0.000 description 3
- 239000000126 substance Substances 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 238000007477 logistic regression Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000011179 visual inspection Methods 0.000 description 2
- 208000003464 asthenopia Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 239000008358 core component Substances 0.000 description 1
- 238000011496 digital image analysis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000003754 machining Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 238000005476 soldering Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Landscapes
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
A surface processing method of a printed circuit board is disclosed. Firstly, coating a conductive material on the surface of a printed circuit board to form a conductive layer, then, covering photoresist on the conductive layer, etching the photoresist into a required pattern by means of exposure and development to form a photoetching layer, then, removing the part of the conductive layer which is not covered by the photoetching layer by means of chemistry or electrochemistry to form a surface pattern of the printed circuit board, then, performing quality inspection on the surface pattern of the printed circuit board to judge whether the surface pattern has defects, and finally, removing the photoetching layer to obtain a surface processing finished product of the printed circuit board. Therefore, the automation degree of the manufacturing process of the printed circuit board and the stability of the product quality can be improved, and the production efficiency and the product quality are optimized.
Description
Technical Field
The present application relates to the field of printed circuit boards, and more particularly, to a surface processing method of a printed circuit board.
Background
A printed circuit board (Printed Circuit Board, PCB) is a common base component in electronic devices for connecting and supporting electronic components. In the manufacturing process, the surface processing of the PCB is a key step, which involves forming a conductive layer and a patterned circuit structure on the surface of the PCB, and the quality of the surface pattern of the printed circuit board directly affects the performance and reliability of the electronic device, so that it is necessary to perform effective quality inspection on the surface pattern of the printed circuit board.
However, the conventional surface pattern inspection method of the printed circuit board generally relies on manual visual inspection, which requires inspection personnel to inspect the surface pattern of each printed circuit board one by one, which is inefficient and cannot meet the demands of mass production and high-speed production lines. In addition, the manual visual detection is easily affected by factors such as personnel fatigue, subjective judgment, visual fatigue and the like, so that the accuracy of the detection result is limited. Different operators may have different criteria for judgment, resulting in inconsistent results. In addition, conventional quality inspection methods generally detect only a few simple defects, such as scratches, cracks, and the like. For more complex defects, such as minor soldering problems or circuit connection errors, manual visual inspection is often not accurately detected and identified.
Thus, an optimized surface finish scheme for printed circuit boards is desired.
Disclosure of Invention
The present application has been made to solve the above-mentioned technical problems. The embodiment of the application provides a surface processing method of a printed circuit board, which can realize automatic defect detection of a printed circuit board surface pattern so as to reduce the problems caused by labor cost and subjectivity, improve the automation degree of a PCB manufacturing process and the stability of product quality, and optimize the production efficiency and the product quality.
According to an aspect of the present application, there is provided a surface processing method of a printed circuit board, comprising:
coating the surface of the printed circuit board with a conductive material to form a conductive layer;
covering photoresist on the conductive layer, and etching the photoresist into a required pattern by means of exposure and development to form a photoetching layer;
chemically or electrochemically removing the portions of the conductive layer not covered by the photolithographic layer to form a surface pattern of a printed circuit board;
Quality inspection is carried out on the surface pattern of the printed circuit board so as to judge whether the surface pattern has defects or not; and
And removing the photoetching layer to obtain a finished surface processing product of the printed circuit board.
Compared with the prior art, the surface processing method of the printed circuit board comprises the steps of firstly coating a conductive material on the surface of the printed circuit board to form a conductive layer, then covering photoresist on the conductive layer, etching the photoresist into a required pattern by means of exposure and development to form a photoetching layer, then removing the part of the conductive layer which is not covered by the photoetching layer by means of chemistry or electrochemistry to form a surface pattern of the printed circuit board, then performing quality inspection on the surface pattern of the printed circuit board to judge whether the surface pattern has defects, and finally removing the photoetching layer to obtain a surface processing finished product of the printed circuit board. Therefore, the automation degree of the manufacturing process of the printed circuit board and the stability of the product quality can be improved, and the production efficiency and the product quality are optimized.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the description of the embodiments will be briefly introduced below, the following drawings not being drawn to scale with respect to actual dimensions, emphasis instead being placed upon illustrating the gist of the present application.
Fig. 1 is a flowchart of a surface processing method of a printed circuit board according to an embodiment of the present application.
Fig. 2 is a flowchart of substep S140 of the surface processing method of the printed circuit board according to the embodiment of the present application.
Fig. 3 is a schematic diagram of a sub-step S140 of the surface processing method of the printed circuit board according to an embodiment of the application.
Fig. 4 is a block diagram of a surface finishing system for a printed circuit board according to an embodiment of the present application.
Fig. 5 is an application scenario diagram of a surface processing method of a printed circuit board according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some, but not all embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are also within the scope of the application.
As used in the specification and in the claims, the terms "a," "an," "the," and/or "the" are not specific to a singular, but may include a plurality, unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that the steps and elements are explicitly identified, and they do not constitute an exclusive list, as other steps or elements may be included in a method or apparatus.
Although the present application makes various references to certain modules in a system according to embodiments of the present application, any number of different modules may be used and run on a user terminal and/or server. The modules are merely illustrative, and different aspects of the systems and methods may use different modules.
