Circuit board FQC automatic inspection system and method

Abstract

The invention discloses a circuit board FQC automatic inspection system and method, and relates to the technical field of circuit boards, and the method comprises the following steps: 1, in an image collection stage, the system adaptively adjusts the intensity and angle of a light source according to the material, reflection characteristics and ambient light of a circuit board, and carries out the image collection through a multi-angle camera array; the system can capture all-dimensional image data of the circuit board; 2, an image preprocessing and enhancing module carries out denoising and enhancing on the image data, and the image definition is optimized through self-adaptive adjustment of the contrast ratio and the brightness; 3, a multi-layer defect detection and dynamic optimization module uses a multi-layer deep learning network to carry out defect identification, the system adopts a model combining a regional convolutional neural network and a deep neural network, and the detection process is optimized through a loss function and a regularization item; 4, dynamically adjusting a light source and a camera shooting angle by the system according to each detection result in a self-adaptive cyclic feedback mechanism; model parameters are dynamically optimized in each cycle.

Description

Technical Field

[0001] The present invention relates to the technical field of circuit boards, and in particular to a circuit board FQC automated inspection system and method. Background Art

[0002] Circuit boards are key components used in electronic devices to support electronic components and provide electrical connections. They are usually composed of an insulating substrate (such as glass fiber, epoxy resin, etc.) and conductive paths arranged on it. These conductive paths form circuits by printing or etching copper foil, supporting the welding of various electronic components to ensure the functionality of electronic devices.

[0003] In the electronics industry, the production requirements for circuit boards are becoming increasingly complex, and manual inspections are gradually exposing drawbacks such as high misjudgment rates and low efficiency. Existing automated inspection systems are difficult to adapt to the needs of rapidly changing production environments, product types, or lighting conditions.

[0004] Therefore, in response to the above problems, there is an urgent need for a circuit board FQC automated inspection system to improve data processing efficiency, ensure data privacy, and achieve more flexible system integration. Summary of the invention

[0005] The object of the present invention is to provide a circuit board FQC automated inspection system and method to solve the problems raised in the above background technology.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A circuit board FQC automated inspection system, comprising an image acquisition and intelligent light source control module, an image preprocessing and enhancement module, a multi-layer defect detection and dynamic optimization module, and an adaptive loop feedback and parameter optimization module;

[0008] The image acquisition and intelligent light source control module includes an image acquisition unit and an intelligent light source control unit. The image acquisition unit includes a multi-angle camera. The intelligent light source control unit includes a dynamic light source controller. The multi-angle camera array acquires raw image data. The dynamic light source controller adaptively adjusts the lighting conditions according to the material properties, reflectivity and ambient light brightness of the circuit board.

[0009] The image preprocessing and enhancement module performs noise reduction and enhancement detection on the image data of the image acquisition and intelligent light source control module through the image preprocessing and enhancement module to ensure that the data is suitable for model analysis;

[0010] The multi-layer defect detection and dynamic optimization module: The multi-layer defect detection module uses a deep convolutional neural network model to perform feature extraction and defect recognition on the enhanced image;

[0011] The adaptive loop feedback and parameter optimization module: The adaptive loop feedback mechanism dynamically adjusts system parameters according to the defect detection results, so that the detection results gradually approach the optimal state.

[0012] A further improvement of the technical solution of the present invention is that: the image acquisition and intelligent light source control module includes an image acquisition unit, an intelligent light source control unit and an image acquisition verification unit; the image acquisition unit includes a multi-angle camera, the multi-angle camera acquires raw image data, and records the acquired data as I raw The intelligent light source control unit automatically adjusts the light source angle and intensity based on the image contrast and brightness feedback. The adjusted image data formula is as follows:

[0013]

[0014] Among them I optimized is the optimized image data, θ i Indicates the light intensity at different angles; G(θ i ) is the light source correction factor; ω i To adaptively weight and highlight the contrast of key areas;

[0015] Image acquisition verification unit: detects image quality through preset brightness and contrast thresholds. If it does not meet the standards, it returns to the light source control module for further adjustment until the optimal image acquisition conditions are achieved.