A flowchart is used in the present application to describe the operations performed by a system according to embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in order precisely. Rather, the various steps may be processed in reverse order or simultaneously, as desired. Also, other operations may be added to or removed from these processes.
Hereinafter, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. It should be apparent that the described embodiments are only some embodiments of the present application and not all embodiments of the present application, and it should be understood that the present application is not limited by the example embodiments described herein.
In the technical scheme of the application, a surface processing method of a printed circuit board is provided, and correspondingly, fig. 1 is a flow chart of the surface processing method of the printed circuit board according to an embodiment of the application. As shown in fig. 1, a surface processing method of a printed circuit board according to an embodiment of the present application includes the steps of: s110, coating a conductive material on the surface of the printed circuit board to form a conductive layer; s120, covering photoresist on the conductive layer, and etching the photoresist into a required pattern by means of exposure and development to form a photoresist layer; s130, removing the part of the conductive layer which is not covered by the photoetching layer in a chemical or electrochemical mode to form a surface pattern of the printed circuit board; s140, quality inspection is carried out on the surface pattern of the printed circuit board so as to judge whether the surface pattern has defects or not; and S150, removing the photoetching layer to obtain a finished surface processing product of the printed circuit board.
Accordingly, considering that the surface processing method of the PCB mainly includes steps of conductive material coating, photoresist etching, and chemical or electrochemical etching to form a surface pattern of the printed circuit board, the steps require high-precision operations and strict quality inspection to ensure quality and reliability of the surface pattern of the PCB. Aiming at the defect of surface quality inspection of the printed circuit board by the traditional manual work, the technical concept of the application is that the digital image of the surface pattern of the printed circuit board is acquired by a high-definition camera, and the digital image analysis of the surface pattern is carried out by introducing an image processing and analyzing algorithm at the rear end, so that the defect inspection of the surface pattern of the printed circuit board is automatically carried out, and the detection efficiency and accuracy are improved. Therefore, the automatic defect detection of the surface pattern of the printed circuit board can be realized, so that the problems caused by labor cost and subjectivity are reduced, the automation degree of the PCB manufacturing process and the stability of the product quality are improved, and the production efficiency and the product quality are optimized.
Fig. 2 is a flowchart of substep S140 of the surface processing method of the printed circuit board according to the embodiment of the present application. Fig. 3 is a schematic diagram of a sub-step S140 of the surface processing method of the printed circuit board according to an embodiment of the application. As shown in fig. 2 and 3, according to the surface processing method of a printed circuit board in an embodiment of the present application, quality inspection is performed on a surface pattern of the printed circuit board to determine whether the surface pattern has a defect, including the steps of: s141, acquiring a digital image of the surface pattern of the printed circuit board acquired by a high-definition camera; s142, performing image blocking processing on the digital image of the surface pattern to obtain a sequence of digital image blocks of the surface pattern; s143, respectively extracting the characteristics of each surface pattern digital image block in the sequence of the surface pattern digital image blocks by a surface pattern local characteristic extractor based on a deep neural network model to obtain a sequence of surface pattern local area characteristic diagrams; s144, performing autocorrelation association reinforcement on each surface pattern local area characteristic diagram in the sequence of the surface pattern local area characteristic diagrams to obtain a sequence of reinforced surface pattern local area characteristic diagrams; s145, globally arranging and sensing the sequence of the characteristic map of the local area of the enhanced surface pattern to obtain the characteristic of the extended surface pattern of the sensing domain; and S146, determining whether the surface pattern of the printed circuit board has defects or not based on the perception domain expansion surface pattern characteristics.
Specifically, in the technical scheme of the application, firstly, a digital image of a surface pattern of a printed circuit board acquired by a high-definition camera is acquired. The feature extraction of the digital image of the surface pattern is then performed using a convolutional neural network model that has excellent performance in implicit feature extraction of the image. In particular, it is considered that different local areas of the surface pattern of the printed circuit board may have different defect characteristics, and that processing the entire image may result in an excessive amount of computation. Therefore, in the technical scheme of the application, the digital image of the surface pattern is further subjected to image blocking processing to obtain a sequence of digital image blocks of the surface pattern. It will be appreciated that by dividing the digital image of the surface pattern into individual patches, the local features of each patch image may be better analyzed, which may help to more accurately detect and classify different types of pattern defects, improving the accuracy of overall surface pattern defect detection for a printed circuit board.
And then, respectively carrying out feature mining on the sequence of the surface pattern digital image blocks in a surface pattern local feature extractor based on a depth neural network model so as to respectively extract hidden feature distribution information of the surface pattern local area related to the printed circuit board in each surface pattern digital image block, thereby obtaining the sequence of the surface pattern local area feature map.