[0016] A further improvement of the technical solution of the present invention is that the image preprocessing and enhancement module includes a noise reduction processing unit and a contrast enhancement unit; the noise reduction processing unit performs noise reduction processing on the data verified by the image acquisition and intelligent light source control module, and uses an adaptive noise reduction filtering algorithm to remove image noise. The specific analysis is as follows:

[0017]

[0018] Among them I denoised (x, y) is the pixel value after processing; I optimized (x+i, y+j) represents the pixel value in the original image within the neighborhood window; ω(i, j) is an adaptive weighting coefficient, which is usually dynamically adjusted based on the similarity and distance of pixel values ​​to reduce the impact of noise; is the sum of weighting coefficients and is used for normalization.

[0019] A further improvement of the technical solution of the present invention is that the contrast enhancement unit dynamically adjusts the contrast according to the grayscale distribution, and the enhancement formula is as follows:

[0020]

[0021] Among them I enhancedis the enhanced image; μ and σ are the average brightness and standard deviation of the image; α and β are dynamic adjustment coefficients calculated through intelligent optimization.

[0022] A further improvement of the technical solution of the present invention is that: the multi-layer defect detection and dynamic optimization module includes a feature extraction unit, a defect classification and grading unit and a dynamic optimization unit. The feature extraction unit uses a multi-layer convolutional neural network to extract image features and inputs the feature data into a defect classification model. The defect classification and grading: based on the feature data, the system classifies the defects through the R-CNN model to determine the type and severity of the defects.

[0023] A further improvement of the technical solution of the present invention is that: the error after each detection in the dynamic optimization unit is calculated by a loss function and used to optimize the model parameters. The loss function is defined as:

[0024]

[0025] in is a regularization term to prevent overfitting; y i , They are represented as actual labels and predicted values ​​respectively.

[0026] A further improvement of the technical solution of the present invention is that: the adaptive loop feedback and parameter optimization module dynamically adjusts the system parameters based on the defect detection results of the multi-layer defect detection and dynamic optimization module, and the adaptive loop feedback and parameter optimization module includes a defect feedback control unit, a light source and detection parameter adaptive adjustment unit, a loop termination condition unit and a detection report generation and optimization unit: the defect feedback control unit, the multi-layer defect detection and dynamic optimization module feeds back the identified defect information to the control module to determine whether it is necessary to re-shoot or adjust the light source angle; the light source and detection parameter adaptive adjustment unit: the control module optimizes the light source and detection parameters based on the feedback information to ensure that the defect areas that are difficult to detect are gradually clear in the cycle.

[0027] The further improvement of the technical solution of the present invention is that: the loop termination condition unit: the system determines whether the loop is terminated according to the change of the detection error and the similarity coefficient. If the following conditions are met, the loop is terminated. The specific analysis is as follows:

[0028]

[0029] Where ΔD j is the detection error change of the jth cycle, δ is the error convergence threshold; k is the current cycle number, N limit is the maximum number of cycles.

[0030] A further improvement of the technical solution of the present invention is that: the test report generation and optimization unit: after the cycle is completed, the system generates a complete test report, including the defect type, location, quantity and severity. Based on the test report, the test data and parameters are accumulated through multiple feedback cycles to form an adaptive adjustment strategy to adapt to different types of circuit board products.

[0031] A further improvement of the technical solution of the present invention is: a circuit board FQC automated inspection method, used to implement a circuit board FQC automated inspection system according to any one of claims 1 to 9, comprising the following steps:

[0032] Step 1: During the image acquisition phase, the system adaptively adjusts the light source intensity and angle according to the material, reflection characteristics and ambient light of the circuit board to ensure the best image acquisition quality. Through the multi-angle camera array, the system can capture all-round image data of the circuit board;

[0033] Step 2: The image preprocessing and enhancement module denoises and enhances the image data, and optimizes the image clarity by adaptively adjusting the contrast and brightness for better defect detection;

[0034] Step 3: The multi-layer defect detection and dynamic optimization module uses a multi-layer deep learning network for defect identification. The system adopts a model that combines regional convolutional neural networks and deep neural networks. It optimizes the detection process through loss functions and regularization terms to ensure the accuracy and stability of the detection model.