Accordingly, in step S143, the deep neural network model is a convolutional neural network model. It should be noted that the convolutional neural network (Convolutional Neural Network, abbreviated as CNN) is a deep learning model, which is mainly used for processing data with a similar grid structure, and the core idea of CNN is to extract the characteristics of input data through a convolutional layer and a pooling layer, and classify or regress through a fully connected layer. The following are the main components of CNN: 1. convolution layer (Convolutional Layer): the convolutional layer is the core component of the CNN that extracts local features of the input data by applying a series of convolutional kernels (also called filters) to the input data. The convolution operation may effectively capture spatial relationships in the image, such as edges, textures, and the like. 2. Activation function (Activation Function): after the convolution layer, a nonlinear activation function, such as ReLU (RECTIFIED LINEAR Unit), is typically applied to introduce nonlinear characteristics to increase the expression of the model. 3. Pooling layer (Pooling Layer): the pooling layer is used for reducing the space dimension of the output of the convolution layer, reducing the number of parameters and simultaneously retaining important characteristics. Common pooling operations include maximum pooling (Max Pooling) and average pooling (Average Pooling). 4. Full tie layer (Fully Connected Layer): the full join layer connects the outputs of the convolutional and pooling layers and classifies or regresses through a series of full join operations. The output of the fully connected layer may be used to predict the target class or generate a corresponding output. Dropout: to reduce the over-fitting, a common technique in CNNs is Dropout. Dropout is an operation that randomly discards a portion of neurons, forcing model learning to be a more robust and generalized feature. The training process of CNNs typically uses a back-propagation algorithm and a gradient descent optimization algorithm to update network parameters to minimize the loss function between the predicted output and the real labels. In general, convolutional neural networks can automatically learn the characteristic representation of input data through operations such as convolution, pooling, full connection and the like, and make important breakthroughs in image processing tasks.
Further, in order to enhance the expression and discrimination capability of the individual local features of the surface pattern, it is necessary to extract more discriminative features so as to perform defect detection and classification more accurately. Therefore, in the technical scheme of the application, the sequence of the surface pattern local area characteristic map is further respectively passed through a characteristic autocorrelation correlation strengthening module to obtain the sequence of the strengthening surface pattern local area characteristic map. Through the feature autocorrelation associated coding process, each local region feature of the surface pattern can be associated with related features around the local region feature, so that the expressive power of the features is enhanced, more context information and associated features can be captured, and the distinguishing degree and the robustness of the features are improved. Meanwhile, the characteristic autocorrelation can enable the characteristics of adjacent areas to be more consistent, which is very important for detecting defects of the surface patterns of the printed circuit board, because the defects generally cause local characteristic changes, and the characteristic differences caused by the defects can be reduced through characteristic autocorrelation enhancement, so that the defects are easier to detect.
Next, considering that it is difficult to extract global semantic related features in an image due to inherent limitations of CNN, global perception of the entire pattern is particularly important in performing surface pattern defect detection of a printed circuit board. Therefore, in the technical scheme of the application, the sequence of the local area characteristic map of the enhanced surface pattern is further arranged into a global surface pattern characteristic map, and then the perception domain expansion surface pattern characteristic map is obtained through a characteristic global perceptron based on a non-local neural network model. That is, by arranging the sequence of the enhanced surface pattern local area feature map as the global surface pattern feature map and processing in the non-local neural network model, global perception of the entire surface pattern can be achieved, which is very important for defect detection and classification of the pattern, since some defects may relate to global features of the entire pattern, through global perception, these global features can be better captured, and accuracy of defect detection is improved. In particular, here, the feature global perceptron based on the non-local neural network model captures hidden dependency information by calculating the similarity between the features of each local area of the surface pattern, so as to model the context features, so that the network focuses on the overall content between the features of each local area of the surface pattern, and further, the feature extraction capability of the backbone network is improved in classification and detection tasks, thereby being beneficial to improving the robustness of the features and the detection capability of complex defects.
Accordingly, in step S144, performing autocorrelation enhancement on each surface pattern local area feature map in the sequence of surface pattern local area feature maps to obtain a sequence of enhanced surface pattern local area feature maps, including: and respectively passing the sequence of the surface pattern local area characteristic diagrams through a characteristic autocorrelation correlation strengthening module to obtain the sequence of the strengthening surface pattern local area characteristic diagrams.
Accordingly, in step S145, globally arranging and sensing the sequence of the local area feature map of the enhanced surface pattern to obtain a sensing domain expansion surface pattern feature, including: and after the sequence of the local area characteristic diagrams of the enhanced surface pattern is arranged into a global surface pattern characteristic diagram, a perception domain expansion surface pattern characteristic diagram serving as the perception domain expansion surface pattern characteristic is obtained through a characteristic global sensor based on a non-local neural network model. It should be appreciated that the Non-local neural network model (Non-local Neural Network) is a deep learning model for image processing and computer vision tasks, the main purpose of which is to build long-range global dependencies in an image to capture global context information in the image. Conventional Convolutional Neural Networks (CNNs) focus mainly on feature extraction of local regions while ignoring global context information when processing images. This may lead to inaccuracy or incompleteness of the local features in certain tasks. The non-local neural network can effectively model long-range dependency relationships among pixels by introducing non-local operations, so that global features of an image are better captured. The main feature of the Non-local neural network model is the introduction of Non-local blocks (Non-local blocks) that perform self-attention operations on the input feature map. The self-attention mechanism allows each pixel to interact with other pixels, weighting the aggregate global context information by computing the similarity between pixels. This may enable each pixel to perceive the characteristics of the entire image, rather than being limited to a local neighborhood. Applications of the non-local neural network model include tasks such as image classification, object detection, image segmentation, and the like. By introducing global context information, the non-local neural network can improve the image understanding performance and improve the expressive force of the model in a complex scene. It achieves good effect in some tasks with long-range dependency, such as object relation modeling in images, video analysis, etc. In summary, the non-local neural network model can establish global dependency relationships in the image by introducing non-local operations, so that global features of the image are better captured. The method has important roles in image processing and computer vision tasks, and can improve the performance and expression capacity of the model.
And then, the perception domain expansion surface pattern feature map is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the surface pattern of the printed circuit board has defects. That is, the global perception enhancement characteristic information of the surface pattern of the printed circuit board is utilized to carry out classification processing, so that defect detection is automatically carried out on the surface pattern of the printed circuit board, and the detection efficiency and accuracy are improved.
Accordingly, determining whether a surface pattern of the printed circuit board is defective based on the perception domain-extended surface pattern features, comprising: and the perception domain expansion surface pattern feature diagram is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the surface pattern of the printed circuit board has defects.
More specifically, the perception domain expansion surface pattern feature map is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the surface pattern of the printed circuit board has defects or not, and the classification result comprises the following steps: expanding the perception domain expansion surface pattern feature map into a perception domain expansion surface pattern feature vector according to a row vector or a column vector; performing full-connection coding on the perception domain expansion surface pattern feature vector by using a full-connection layer of the classifier to obtain a coding classification feature vector; and inputting the coding classification feature vector into a Softmax classification function of the classifier to obtain the classification result.
That is, in the technical solution of the present application, the label of the classifier includes that the surface pattern of the printed circuit board is defective (first label) and that the surface pattern of the printed circuit board is not defective (second label), wherein the classifier determines to which classification label the perception domain expansion surface pattern feature map belongs through a soft maximum function. It should be noted that the first label p1 and the second label p2 do not include the concept of artificial setting, and in fact, during the training process, the computer model does not have the concept of "whether the surface pattern of the printed circuit board is defective", which is just two kinds of classification labels, and the probability that the output feature is the sum of the two classification labels sign, i.e., p1 and p2 is one. Therefore, the classification result of whether the surface pattern of the printed circuit board has defects is actually that the classification label is converted into the classification probability distribution conforming to the natural rule, and the physical meaning of the natural probability distribution of the label is essentially used instead of the language text meaning of whether the surface pattern of the printed circuit board has defects.
It should be appreciated that the role of the classifier is to learn the classification rules and classifier using a given class, known training data, and then classify (or predict) the unknown data. Logistic regression (logistics), SVM, etc. are commonly used to solve the classification problem, and for multi-classification problems (multi-class classification), logistic regression or SVM can be used as well, but multiple bi-classifications are required to compose multiple classifications, but this is error-prone and inefficient, and the commonly used multi-classification method is the softmax classification function.
Further, in the technical scheme of the application, the surface processing method of the printed circuit board further comprises the training steps of: the surface pattern local feature extractor is used for training the surface pattern local feature extractor based on the deep neural network model, the feature autocorrelation correlation strengthening module, the feature global perceptron of the non-local neural network model and the classifier. It should be understood that the training step refers to training each module involved using some marked data samples, including a surface pattern local feature extractor based on a deep neural network model, a feature autocorrelation correlation enhancement module, a feature global perceptron of a non-local neural network model, and a classifier, and is aimed at enabling the model to learn effective feature representation and classification capabilities for accurate surface finishing of printed circuit boards in practical applications. Specifically, the training step works as follows: 1. surface pattern local feature extractor training: by training the surface pattern local feature extractor based on the deep neural network model, the model can learn the ability to effectively extract local features of the surface pattern. These local features may include lines, corner points, textures, etc., which are important for subsequent feature processing and classification tasks. 2. Feature autocorrelation associated reinforcement module training: the training purpose of the feature autocorrelation correlation enhancement module is to let the model learn how to enhance the correlation between features through the self-attention mechanism. Therefore, the model can better capture the global dependency relationship in the surface pattern, and the expression capability and the discriminant of the features are improved. 3. Feature global perceptron training of non-local neural network models: by training the feature global perceptron of the non-local neural network model, the model can learn how to perceive image features in a global scope. This may allow the model to better understand the surface pattern of the entire printed circuit board, thereby improving performance of subsequent tasks. 4. Training a classifier: the classifier is trained to enable the model to accurately classify the surface pattern of the printed circuit board. By training using the marked data samples, the model can learn distinguishing features between different categories, thereby achieving accurate classification of the surface pattern. In general, the purpose of the training step is to enable the model to learn effective feature representation and classification capabilities by training the various modules using labeled data samples. In practical application, the trained model can accurately extract and classify the surface pattern of the printed circuit board, so that accurate surface processing is realized.
Wherein, in one example, the training step comprises: acquiring training data, wherein the training data comprises a digital image of a training surface pattern of a printed circuit board, and a true value of whether the surface pattern of the printed circuit board has defects or not; performing image blocking processing on the digital image of the training surface pattern to obtain a sequence of training surface pattern digital image blocks; respectively passing the sequence of training surface pattern digital image blocks through the surface pattern local feature extractor based on the deep neural network model to obtain a sequence of training surface pattern local region feature map; respectively passing the sequence of the training surface pattern local area characteristic diagram through the characteristic autocorrelation correlation strengthening module to obtain the sequence of the training strengthening surface pattern local area characteristic diagram; the sequence of the training enhancement surface pattern local area feature map is arranged into a training global surface pattern feature map, and then the training perception domain expansion surface pattern feature map is obtained through the feature global perceptron based on the non-local neural network model; optimizing the training perception domain expansion surface pattern feature vector obtained after the training perception domain expansion surface pattern feature map is expanded to obtain an optimized training perception domain expansion surface pattern feature vector; the optimized training perception domain expansion surface pattern feature vector passes through the classifier to obtain a classification loss function value; and training the surface pattern local feature extractor, the feature autocorrelation correlation enhancement module, the feature global perceptron of the non-local neural network model, and the classifier based on the classification loss function value and traveling in the direction of gradient descent.