[0035] Step 4: In the adaptive loop feedback mechanism, the system dynamically adjusts the light source and camera angle according to the results of each detection to obtain higher quality data; at the same time, the model parameters are dynamically optimized in each cycle, and the loop feedback is terminated when the detection error reaches the set threshold or the maximum number of cycles is reached, ensuring the optimal detection effect.

[0036] Due to the adoption of the above technical solution, the present invention has the following technical advances compared with the prior art:

[0037] 1. The present invention provides an innovative circuit board FQC automated inspection system and method. The high-precision camera array and intelligent light source controller of the present invention ensure image clarity through real-time light source adjustment;

[0038] 2. The circuit board FQC automated inspection system and method provided by the present invention reduces noise, filters and enhances the collected images to ensure that the image data is suitable for subsequent analysis;

[0039] 3. The circuit board FQC automated inspection system and method provided by the present invention classifies and grades defects by adopting a regional convolutional neural network (R-CNN) and a deep neural network integrated model;

[0040] 4. The circuit board FQC automated inspection system and method provided by the present invention gradually optimizes the system inspection parameters by dynamically feeding back the inspection results to the image acquisition and defect detection modules. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0042] Figure 1 A flow chart of a circuit board FQC automated inspection system and method according to the present invention; DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] Examples, such as Figure 1 As shown, a circuit board FQC automated inspection system includes an image acquisition and intelligent light source control module, an image preprocessing and enhancement module, a multi-layer defect detection and dynamic optimization module, and an adaptive loop feedback and parameter optimization module;

[0045] The image acquisition and intelligent light source control module includes an image acquisition unit and an intelligent light source control unit. The image acquisition unit includes a multi-angle camera. The intelligent light source control unit includes a dynamic light source controller. The multi-angle camera array acquires raw image data. The dynamic light source controller adaptively adjusts the lighting conditions according to the material characteristics, reflectivity and ambient light brightness of the circuit board.

[0046] The image preprocessing and enhancement module performs noise reduction and enhancement detection on the image data of the image acquisition and intelligent light source control modules through the image preprocessing and enhancement module to ensure that the data is suitable for model analysis;

[0047] Multi-layer defect detection and dynamic optimization module: The multi-layer defect detection module uses a deep convolutional neural network model to extract features and identify defects from enhanced images;

[0048] Adaptive loop feedback and parameter optimization module: The adaptive loop feedback mechanism dynamically adjusts system parameters according to the defect detection results, so that the detection results gradually approach the optimal state.

[0049] Specifically, the image acquisition and intelligent light source control module includes an image acquisition unit, an intelligent light source control unit and an image acquisition verification unit; the image acquisition unit includes a multi-angle camera, which acquires raw image data and records the acquired data as I raw ; The intelligent light source control unit automatically adjusts the light source angle and intensity based on image contrast and brightness feedback. The adjusted image data formula is as follows:

[0050]

[0051] Among them I optimized is the optimized image data, θ i Indicates the light intensity at different angles; G(θ i ) is the light source correction factor; ω i To adaptively weight and highlight the contrast of key areas;

[0052] Image acquisition verification unit: detects image quality through preset brightness and contrast thresholds. If it does not meet the standards, it returns to the light source control module for further adjustment until the optimal image acquisition conditions are achieved.

[0053] The multi-angle camera array is used to obtain multi-angle images of the circuit board. The system adjusts the light source intensity and angle each time an image is captured to ensure that image data with ideal clarity and contrast can be obtained under different lighting conditions.

[0054] Specifically, the image preprocessing and enhancement module includes a noise reduction processing unit and a contrast enhancement unit; the noise reduction processing unit performs noise reduction processing on the data verified by the image acquisition and intelligent light source control module, and uses an adaptive noise reduction filtering algorithm to remove image noise. The specific analysis is as follows:

[0055]

[0056] Among them I denoised (x, y) is the pixel value after processing; I optimized (x+i, y+j) represents the pixel value in the original image within the neighborhood window; ω(i, j) is an adaptive weighting coefficient, which is usually dynamically adjusted based on the similarity and distance of pixel values ​​to reduce the impact of noise; is the sum of weighting coefficients and is used for normalization.