In particular, in the technical solution of the present application, the feature matrix of each training enhanced surface pattern local area feature map in the sequence of training enhanced surface pattern local area feature maps is used to express the image semantic features of the feature autocorrelation correlation enhancement of the corresponding surface pattern digital image block in the local image space domain, so that after the sequence of training enhanced surface pattern local area feature maps is arranged as a training global surface pattern feature map, through the feature global perceptron based on the non-local neural network model, each feature matrix of the training perception domain expansion surface pattern feature map can express the image semantic feature representation of the global-local correlation enhancement of the digital image of the surface pattern, but considering that the channel distribution of the local feature extractor and the feature global perceptron is followed between each feature matrix of the training perception domain expansion surface pattern feature map, the training perception domain expansion surface pattern feature map as a whole has imbalance for the channel correlation expression of the local features and the global features. That is, in the case where the training perception domain expansion surface pattern feature map fuses the image semantic feature representations of global-local association reinforcement, the feature distribution information saliency of the image semantic feature representations at the respective association scales corresponding to the specific feature distributions thereof is also affected, so that it is difficult to stably focus on the prominent local distribution of the features when the training perception domain expansion surface pattern feature map is classified by the classifier, thereby affecting the training speed.
Therefore, the application carries out the iteration of classification regression on the training perception domain expansion surface pattern feature vector obtained after each time the training perception domain expansion surface pattern feature map is expanded through the classifierAnd (5) optimizing.
In one example, optimizing the training perception domain expansion surface pattern feature vector obtained after the training perception domain expansion surface pattern feature map is expanded to obtain an optimized training perception domain expansion surface pattern feature vector includes: optimizing the training perception domain expansion surface pattern feature vector by using the following optimization formula to obtain the optimized training perception domain expansion surface pattern feature vector; wherein, the optimization formula is:
wherein, Is the training perception domain extended surface pattern feature vector,/>And/>The training perception domain expands the surface pattern feature vector/>, respectivelySquare of 1-norm and 2-norm,/>Is the eigenvalue of the training perception domain expansion surface pattern eigenvector,/>Is the training perception domain expansion surface pattern feature vector/>Length of/(I)Represents a logarithmic function value based on 2, and/>Is a weight superparameter,/>Is the eigenvalue of the optimized training perception domain expansion surface pattern eigenvector.
Specifically, by extending the surface pattern feature vector based on the training perception domainGeometric registration of its high-dimensional feature manifold shape may be performed with respect to the training perceptron extended surface pattern feature vector/>Features with rich feature semantic information in the feature set formed by the feature values of (a), namely distinguishable stable interest features representing dissimilarity based on local context information when the classifier classifies, thereby realizing the training perception domain expansion surface pattern feature vector/>And the feature information significance is marked in the classification process, so that the training speed of the classifier is improved. Therefore, the automatic defect detection of the surface pattern of the printed circuit board can be realized, so that the detection efficiency and accuracy are improved, the degree of automation of the PCB manufacturing process and the stability of the product quality are improved, and the surface machining efficiency and the product quality of the printed circuit board are optimized.
Further, passing the optimized training perception domain expanded surface pattern feature vector through the classifier to obtain a classification loss function value, comprising: expanding surface pattern feature vectors for the optimized training perception domain by the classifier in a training classification formula to generate a training classification result; wherein, training classification formula is:
wherein, Representing the optimized training perception domain extended surface pattern feature vector,/>To/>For the weight matrix of each layer of full-connection layer,/>To/>Representing the bias matrix of each fully connected layer; and calculating a cross entropy value between the training classification result and the true value as the classification loss function value.
In summary, the surface processing method of the printed circuit board based on the embodiment of the application is explained, which can realize the automatic defect detection of the surface pattern of the printed circuit board, reduce the labor cost, improve the automation degree of the PCB manufacturing process and the stability of the product quality, and optimize the production efficiency and the product quality.
Fig. 4 is a block diagram of a surface finishing system 100 for a printed circuit board according to an embodiment of the present application. As shown in fig. 4, a surface processing system 100 of a printed circuit board according to an embodiment of the present application includes: a conductive material coating module 110 for coating a surface of the printed circuit board with a conductive material to form a conductive layer; an etching module 120, configured to cover the conductive layer with photoresist, and etch the photoresist into a desired pattern by exposing and developing to form a photoresist layer; a chemical removal module 130 for chemically or electrochemically removing the portion of the conductive layer not covered by the photoresist layer to form a surface pattern of the printed circuit board; the quality inspection module 140 is configured to perform quality inspection on the surface pattern of the printed circuit board to determine whether the surface pattern has a defect; and a photoresist layer removing module 150 for removing the photoresist layer to obtain a finished surface processing product of the printed circuit board.
In one example, in the above-described surface finishing system 100 for a printed circuit board, the quality inspection module 140 includes: the digital image acquisition unit is used for acquiring a digital image of the surface pattern of the printed circuit board acquired by the high-definition camera; the image blocking processing unit is used for carrying out image blocking processing on the digital image of the surface pattern to obtain a sequence of digital image blocks of the surface pattern; a feature extraction unit, configured to perform feature extraction on each surface pattern digital image block in the sequence of surface pattern digital image blocks by using a surface pattern local feature extractor based on a deep neural network model, so as to obtain a sequence of surface pattern local area feature maps; the self-correlation strengthening unit is used for carrying out self-correlation strengthening on each surface pattern local area characteristic diagram in the sequence of the surface pattern local area characteristic diagrams to obtain a sequence of strengthening surface pattern local area characteristic diagrams; the global arrangement and perception unit is used for globally arranging and perceiving the sequence of the local area characteristic map of the enhanced surface pattern to obtain perception domain expansion surface pattern characteristics; and a defect analysis unit for determining whether a surface pattern of the printed circuit board is defective based on the perception domain extended surface pattern feature.
Here, it will be understood by those skilled in the art that the specific functions and operations of the respective modules in the above-described surface processing system 100 for a printed circuit board have been described in detail in the above description of the surface processing method for a printed circuit board with reference to fig. 1 to 3, and thus, repetitive descriptions thereof will be omitted.
As described above, the surface processing system 100 of a printed circuit board according to an embodiment of the present application may be implemented in various wireless terminals, such as a server having a surface processing algorithm of a printed circuit board, and the like. In one example, the surface finishing system 100 for a printed circuit board according to an embodiment of the present application may be integrated into a wireless terminal as a software module and/or a hardware module. For example, the surface finishing system 100 of the printed circuit board may be a software module in the operating system of the wireless terminal or may be an application developed for the wireless terminal; of course, the surface finishing system 100 of the printed circuit board may also be one of a number of hardware modules of the wireless terminal.
Alternatively, in another example, the surface finishing system 100 of the printed circuit board and the wireless terminal may be separate devices, and the surface finishing system 100 of the printed circuit board may be connected to the wireless terminal through a wired and/or wireless network and transmit interactive information in accordance with a agreed data format.
Fig. 5 is an application scenario diagram of a surface processing method of a printed circuit board according to an embodiment of the present application. As shown in fig. 5, in this application scenario, first, a digital image of a surface pattern of a printed circuit board (for example, D illustrated in fig. 5) acquired by a high-definition camera is acquired, and then, the digital image of the surface pattern is input to a server (for example, S illustrated in fig. 5) where a surface processing algorithm of the printed circuit board is deployed, wherein the server can process the digital image of the surface pattern using the surface processing algorithm of the printed circuit board to obtain a classification result for indicating whether or not the surface pattern of the printed circuit board is defective.
Furthermore, those skilled in the art will appreciate that the various aspects of the application are illustrated and described in the context of a number of patentable categories or circumstances, including any novel and useful procedures, machines, products, or materials, or any novel and useful modifications thereof. Accordingly, aspects of the application may be performed entirely by hardware, entirely by software (including firmware, resident software, micro-code, etc.) or by a combination of hardware and software. The above hardware or software may be referred to as a "data block," module, "" engine, "" unit, "" component, "or" system. Furthermore, aspects of the application may take the form of a computer product, comprising computer-readable program code, embodied in one or more computer-readable media.
Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
The foregoing is illustrative of the present invention and is not to be construed as limiting thereof. Although a few exemplary embodiments of this invention have been described, those skilled in the art will readily appreciate that many modifications are possible in the exemplary embodiments without materially departing from the novel teachings and advantages of this invention. Accordingly, all such modifications are intended to be included within the scope of this invention as defined in the following claims. It is to be understood that the foregoing is illustrative of the present invention and is not to be construed as limited to the specific embodiments disclosed, and that modifications to the disclosed embodiments, as well as other embodiments, are intended to be included within the scope of the appended claims. The invention is defined by the claims and their equivalents.
Claims (8)
1. A surface processing method of a printed circuit board, comprising:
coating the surface of the printed circuit board with a conductive material to form a conductive layer;
covering photoresist on the conductive layer, and etching the photoresist into a required pattern by means of exposure and development to form a photoetching layer;
chemically or electrochemically removing the portions of the conductive layer not covered by the photolithographic layer to form a surface pattern of a printed circuit board;
Quality inspection is carried out on the surface pattern of the printed circuit board so as to judge whether the surface pattern has defects or not; and
Removing the photoetching layer to obtain a surface processing finished product of the printed circuit board;
the quality inspection of the surface pattern of the printed circuit board to judge whether the surface pattern has defects or not comprises the following steps:
Acquiring a digital image of the surface pattern of the printed circuit board acquired by a high-definition camera;
performing image blocking processing on the digital image of the surface pattern to obtain a sequence of digital image blocks of the surface pattern;
Respectively extracting the characteristics of each surface pattern digital image block in the sequence of the surface pattern digital image blocks by a surface pattern local characteristic extractor based on a depth neural network model to obtain a sequence of surface pattern local area characteristic diagrams;
Performing autocorrelation association strengthening on each surface pattern local area characteristic diagram in the sequence of surface pattern local area characteristic diagrams respectively to obtain a sequence of strengthened surface pattern local area characteristic diagrams;
globally arranging and sensing the sequence of the local area characteristic map of the enhanced surface pattern to obtain a sensing domain expansion surface pattern characteristic; and
And determining whether a surface pattern of the printed circuit board is defective based on the perception domain expansion surface pattern features.