[0057] Specifically, the contrast enhancement unit dynamically adjusts the contrast according to the grayscale distribution, and the enhancement formula is as follows:

[0058]

[0059] Among them I enhanced is the enhanced image; μ and σ are the average brightness and standard deviation of the image; α and β are dynamic adjustment coefficients calculated through intelligent optimization.

[0060] The collected images are processed through noise reduction algorithms and contrast enhancement to remove noise and improve detail clarity. An adaptive contrast adjustment formula is used to optimize the image to make the features of different areas more prominent, facilitating subsequent defect detection.

[0061] Specifically, the multi-layer defect detection and dynamic optimization module includes a feature extraction unit, a defect classification and grading unit and a dynamic optimization unit. The feature extraction unit uses a multi-layer convolutional neural network to extract image features and inputs the feature data into the defect classification model. Defect classification and grading: Based on the feature data, the system classifies the defects through the R-CNN model to determine the type and severity of the defects.

[0062] The preprocessed image data is sent to the defect detection module, which performs feature extraction, defect identification and classification on the image based on a multi-layer deep learning algorithm (such as R-CNN). The defect types include poor solder joints, broken wires and short circuits. The result of each detection is optimized by the loss function to reduce detection errors and improve model robustness.

[0063] Specifically, in the dynamic optimization unit, the error after each detection is calculated by the loss function and used to optimize the model parameters. The loss function is defined as:

[0064]

[0065] in is a regularization term to prevent overfitting; y i , They are represented as actual labels and predicted values ​​respectively.

[0066] Specifically, the adaptive loop feedback and parameter optimization module dynamically adjusts system parameters based on the defect detection results of the multi-layer defect detection and dynamic optimization module. The adaptive loop feedback and parameter optimization module includes a defect feedback control unit, a light source and detection parameter adaptive adjustment unit, a loop termination condition unit and a detection report generation and optimization unit: a defect feedback control unit, the multi-layer defect detection and dynamic optimization module feeds back the identified defect information to the control module to determine whether it is necessary to re-shoot or adjust the light source angle; a light source and detection parameter adaptive adjustment unit: the control module optimizes the light source and detection parameters based on the feedback information to ensure that the defect areas that are difficult to detect are gradually clear in the cycle.

[0067] Specifically, the loop termination condition unit: the system determines whether the loop is terminated based on the change of the detection error and the similarity coefficient. If the following conditions are met, the loop is terminated. The specific analysis is as follows:

[0068]

[0069] Where ΔD j is the detection error change of the jth cycle, δ is the error convergence threshold; k is the current cycle number, N limit is the maximum number of cycles.

[0070] The detection results are fed back to the control module. The system adjusts the light source angle, intensity and camera angle according to the detection error and re-captures the image after optimization. The process enters a loop and the system automatically performs feedback adjustments, gradually approaching the optimal detection effect until the detection accuracy reaches the preset requirements or the number of cycles reaches the limit.

[0071] Specifically, the test report generation and optimization unit: After the cycle is completed, the system generates a complete test report, including the defect type, location, quantity and severity. Based on the test report, it accumulates test data and parameters through multiple feedback cycles to form an adaptive adjustment strategy to adapt to different types of circuit board products.

[0072] After the cycle inspection is completed, the system generates an inspection report containing the defect type, location, quantity and severity, providing data support for subsequent production process adjustments and quality control. The data and adjustment parameters of each inspection are recorded to form an adaptive adjustment strategy to facilitate rapid inspection and automated adjustment of similar circuit boards in the future, further improving inspection efficiency and accuracy.

[0073] Specifically, a circuit board FQC automated inspection method is used to implement a circuit board FQC automated inspection system according to any one of claims 1 to 9, comprising the following steps:

[0074] Step 1: During the image acquisition phase, the system adaptively adjusts the light source intensity and angle according to the material, reflection characteristics and ambient light of the circuit board to ensure the best image acquisition quality. Through the multi-angle camera array, the system can capture all-round image data of the circuit board;

[0075] Step 2: The image preprocessing and enhancement module denoises and enhances the image data, and optimizes the image clarity by adaptively adjusting the contrast and brightness for better defect detection;

[0076] Step 3: The multi-layer defect detection and dynamic optimization module uses a multi-layer deep learning network for defect identification. The system adopts a model that combines regional convolutional neural networks and deep neural networks. It optimizes the detection process through loss functions and regularization terms to ensure the accuracy and stability of the detection model.

[0077] Step 4: In the adaptive loop feedback mechanism, the system dynamically adjusts the light source and camera angle according to the results of each detection to obtain higher quality data; at the same time, the model parameters are dynamically optimized in each cycle, and the loop feedback is terminated when the detection error reaches the set threshold or the maximum number of cycles is reached, ensuring the optimal detection effect.

[0078] Working principle: Through the multi-angle camera array, the system can capture all-round image data of the circuit board. The system adaptively adjusts the light source intensity and angle according to the material, reflective characteristics and ambient light of the circuit board. The collected data is denoised and enhanced by the image preprocessing and enhancement module, and the image clarity is optimized by adaptively adjusting the contrast and brightness. Then the defect detection and model optimization module: using a multi-layer deep learning network for defect identification, the system adopts a model combining regional convolutional neural network and deep neural network, and optimizes the detection process through loss function and regularization term. The system dynamically adjusts the light source and camera angle according to the results of each detection to obtain higher quality data; the data and adjusted parameters of each detection are recorded to form an adaptive adjustment strategy, so as to facilitate rapid detection and automatic adjustment of similar circuit boards in the future, and further improve detection efficiency and accuracy.

[0079] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A circuit board FQC automated inspection system, characterized by: It includes image acquisition and intelligent light source control module, image preprocessing and enhancement module, multi-layer defect detection and dynamic optimization module and adaptive loop feedback and parameter optimization module; The image acquisition and intelligent light source control module includes an image acquisition unit and an intelligent light source control unit. The image acquisition unit includes a multi-angle camera. The intelligent light source control unit includes a dynamic light source controller. The multi-angle camera array acquires raw image data. The dynamic light source controller adaptively adjusts the lighting conditions according to the material properties, reflectivity and ambient light brightness of the circuit board. The image preprocessing and enhancement module performs noise reduction and enhancement detection on the image data of the image acquisition and intelligent light source control module through the image preprocessing and enhancement module to ensure that the data is suitable for model analysis; The multi-layer defect detection and dynamic optimization module: The multi-layer defect detection module uses a deep convolutional neural network model to perform feature extraction and defect recognition on the enhanced image; The adaptive loop feedback and parameter optimization module: The adaptive loop feedback mechanism dynamically adjusts system parameters according to the defect detection results, so that the detection results gradually approach the optimal state.

2. According to claim 1, a circuit board FQC automated inspection system is characterized by: The image acquisition and intelligent light source control module includes an image acquisition unit, an intelligent light source control unit and an image acquisition verification unit; the image acquisition unit includes a multi-angle camera, the multi-angle camera acquires raw image data and records the acquired data as I raw The intelligent light source control unit automatically adjusts the light source angle and intensity based on the image contrast and brightness feedback. The adjusted image data formula is as follows: Among them I optimized is the optimized image data, θ i Represents the light intensity at different angles; G(θ i ) is the light source correction factor; ω i To adaptively weight and highlight the contrast of key areas; Image acquisition verification unit: detects image quality through preset brightness and contrast thresholds. If it does not meet the standards, it returns to the light source control module for further adjustment until the optimal image acquisition conditions are achieved.

3. A circuit board FQC automated inspection system according to claim 1, characterized in that: The image preprocessing and enhancement module includes a noise reduction processing unit and a contrast enhancement unit; the noise reduction processing unit performs noise reduction processing on the data verified by the image acquisition and intelligent light source control module, and uses an adaptive noise reduction filtering algorithm to remove image noise. The specific analysis is as follows: Among them I denoised (x, y) is the pixel value after processing; I optimized (x+i, y+j) represents the pixel value in the original image within the neighborhood window; ω(i,j) is an adaptive weighting coefficient, which is usually dynamically adjusted based on the similarity and distance of pixel values ​​to reduce the impact of noise; is the sum of weighting coefficients and is used for normalization.

4. A circuit board FQC automated inspection system according to claim 3, characterized in that: The contrast enhancement unit dynamically adjusts the contrast according to the grayscale distribution, and the enhancement formula is as follows: Among them I enhanced is the enhanced image; μ and σ are the average brightness and standard deviation of the image; α and β are dynamic adjustment coefficients calculated through intelligent optimization.

5. The circuit board FQC automated inspection system according to claim 1, characterized in that: The multi-layer defect detection and dynamic optimization module includes a feature extraction unit, a defect classification and grading unit and a dynamic optimization unit. The feature extraction unit uses a multi-layer convolutional neural network to extract image features and inputs the feature data into a defect classification model. The defect classification and grading: based on the feature data, the system classifies the defects through the R-CNN model to determine the type and severity of the defects.

6. A circuit board FQC automated inspection system according to claim 5, characterized in that: In the dynamic optimization unit, the error after each detection is calculated by the loss function and used to optimize the model parameters. The loss function is defined as: in is a regularization term to prevent overfitting; y i , They are represented as actual labels and predicted values ​​respectively.

7. The circuit board FQC automated inspection system according to claim 1, characterized in that: The adaptive loop feedback and parameter optimization module dynamically adjusts system parameters based on the defect detection results of the multi-layer defect detection and dynamic optimization module. The adaptive loop feedback and parameter optimization module includes a defect feedback control unit, a light source and detection parameter adaptive adjustment unit, a loop termination condition unit and a detection report generation and optimization unit: the defect feedback control unit, the multi-layer defect detection and dynamic optimization module feeds back the identified defect information to the control module to determine whether it is necessary to re-shoot or adjust the light source angle; the light source and detection parameter adaptive adjustment unit: the control module optimizes the light source and detection parameters based on the feedback information to ensure that the defect areas that are difficult to detect are gradually clear in the cycle.

8. The circuit board FQC automated inspection system according to claim 7, characterized in that: The loop termination condition unit: The system determines whether the loop is terminated based on the change of the detection error and the similarity coefficient. If the following conditions are met, the loop is terminated. The specific analysis is as follows: Where ΔD j is the detection error change of the jth cycle, δ is the error convergence threshold; k is the current cycle number, N limit is the maximum number of cycles.

9. The circuit board FQC automated inspection system according to claim 1, characterized in that: The test report generation and optimization unit: After the cycle is completed, the system generates a complete test report including defect type, location, quantity and severity. Based on the test report, the system accumulates test data and parameters through multiple feedback cycles to form an adaptive adjustment strategy to adapt to different types of circuit board products.

10. A circuit board FQC automated inspection method, used to implement a circuit board FQC automated inspection system according to any one of claims 1 to 9, comprising the following steps: Step 1: During the image acquisition phase, the system adaptively adjusts the light source intensity and angle according to the material, reflection characteristics and ambient light of the circuit board to ensure the best image acquisition quality. Through the multi-angle camera array, the system can capture all-round image data of the circuit board; Step 2: The image preprocessing and enhancement module denoises and enhances the image data, and optimizes the image clarity by adaptively adjusting the contrast and brightness for better defect detection; Step 3: The multi-layer defect detection and dynamic optimization module uses a multi-layer deep learning network for defect identification. The system adopts a model that combines regional convolutional neural networks and deep neural networks. It optimizes the detection process through loss functions and regularization terms to ensure the accuracy and stability of the detection model. Step 4: In the adaptive loop feedback mechanism, the system dynamically adjusts the light source and camera angle according to the results of each detection to obtain higher quality data; at the same time, the model parameters are dynamically optimized in each cycle, and the loop feedback is terminated when the detection error reaches the set threshold or the maximum number of cycles is reached, ensuring the optimal detection effect.

Images