2. The surface processing method of a printed circuit board according to claim 1, wherein the deep neural network model is a convolutional neural network model.
3. The surface processing method of a printed circuit board according to claim 2, wherein performing autocorrelation strengthening on each surface pattern local area feature map in the sequence of surface pattern local area feature maps to obtain a sequence of strengthened surface pattern local area feature maps, respectively, comprises:
And respectively passing the sequence of the surface pattern local area characteristic diagrams through a characteristic autocorrelation correlation strengthening module to obtain the sequence of the strengthening surface pattern local area characteristic diagrams.
4. A surface processing method of a printed circuit board according to claim 3, wherein globally arranging and sensing the sequence of the enhanced surface pattern local area feature map to obtain a sensing domain expansion surface pattern feature comprises:
and after the sequence of the local area characteristic diagrams of the enhanced surface pattern is arranged into a global surface pattern characteristic diagram, a perception domain expansion surface pattern characteristic diagram serving as the perception domain expansion surface pattern characteristic is obtained through a characteristic global sensor based on a non-local neural network model.
5. The surface processing method of a printed circuit board of claim 4, wherein determining whether a surface pattern of the printed circuit board is defective based on the perception domain expansion surface pattern feature comprises:
and the perception domain expansion surface pattern feature diagram is passed through a classifier to obtain a classification result, wherein the classification result is used for indicating whether the surface pattern of the printed circuit board has defects.
6. The surface finishing method of a printed circuit board according to claim 5, further comprising a training step of: the surface pattern local feature extractor is used for training the surface pattern local feature extractor based on the deep neural network model, the feature autocorrelation correlation strengthening module, the feature global perceptron of the non-local neural network model and the classifier.
7. The surface processing method of a printed circuit board according to claim 6, wherein the training step comprises:
acquiring training data, wherein the training data comprises a digital image of a training surface pattern of a printed circuit board, and a true value of whether the surface pattern of the printed circuit board has defects or not;
performing image blocking processing on the digital image of the training surface pattern to obtain a sequence of training surface pattern digital image blocks;
respectively passing the sequence of training surface pattern digital image blocks through the surface pattern local feature extractor based on the deep neural network model to obtain a sequence of training surface pattern local region feature map;
respectively passing the sequence of the training surface pattern local area characteristic diagram through the characteristic autocorrelation correlation strengthening module to obtain the sequence of the training strengthening surface pattern local area characteristic diagram;
the sequence of the training enhancement surface pattern local area feature map is arranged into a training global surface pattern feature map, and then the training perception domain expansion surface pattern feature map is obtained through the feature global perceptron based on the non-local neural network model;
Optimizing the training perception domain expansion surface pattern feature vector obtained after the training perception domain expansion surface pattern feature map is expanded to obtain an optimized training perception domain expansion surface pattern feature vector;
the optimized training perception domain expansion surface pattern feature vector passes through the classifier to obtain a classification loss function value; and
Training a surface pattern local feature extractor based on the deep neural network model, the feature autocorrelation correlation strengthening module, a feature global perceptron of the non-local neural network model and the classifier based on the classification loss function value and traveling in the direction of gradient descent.
8. The surface processing method of a printed circuit board of claim 7, wherein passing the optimized training perception domain expansion surface pattern feature vector through the classifier to obtain a classification loss function value comprises:
expanding surface pattern feature vectors for the optimized training perception domain by the classifier in a training classification formula to generate a training classification result; wherein, training classification formula is:
wherein, Representing the optimized training perception domain extended surface pattern feature vector,/>To/>For the weight matrix of each layer of full-connection layer,/>To/>Representing the bias matrix of each fully connected layer; and
And calculating a cross entropy value between the training classification result and the true value as the classification loss function value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410183031.9A CN117750643B (en) | 2024-02-19 | 2024-02-19 | Surface processing method of printed circuit board |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410183031.9A CN117750643B (en) | 2024-02-19 | 2024-02-19 | Surface processing method of printed circuit board |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117750643A CN117750643A (en) | 2024-03-22 |
| CN117750643B true CN117750643B (en) | 2024-04-19 |
Family
ID=90253052
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410183031.9A Active CN117750643B (en) | 2024-02-19 | 2024-02-19 | Surface processing method of printed circuit board |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117750643B (en) |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2000076245A (en) * | 1998-08-28 | 2000-03-14 | Seiko Epson Corp | Image producing method and device |
| DE20004439U1 (en) * | 1999-03-09 | 2000-10-05 | APPLIED PRECISION, INC., Issaquah, Wash. | Probe card examination system for determining the relative positions of probe tips |
| WO2016207875A1 (en) * | 2015-06-22 | 2016-12-29 | Photomyne Ltd. | System and method for detecting objects in an image |
| CN106852033A (en) * | 2017-04-06 | 2017-06-13 | 昆山苏杭电路板有限公司 | High precision part very high current printed circuit board processing method |
| CN111800954A (en) * | 2020-06-16 | 2020-10-20 | 珠海杰赛科技有限公司 | Reworking method for manufacturing abnormal product of copper-clad plate line and printed circuit board |
| KR102266605B1 (en) * | 2020-10-23 | 2021-06-17 | 김철희 | APPARATUS AND METHOd for inspecting printed circuit boards |
| EP3933691A1 (en) * | 2020-07-02 | 2022-01-05 | Robert Bosch GmbH | System and method to alter an image |
| KR20220023952A (en) * | 2020-08-20 | 2022-03-03 | 삼성디스플레이 주식회사 | Display device |
| WO2022065892A1 (en) * | 2020-09-24 | 2022-03-31 | 삼성전자 주식회사 | Electronic device |
| WO2022098058A1 (en) * | 2020-11-06 | 2022-05-12 | 삼성전자 주식회사 | Electronic device including display |
| CN117355038A (en) * | 2023-11-10 | 2024-01-05 | 江西红板科技股份有限公司 | X-shaped hole processing method and system for circuit board soft board |
| CN117528946A (en) * | 2023-11-23 | 2024-02-06 | 江西红板科技股份有限公司 | Preparation method and system of SIP hard board |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11927965B2 (en) * | 2016-02-29 | 2024-03-12 | AI Incorporated | Obstacle recognition method for autonomous robots |
| KR20220036540A (en) * | 2020-09-16 | 2022-03-23 | 삼성전자주식회사 | Patterned cover member and electronic device including the same |
-
2024
- 2024-02-19 CN CN202410183031.9A patent/CN117750643B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2000076245A (en) * | 1998-08-28 | 2000-03-14 | Seiko Epson Corp | Image producing method and device |
| DE20004439U1 (en) * | 1999-03-09 | 2000-10-05 | APPLIED PRECISION, INC., Issaquah, Wash. | Probe card examination system for determining the relative positions of probe tips |
| WO2016207875A1 (en) * | 2015-06-22 | 2016-12-29 | Photomyne Ltd. | System and method for detecting objects in an image |
| CN106852033A (en) * | 2017-04-06 | 2017-06-13 | 昆山苏杭电路板有限公司 | High precision part very high current printed circuit board processing method |
| CN111800954A (en) * | 2020-06-16 | 2020-10-20 | 珠海杰赛科技有限公司 | Reworking method for manufacturing abnormal product of copper-clad plate line and printed circuit board |
| EP3933691A1 (en) * | 2020-07-02 | 2022-01-05 | Robert Bosch GmbH | System and method to alter an image |
| KR20220023952A (en) * | 2020-08-20 | 2022-03-03 | 삼성디스플레이 주식회사 | Display device |
| WO2022065892A1 (en) * | 2020-09-24 | 2022-03-31 | 삼성전자 주식회사 | Electronic device |
| KR102266605B1 (en) * | 2020-10-23 | 2021-06-17 | 김철희 | APPARATUS AND METHOd for inspecting printed circuit boards |
| WO2022098058A1 (en) * | 2020-11-06 | 2022-05-12 | 삼성전자 주식회사 | Electronic device including display |
| CN117355038A (en) * | 2023-11-10 | 2024-01-05 | 江西红板科技股份有限公司 | X-shaped hole processing method and system for circuit board soft board |
| CN117528946A (en) * | 2023-11-23 | 2024-02-06 | 江西红板科技股份有限公司 | Preparation method and system of SIP hard board |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117750643A (en) | 2024-03-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111179251B (en) | Defect detection system and method based on twin neural network and by utilizing template comparison | |
| CN111008618B (en) | Self-attention deep learning end-to-end pedestrian re-identification method | |
| CN117355038B (en) | X-shaped hole processing method and system for circuit board soft board | |
| CN117784710B (en) | Remote state monitoring system and method for numerical control machine tool | |
| CN111598854B (en) | Segmentation method for small defects of complex textures based on rich robust convolution feature model | |
| CN114494780A (en) | Semi-supervised industrial defect detection method and system based on feature comparison | |
| CN117853492B (en) | Intelligent industrial defect detection method and system based on fusion model | |
| CN117392042A (en) | Defect detection method, defect detection apparatus, and storage medium | |
| CN117103790A (en) | Corrugated board production line and control method thereof | |
| Voon et al. | Computer-vision-based integrated circuit recognition using deep learning | |
| CN111882547A (en) | PCB missing part detection method based on neural network | |
| CN117434082A (en) | Quality detection method and system for silica gel product production line | |
| CN117750643B (en) | Surface processing method of printed circuit board | |
| CN117372528A (en) | Visual image positioning method for modularized assembly of mobile phone shell | |
| CN117315578A (en) | Monitoring method and system for rust area expansion by combining classification network | |
| CN114596273B (en) | Intelligent detection method for multiple defects of ceramic substrate by using YOLOV4 network | |
| CN115661042A (en) | Hierarchical classification defect detection method based on attention mechanism guidance | |
| CN114692887A (en) | Semi-supervised learning system and semi-supervised learning method | |
| CN111461060A (en) | Traffic sign identification method based on deep learning and extreme learning machine | |
| CN116525295B (en) | Metallized film for high-frequency pulse capacitor and preparation method thereof | |
| CN117274903B (en) | Intelligent early warning device and method for electric power inspection based on intelligent AI chip | |
| CN115482215B (en) | Method and system suitable for identifying true and false defects of PCB | |
| Wang et al. | A deep learning-based method for aluminium foil-surface defect recognition | |
| CN118470716A (en) | Sensor equipment production monitoring system and method | |
| CN112990145B (en) | Group-sparse-based age estimation method and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |