A PCB micro-defect closed-loop detection method, system and device

Abstract

The present application relates to the technical field of automatic optical inspection, and particularly relates to a PCB micro-defect closed-loop detection method, system and device, the method comprising: positioning and pressing the PCB to be detected based on a preset positioning reference, and collecting a board image; registering and differentiating the board image and a reference image to obtain a candidate defect region; extracting features from the candidate defect region as spatial prior, and adaptively weighting high-frequency components and low-frequency components and fusing residuals to obtain defect-enhanced features; performing defect detection based on the defect-enhanced features to obtain a defect detection result; adjusting imaging parameters based on the defect detection result, and collecting a board image according to the adjusted imaging parameters to perform defect detection until the defect detection result meets an output condition and is output. Through the present application, the problem of insufficient micro-defect detection accuracy and stability caused by the lack of reproducibility of imaging input and the mutual fragmentation of detection algorithms and imaging processes in the existing PCB micro-defect detection is effectively solved.

Description

Technical Field

[0001] This invention relates to the field of automated optical inspection technology, and in particular to a closed-loop inspection method, system and device for micro-defects in PCBs. Background Technology

[0002] Printed circuit board (PCB) micro-defect detection is an automated optical inspection technology that uses machine vision to acquire images of the PCB surface and identifies abnormalities through image processing or intelligent detection algorithms. Existing technologies, such as CN115719326A, determine defect areas by comparing the image of the PCB under test with a standard PCB image, and then use a neural network to identify the defect type; CN119540152A acquires PCB images through a camera, light source, and clamping structure, and combines this with an improved target detection network to achieve defect identification.

[0003] However, existing solutions lack effective coordination between board positioning, imaging conditions, and defect detection processes. Changes in board pose and imaging state during repeated detection can affect reference image comparison and micro-defect feature extraction. Furthermore, the detection results are difficult to apply back to the imaging process, resulting in insufficient consistency of input images and further affecting the accuracy and stability of micro-defect detection. Therefore, there is an urgent need for a PCB micro-defect detection solution that can achieve reproducible imaging and form a collaborative closed loop between the imaging process and the defect detection process.

[0004] The information disclosed in this background section is intended only to enhance the understanding of the general background of this disclosure and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0005] This invention provides a closed-loop detection method, system, and device for minute defects in PCBs, which can effectively solve the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A closed-loop detection method for minute defects in PCBs, the method comprising: The PCB under test is positioned and pressed based on a preset positioning reference, and the first board surface image is acquired according to the first imaging parameters. The first plate image is registered and differencing with a pre-stored reference image to obtain candidate defect regions; The candidate defect region is used as a spatial prior to extract features from the first plate image. The high-frequency and low-frequency components in the extracted features are adaptively weighted and residually fused to obtain defect enhancement features. Based on the defect enhancement features, defect detection is performed on the PCB under test to obtain a first defect detection result; Based on the first defect detection result and the image quality parameters of the first board image, the first imaging parameters are adjusted, and the board image is re-acquired according to the adjusted first imaging parameters for defect detection until the preset output conditions are met and the defect detection result is output.

[0007] Furthermore, the preset positioning reference includes a fixed right-angle reference block that defines the orientation of the PCB board surface under test along a first direction and a second direction, and an adjustable positioning block, and the PCB board under test is held in the orientation of the board surface determined by the preset positioning reference by a clamping mechanism.

[0008] Further, acquiring a first plate surface image according to the first imaging parameters includes: The first imaging parameters are determined based on the thickness and surface reflectivity of the PCB under test, and the imaging distance and lighting conditions between the camera and the PCB under test are adjusted based on the first imaging parameters. In an imaging environment that suppresses stray ambient light, the first board surface image is acquired according to the adjusted first imaging parameters so that the PCB under test of the same board type has reproducible imaging conditions.

[0009] Further, the first plate image and the pre-stored reference image are registered and differencing, including: Based on the reproducible board pose formed by the preset positioning reference, the first board image and the pre-stored reference image are sub-pixel registered to obtain a registered image. The registered image and the pre-stored reference image are subjected to differential processing to obtain the candidate defect region that represents the suspected defect location; The first plate image is subjected to illumination normalization and geometric correction based on the preset positioning reference, and the candidate defect region is mapped to the corrected first plate image to form the spatial prior for feature extraction.

[0010] Furthermore, feature extraction is performed on the first plate image based on the candidate defect regions, including: Multi-layer feature extraction and feature fusion are performed on the first plate image through the backbone network and neck network of the target detection network; After the feature fusion module of the backbone network and the neck network, the salience of each feature position relative to the neighborhood features is determined based on the parameterless attention mechanism, and the corresponding features are weighted according to the salience. The spatial prior is fused with the weighted features to enhance the micro-defect features corresponding to the candidate defect region.

[0011] Furthermore, adaptive weighting and residual fusion are performed on the high-frequency and low-frequency components, including: The low-frequency components are obtained by performing low-frequency extraction on the input features obtained from the feature extraction. The high-frequency component is determined based on the difference between the input feature and the low-frequency component; Frequency enhancement features are obtained by adaptively weighting the high-frequency components and the low-frequency components using learnable weights. The frequency enhancement feature is residually fused with the input feature to obtain the defect enhancement feature.

[0012] Further, when there is a domain difference between the first plate image and the training image used to train the defect detection model, it includes: Unsupervised image domain transformation is performed on the first board image to obtain a transformed image with a corresponding image style to the training image; The transformed image is used to replace the first board image for feature extraction, so that the defect detection model can perform defect detection on PCB images of different board types or different image sources.

[0013] Furthermore, the first imaging parameters are adjusted based on the first defect detection result and the image quality parameters of the first plate surface image, including: The detection confidence level is determined based on the first defect detection result, and the image quality parameter characterizing the current imaging state is determined based on the first plate image. The detection confidence and the image quality parameters are compared with their respective preset thresholds, and the first panel image is determined not to meet the preset output conditions when at least one of them does not meet the corresponding preset threshold. Based on the detection confidence level and the image quality parameters, at least one of the first imaging parameters, namely camera height and light source intensity, is adjusted, and the board surface image is re-acquired according to the adjusted first imaging parameters for defect detection until the preset output conditions are met.

[0014] A closed-loop inspection system for minute defects in PCBs, the system comprising: The positioning and pressing imaging module positions and presses the PCB under test based on a preset positioning reference, and acquires the first board surface image according to the first imaging parameters; The registration and difference processing module registers and differs the first plate image with the pre-stored reference image to obtain candidate defect regions; The prior feature enhancement module uses the candidate defect region as a spatial prior to extract features from the first plate image, and performs adaptive weighting and residual fusion on the high-frequency and low-frequency components in the extracted features to obtain defect enhancement features. The defect detection and identification module performs defect detection on the PCB under test based on defect enhancement features and obtains the first defect detection result. The closed-loop imaging adjustment module adjusts the first imaging parameters based on the first defect detection result and the image quality parameters of the first plate image, and re-acquires the plate image according to the adjusted first imaging parameters to perform defect detection until the preset output conditions are met and then outputs the defect detection result.

[0015] A closed-loop detection device for minute defects in PCBs, the device comprising a base 1, a detection stage 2, a positioning and clamping mechanism, a light-shielding and illumination mechanism, and an imaging adjustment mechanism; The testing stage 2 is disposed on the base 1, and the testing stage 2 is provided with a T-slot 3; The positioning and clamping mechanism includes a fixed right-angle reference block 4, a movable positioning block 5, a locking knob 6, a T-shaped slider 7, and a downward-pressing quick clamp 9. The fixed right-angle reference block 4 is set at the right-angle position of the detection platform 2 and forms a first positioning reference and a second positioning reference that are perpendicular to each other. The movable positioning block 5 slides with the T-shaped slider 7 and the T-shaped groove 3 and is positioned by the locking knob 6 so that the PCB board 8 abuts against the fixed right-angle reference block 4 in two directions. The pressing quick clamp 9 is disposed on the testing platform 2. The pressing end of the pressing quick clamp 9 is provided with a rubber pressure head 10, which is used to press the positioned PCB board 8 onto the testing platform 2. The imaging adjustment mechanism includes a vertical guide rail 12, a lifting slider 13, a horizontal camera frame 14, and a camera light source mounting plate 15. The lifting slider 13 is slidably connected to the vertical guide rail 12, the horizontal camera frame 14 is connected to the lifting slider 13, and the camera light source mounting plate 15 is disposed on the horizontal camera frame 14 so as to adjust the height of the camera light source mounting plate 15 relative to the PCB board 8 by means of the lifting slider 13. The light-shielding illumination mechanism includes a light-shielding plate 11 and a ring light source 18. The light-shielding plate 11 surrounds the imaging area of ​​the detection stage 2, and the ring light source 18 is disposed on the camera light source mounting plate 15 for illuminating the PCB board 8 in the light-shielding environment formed by the light-shielding plate 11. The camera light source mounting plate 15 is also provided with an industrial camera 16 and a lens 17 connected to the industrial camera 16. The industrial camera 16, the lens 17 and the ring light source 18 move up and down synchronously with the lifting slider 13 to acquire the board surface image of the PCB board 8 under the conditions of the reproducible board surface posture formed by the positioning and pressing mechanism and the controlled lighting formed by the light shielding and lighting mechanism.

[0016] The technical solution of this invention can achieve the following technical effects: By performing reproducible positioning and clamping and controlled imaging on the PCB under test, the board surface image and the reference image are registered and differentiated to form a candidate defect space prior. Combined with adaptive fusion of high and low frequency features, the detection of small defects is enhanced. Based on the detection results and image quality feedback, the imaging parameters are adjusted to form a closed-loop detection. This effectively solves the problem of insufficient accuracy and stability of small defect detection in existing PCB small defect detection due to the lack of reproducibility of imaging input and the disconnect between the detection algorithm and the imaging process.

[0017] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating a closed-loop detection method for minute defects in PCBs. Figure 2 This is a schematic diagram of a closed-loop detection device for minor defects in PCBs. Figure 3 This is a logic diagram of a closed-loop detection method for minor defects in PCBs. Figure 4 This is a block diagram of the MicroFuse-YOLOv10 network structure. Figure 5 A schematic diagram showing the embedding location of the SimAM attention module; Figure 6 Block diagram of FGRF frequency-guided residual fusion module; Figure 7 Here is a flowchart of the CycleGAN cross-domain front-end process; Figure 8 This is a flowchart of closed-loop feedback control.

[0020] Reference numerals: 1. Base; 2. Inspection platform; 3. T-slot; 4. Fixed right-angle reference block; 5. Movable positioning block; 6. Locking knob; 7. T-slide block; 8. PCB board; 9. Downward-pressing quick clamp; 10. Rubber indenter; 11. Light shield; 12. Vertical guide rail; 13. Lifting slide block; 14. Horizontal camera stand; 15. Camera light source mounting plate; 16. Industrial camera; 17. Lens; 18. Ring light source. Detailed Implementation

[0021] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0022] Unless otherwise defined, all 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 pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0023] Example 1; as Figure 1 As shown, this application provides a closed-loop detection method for minute defects in PCBs, the method including: S10: Position and clamp the PCB to be tested based on the preset positioning reference, and acquire the first board surface image according to the first imaging parameters; S20: Register and differ the first plate image with the pre-stored reference image to obtain candidate defect regions; S30: The candidate defect region is used as a spatial prior to extract features from the first plate image, and the high-frequency and low-frequency components in the extracted features are adaptively weighted and residual fused to obtain defect enhancement features. S40: Perform defect detection on the PCB under test based on defect enhancement features to obtain the first defect detection result; S50: Adjust the first imaging parameters based on the first defect detection result and the image quality parameters of the first board image, and re-acquire the board image according to the adjusted first imaging parameters to perform defect detection until the preset output conditions are met and the defect detection result is output.

[0024] Specifically, this embodiment provides a closed-loop detection method for PCB micro-defects, used to detect micro-defects such as open circuits, short circuits, rodent bites, burrs, excess copper, and leaks on the PCB board surface; the method forms a closed-loop detection process through reproducible positioning imaging, reference image differentiation, micro-defect feature enhancement, and imaging adjustment driven by detection results. In this embodiment, the preset positioning reference refers to a pre-established mechanical positioning reference that enables different PCBs under test to repeatedly reach a basically consistent board surface position and angle; the first imaging parameter refers to the imaging conditions used when initially acquiring the board surface image, including at least the imaging height of the industrial camera and the illumination intensity of the ring light source; the candidate defect region refers to the suspected defect region obtained after registering and differencing the board surface image of the PCB under test with a defect-free reference image; the spatial prior refers to the suspected defect location information provided by the candidate defect region, which is used to guide the subsequent detection network to focus on processing the corresponding region and filter out large areas of normal regions; the defect enhancement feature refers to the feature formed by weighting the high-frequency and low-frequency components in the detection feature and then performing residual fusion; the image quality parameter refers to the quality index used to characterize whether the current board surface image is suitable for defect detection, including sharpness and brightness uniformity; the preset output condition refers to the state in which the detection confidence and image quality parameters reach the corresponding set threshold. like Figure 2 and Figure 3 As shown, the PCB board 8 to be tested is placed on the testing stage 2, and pushed along the T-slot 3 to the fixed right-angle reference block 4, so that the adjacent two sides of the PCB board 8 to be tested are respectively in contact with the X-direction reference and the Y-direction reference. Then, the movable positioning block 5 is moved to limit the PCB board 8 to be tested from the other two directions. The movable positioning block 5 is fixed by the locking knob 6. Then, the pressing quick clamp 9 and the rubber pressure head 10 are used to press the PCB board 8 to be tested onto the testing stage 2, so that the board surface posture of the PCB board 8 to be tested is consistent with the pre-calibrated reference posture. The fixed right-angle reference block 4 thus constitutes the preset positioning reference, so that the position and angle of the repeatedly placed PCB board 8 in the field of view of the industrial camera 16 are reproducible. Adjust the lifting slider 13 according to the thickness and surface reflectivity of the PCB board 8 under test, so that the industrial camera 16 and lens 17 reach the pre-calibrated imaging height, and turn on the ring light source 18 for illumination; the light shield 11 is set up around the imaging area to isolate ambient stray light; the same board type on the same production line uses the same camera height and illumination parameters to complete batch imaging, and the first board surface image is acquired by the industrial camera 16 under the imaging conditions; the camera height and the illumination intensity of the ring light source 18 together constitute the first imaging parameters; Acquire a pre-stored defect-free reference image corresponding to the current 8-type PCB board under test; since mechanical positioning has made the first board image and the pre-stored defect-free reference image have basically the same board pose, there is a small residual positional deviation between the two images. Therefore, subpixel registration is first used to align the first board image and the pre-stored defect-free reference image, and then the two registered images are subdivided to obtain candidate difference regions; the candidate difference regions are used to roughly determine the location of suspected defects and serve as the spatial prior for subsequent feature extraction to reduce the interference of normal board areas on the detection of minor defects; Before extracting defect features, the first board image is subjected to illumination homogenization and highlight suppression to make the brightness differences caused by reflections from different times or different board surfaces tend to be consistent; the board pose is obtained based on the board edge position determined by the fixed right-angle reference block 4, and the detection area is aligned to a unified coordinate system through rotation correction and perspective correction so that the defect position obtained later can correspond to the physical position of the PCB board. like Figure 4 , Figure 5 and Figure 6 As shown, this embodiment uses MicroFuse-YOLOv10 for defect feature extraction and defect detection, and the base network uses YOLOv10n; the suspected defect location provided by the candidate defect region is used as a spatial prior, so that the detection network focuses on processing the corresponding region; after each C2f feature fusion module of the backbone network and the neck network, a SimAM parameterless attention module is embedded. SimAM determines the importance of features by calculating the difference between neurons and other neurons in the same channel, without adding additional trainable weights; The energy function of SimAM is: e t ∗ = 4(σ 2 +λ) / [(t−μ) 2 +2σ 2 +2λ]; Where t is the target neuron, μ, σ 2 Let be the mean and variance of the remaining neurons within the same channel, and λ be the regularization term; the feature weighting formula is: X' = ​​sigmoid(1 / e t ∗ ) · X; The above treatments improve the response of local mutation features such as mouse bites and spiking in the network. To further preserve edge and texture information that is easily lost during downsampling due to minor defects, a Frequency-Guided Residual Fusion (FGRF) module is set up in the feature fusion stage. For the input feature X, a low-frequency component X_low is extracted using average pooling with a size of 3×3, a stride of 1, and padding of 1. Then, the high-frequency component is obtained based on the difference between the input feature X and the low-frequency component X_low. The calculation formula for the high-frequency component is as follows: X_high = X − X_low; The high-frequency and low-frequency components are recombine using two learnable coefficients α and β to form an enhancement term, as shown in the formula: E = α·X_high + β·X_low; The enhancement term is then added to the input features as a residual, using the following formula: X' = ​​X + E; The two learnable coefficients are initialized as follows: α=1, β=0; In the early stages of training, high-frequency details are mainly enhanced. During training, the network learns the combination ratio of high-frequency and low-frequency components to compensate for the edge and texture information corresponding to minor defects. The fused X' is used as the defect enhancement feature. Defect category and defect location are predicted using defect enhancement features to obtain the first defect detection result; the first defect detection result includes at least the defect category, defect bounding box and detection confidence, and the defect categories include open circuit, short circuit, rodent bite, burr, excess copper and leak; After completing a defect detection, the detection confidence level corresponding to the first defect detection result is read. At the same time, the sharpness and brightness uniformity are obtained from the first panel image as image quality parameters. The detection confidence level and image quality parameters are compared with the corresponding preset thresholds. When the detection confidence level or image quality parameter is lower than the corresponding threshold, it is determined that the current imaging state is not conducive to the detection of small defects. When the current imaging state does not meet the preset output conditions, the height of the lifting slider 13 and the illumination intensity of the ring light source 18 are finely adjusted according to the detection confidence and image quality parameters to update the first imaging parameters; the board surface image is re-acquired according to the updated imaging parameters, and registration difference, spatial prior acquisition, feature enhancement and defect detection are performed again; when the newly obtained detection confidence and image quality parameters reach the corresponding threshold, the imaging parameter adjustment is stopped and the current defect detection result is output, thus forming a closed-loop detection process from board surface imaging, image analysis, defect detection to imaging parameter feedback adjustment and re-imaging; In other embodiments, the preset positioning reference can also achieve reproducible positioning through vacuum adsorption positioning, pin hole positioning, or visual positioning marks; low-frequency extraction in high-low frequency separation can also use Gaussian filtering or wavelet transform; image quality parameters can also include contrast or defect quantity anomaly indicators; adjustable imaging parameters can also include light source angle or exposure time; the above alternative methods all maintain the technical method consisting of reproducible imaging, defect space prior, frequency feature enhancement, and detection feedback adjustment.

[0025] The technical solution of this invention enables reproducible positioning, clamping, and controlled imaging of the PCB under test. The board surface image and the reference image are registered and differentiated to form a candidate defect space prior. The high- and low-frequency features are combined with adaptive fusion to achieve enhanced detection of micro-defects. Based on the detection results and image quality feedback, the imaging parameters are adjusted to form a closed-loop detection. This effectively solves the problem of insufficient accuracy and stability of micro-defect detection in existing PCB micro-defect detection due to the lack of reproducibility of imaging input and the disconnect between the detection algorithm and the imaging process.

[0026] Furthermore, the preset positioning reference includes a fixed right-angle reference block that defines the orientation of the PCB board surface under test along the first direction and the second direction, and an adjustable positioning block, and the PCB board under test is held in the orientation of the board surface determined by the preset positioning reference by a clamping mechanism.

[0027] As a preferred embodiment of the above, the preset positioning reference is used to define the planar position and rotation angle of the PCB under test on the detection platform. The board surface pose refers to the position and angle state formed by the PCB under test relative to the detection platform. The first direction and the second direction correspond to the mutually perpendicular X direction and Y direction in the plane of the detection platform, respectively. The testing stage 2 is mounted on the base 1. A fixed right-angle reference block 4 is fixedly installed at a right-angle position of the testing stage 2. The two mutually perpendicular reference surfaces of the fixed right-angle reference block 4 form positioning references in the first direction and the second direction, respectively. Several T-slots 3 are opened on the surface of the testing stage 2. The movable positioning block 5 is installed in the T-slot 3 through the T-slot slider 7 and can move along the T-slot 3. The adjusted position is fixed by the locking knob 6. During positioning, the PCB board 8 is placed on the testing stage 2, so that the two adjacent sides of the PCB board 8 are respectively attached to the two reference surfaces of the fixed right-angle reference block 4. Then, the movable positioning block 5 is moved according to the size of the PCB board 8, and the PCB board 8 is limited from another direction and locked by the locking knob 6, so that PCB boards 8 of different sizes can be positioned to the same reference position. After completing the planar positioning, the downward-pressing quick clamp 9 is operated so that the rubber pressure head 10 set at the clamping end presses down against the PCB board 8, holding the PCB board 8 on the detection stage 2 and suppressing board surface warping, thereby ensuring that the positioned PCB board 8 is consistent with the pre-calibrated reference posture; the rubber pressure head 10 adopts a flexible contact method to reduce damage to the PCB board surface during the clamping process; in other embodiments, the clamping end can also use a polyurethane pressure head or a vacuum suction cup to form flexible clamping; The fixed right-angle reference block 4 provides a defined two-dimensional reference, the movable positioning block 5 adapts to PCB boards 8 of different sizes, and the downward quick clamp 9 maintains the positioning state, so that the position and angle of the PCB board 8 in the imaging field of view can be reproduced by repeated clamping, providing a stable physical positioning basis for subsequent registration and differential of board surface image and reference image.

[0028] Furthermore, acquiring the first plate surface image according to the first imaging parameters includes: The first imaging parameters are determined based on the thickness and surface reflectivity of the PCB under test, and the imaging distance between the camera and the PCB under test, as well as the lighting conditions, are adjusted based on the first imaging parameters. In an imaging environment that suppresses stray ambient light, the first board image is acquired according to the adjusted first imaging parameters so that the PCB under test of the same board type has reproducible imaging conditions.

[0029] As a preferred embodiment of the above, the first imaging parameter refers to the parameter used to determine the imaging state when the PCB board image is acquired for the first time, including the imaging height of the industrial camera 16 and the illumination parameter of the ring light source 18; the reproducible imaging condition refers to the use of the same camera height and illumination parameter in the repeated detection process of the same production line and the same board type, so that the board images obtained in different batches have a stable imaging state. After the PCB positioning and clamping is completed, the lifting slider 13 is adjusted according to the thickness of the PCB board 8 and the surface reflective characteristics. The lifting slider 13 moves along the vertical guide rail 12 and drives the horizontal camera frame 14 and the camera light source mounting plate 15 to rise and fall, so that the industrial camera 16, lens 17 and ring light source 18 mounted on the camera light source mounting plate 15 reach the corresponding calibration height, thereby determining the imaging distance between the industrial camera 16 and the PCB board 8. The ring light source 18 is turned on to illuminate the PCB board 8, and the light shield 11 surrounding the imaging area is used to isolate the stray light from the environment. Under the condition that the camera height and illumination parameters are determined, the industrial camera 16 acquires the first image of the board surface. The same camera height and illumination parameters are maintained for the same board type on the same production line, so that the imaging conditions are consistent in the repeated inspection process, providing a stable image input for subsequent image registration, differential and defect feature extraction.

[0030] Furthermore, the registration and difference between the first plate image and the pre-stored reference image include: Based on the reproducible board pose formed by the preset positioning reference, the first board image and the pre-stored reference image are sub-pixel registered to obtain the registered image. The registered image and the pre-stored reference image are differentially processed to obtain candidate defect regions that represent the suspected defect locations; The first plate image is subjected to illumination normalization and geometric correction based on a preset positioning reference, and the candidate defect region is mapped to the corrected first plate image to form a spatial prior for feature extraction.

[0031] As a preferred embodiment of the above, subpixel registration refers to correcting the residual positional deviation between the first board image and the pre-stored reference image with a positioning accuracy of less than one image pixel; the candidate defect region refers to the suspected abnormal region generated after the two images are registered and differentiated; geometric correction refers to rotating and perspective transforming the image according to the PCB board pose to unify the detection area to the preset coordinate system. Since the fixed positioning reference makes the position and angle of different PCB boards 8 in the imaging field of view highly repeatable, there is a small residual offset between the first board image and the pre-stored defect-free reference image. Sub-pixel registration is performed on the first board image and the pre-stored reference image, and the registered image is obtained after correcting the residual offset, so as to reduce the impact of board edge misalignment on subsequent differential results. The registered image is differentially divided with the pre-stored reference image. Candidate defect regions are obtained based on the differences between corresponding positions in the two images. The candidate defect regions are then used to roughly determine the location of suspected defects. The candidate defect regions serve as spatial priors for subsequent defect detection, enabling the detection network to focus on the locations where there are image differences and reduce the interference of normal regions on the identification of minor defects. The first plate image is subjected to illumination normalization processing, including illumination homogenization and highlight suppression of the imaging area, so that the brightness difference caused by reflection at different times or different plate surfaces tends to be consistent, thereby reducing the impact of illumination changes on candidate defect areas and subsequent feature extraction. The PCB board 8 is obtained by determining the edge position of the PCB board based on the fixed right-angle reference block 4. The first board image is then rotated and perspective corrected based on the board image to align the detection area to a unified coordinate system. The candidate defect area is then established to correspond to the corresponding image position in the unified coordinate system, so that the defect position obtained later can be converted into the physical position of the PCB board.

[0032] Furthermore, feature extraction is performed on the first plate image based on the candidate defect region, including: Multi-layer feature extraction and feature fusion are performed on the first plate image through the backbone network and neck network of the object detection network; After the feature fusion module of the backbone network and the neck network, the salience of each feature position relative to the neighborhood features is determined based on the parameterless attention mechanism, and the corresponding features are weighted according to the salience. Spatial priors are fused with weighted features to enhance the features of minute defects corresponding to candidate defect regions.

[0033] As a preferred embodiment of the above embodiments, in this embodiment, MicroFuse-YOLOv10 is used to complete the extraction of small defect features, and the basic detection network is YOLOv10n. MicroFuse-YOLOv10 refers to the target detection network formed by introducing small defect attention enhancement and frequency feature enhancement into the YOLOv10n basic network. The first plate image is input into the backbone network for multi-layer feature extraction, and the neck network completes the fusion of features at different scales. The spatial prior provided by the candidate defect region is used to indicate the location of suspected defects, so that the feature extraction process focuses on the corresponding region and reduces the interference of normal plate area on the features of small defects. SimAM parameterless attention modules are embedded after each C2f feature fusion module in the backbone network and the neck network. The C2f feature fusion module is used to realize the information transmission and fusion of features at different levels. The SimAM parameterless attention module is used to determine the importance of the corresponding position based on the difference between the target neuron and other neurons in the same channel in the feature map, and to reweight the features output by the C2f feature fusion module according to the energy weighting method described in the aforementioned implementation method. After processing by the SimAM parameterless attention module, the feature locations with high local mutation degree receive a strong response, so that small defect features such as mouse bites and burrs with weak difference from the background are preserved and enhanced in the subsequent feature fusion process, without introducing additional trainable weights. In other implementations, the base detection network may also employ YOLOv8, YOLOv11, or other single-stage or two-stage object detection networks, and SimAM may be set in the corresponding feature extraction or feature fusion stage; SimAM may also be replaced with SE, CBAM, or ECA lightweight attention modules.

[0034] Furthermore, adaptive weighting and residual fusion of high-frequency and low-frequency components include: Low-frequency components are extracted from the input features obtained by feature extraction. High-frequency components are determined based on the difference between input features and low-frequency components; Frequency enhancement features are obtained by adaptively weighting high-frequency and low-frequency components using learnable weights. The frequency enhancement features are residually fused with the input features to obtain the defect enhancement features.

[0035] As a preferred embodiment of the above, frequency-guided residual fusion refers to a feature enhancement method that decomposes the input features into low-frequency components that characterize the overall structural information and high-frequency components that characterize edge and texture changes, and then recombines the two types of frequency information and injects the input features back into the residual method to compensate for the information loss caused by minor defect features during network downsampling and feature fusion. In the feature fusion stage of the target detection network, an FGRF frequency-guided residual fusion module is set up. The detection feature to be enhanced is used as the input feature X. The low-frequency component X_low is extracted by average pooling with a size of 3×3, a stride of 1, and a padding of 1. The high-frequency component X_high is obtained by using the difference between the input feature X and the low-frequency component X_low. Average pooling is used to extract structural information with relatively gentle changes. The high-frequency component mainly retains the edge and texture changes corresponding to small defects such as mouse bites, burrs, and holes. According to the FGRF calculation relationship disclosed in the aforementioned implementation method, the contribution levels of high-frequency component X_high and low-frequency component X_low are adjusted by learning coefficients α and β respectively, and the weighted enhancement term E is superimposed on the input feature X in the form of residuals to obtain the defect enhancement feature X'; α and β participate in network training and adaptively adjust the fusion ratio of high and low frequency information according to the detection task, wherein α is initialized to 1 and β is initialized to 0, so that high-frequency details corresponding to small defects are retained first in the early stage of training, and then the high and low frequency combination ratio suitable for defect detection is gradually determined through training; The defect enhancement features after frequency-guided residual fusion continue to participate in subsequent feature fusion and defect prediction, so that the small defect edges and texture information that are easily weakened during downsampling are compensated, and the effective information in the original input features is retained by residual connection. In other implementations, low-frequency components can also be obtained through Gaussian filtering or wavelet transform, achieving high- and low-frequency feature separation while maintaining the relationships between low-frequency extraction, high-frequency difference acquisition, learnable weighting, and residual fusion processing.

[0036] Furthermore, such as Figure 7 As shown, when there is a domain difference between the first plate image and the training image used to train the defect detection model, it includes: Unsupervised image domain transformation is performed on the first board image to obtain a transformed image with the same image style as the training image; The transformed image is used to replace the first board image for feature extraction, so that the defect detection model can perform defect detection on PCB images of different board types or different image sources.

[0037] As a preferred embodiment of the above, domain difference refers to the distributional differences between the PCB board image under test and the training image of the defect detection model in terms of substrate color, image source, or imaging style; cross-domain front end refers to an image conversion module set before feature extraction to reduce the above-mentioned domain difference. When the substrate color or source of the first panel image is significantly different from that of the training image, a cross-domain front-end based on CycleGAN is connected before the defect detection model to convert the first panel image with heterogeneous color into a binary style image consistent with the training image. The converted image is then used as the input of the defect detection model for feature extraction and defect detection, thereby reducing the distribution difference between the input image and the training data. CycleGAN employs a generator containing 6 residual blocks, with the base number of channels for both the generator and the discriminator set to 32. It completes the transformation of the target domain image to the training domain image style through unsupervised style transfer, thereby improving the defect detection adaptability under cross-board type and cross-production line conditions without re-annotating the target domain PCB image. In practical applications, the defect detection model trained on binary style PCB images can be applied to color PCB images. First, the color PCB image is converted into the corresponding binary style image through a cross-domain front-end, and then subsequent feature extraction and defect detection are performed. The results show that introducing this cross-domain front-end can effectively reduce the distribution differences between images from different sources. In other implementations, the cross-domain front-end can also use other unsupervised style transfer methods or domain adaptation methods to achieve image domain transformation; when the domain difference between the PCB board image to be tested and the training image is small, the cross-domain front-end can be omitted and feature extraction and defect detection can be performed directly.

[0038] Furthermore, such as Figure 8 As shown, adjusting the first imaging parameters based on the first defect detection result and the image quality parameters of the first plate surface image includes: The detection confidence level is determined based on the first defect detection result, and the image quality parameters characterizing the current imaging state are determined based on the first plate image. The detection confidence and image quality parameters are compared with their respective preset thresholds, and the first panel image is determined not to meet the preset output conditions when at least one of them does not meet the corresponding preset threshold. Based on the detection confidence level and image quality parameters, at least one of the camera height and light source intensity is adjusted as a first imaging parameter, and the board surface image is re-acquired according to the adjusted first imaging parameters for defect detection until the preset output conditions are met.

[0039] As a preferred embodiment of the above, the detection confidence level is used to characterize the credibility of the defect detection model for the current detection result; the image quality parameter is used to characterize whether the board image has imaging quality suitable for the identification of minor defects, including sharpness and brightness uniformity; the preset output condition refers to the state in which both the detection confidence level and the image quality parameter reach the corresponding threshold; the first imaging parameter includes the imaging height of the industrial camera 16 and the light source intensity of the ring light source 18. After completing the defect detection of the first panel image, the detection confidence is read from the first defect detection result. At the same time, the sharpness and brightness uniformity of the first panel image are calculated and compared with the corresponding preset thresholds. When the detection confidence or any image quality parameter is lower than the corresponding threshold, it is determined that the current panel image does not meet the preset output conditions, that is, the current imaging state is not conducive to the detection of small defects. When the current board image does not meet the preset output conditions, the lifting slider 13 is finely adjusted according to the evaluation results of the detection confidence and image quality parameters, so that the industrial camera 16, lens 17 and ring light source 18 are raised and lowered synchronously relative to the PCB board 8 to change the imaging height; at the same time, the light intensity of the ring light source 18 is adjusted to form updated imaging parameters; the board image is re-acquired according to the updated imaging parameters, and image processing and defect detection are performed again to obtain the detection confidence and image quality parameters again. Repeatedly perform imaging parameter adjustment, board image acquisition and defect detection until the detection confidence and image quality parameters reach the corresponding threshold, then stop the adjustment and output the defect detection result. This allows the defect detection result to have a reverse effect on the board imaging process, forming a closed-loop feedback between imaging acquisition and defect detection. In other embodiments, image quality evaluation may also include anomaly indicators such as contrast or number of defects, and imaging parameters may also include light source angle or exposure time, and the corresponding imaging parameters may be adjusted according to the evaluation results.

[0040] Example 2; Based on the same inventive concept as the PCB micro-defect closed-loop detection method in the foregoing examples, the present invention also provides a PCB micro-defect closed-loop detection system, the system comprising: The positioning and pressing imaging module positions and presses the PCB under test based on a preset positioning reference, and acquires the first board surface image according to the first imaging parameters; The registration and difference processing module registers and differs the first plate image with the pre-stored reference image to obtain candidate defect regions; The prior feature enhancement module uses the candidate defect region as a spatial prior to extract features from the first plate image, and performs adaptive weighting and residual fusion on the high-frequency and low-frequency components in the extracted features to obtain defect enhancement features. The defect detection and identification module performs defect detection on the PCB under test based on defect enhancement features and obtains the first defect detection result. The closed-loop imaging adjustment module adjusts the first imaging parameters based on the first defect detection result and the image quality parameters of the first plate image, and re-acquires the plate image according to the adjusted first imaging parameters to perform defect detection until the preset output conditions are met and then outputs the defect detection result.

[0041] The system described above in this invention can effectively realize a closed-loop detection method for minor defects in PCBs, and the technical effects it can achieve are as described in the above embodiments, which will not be repeated here.

[0042] Example 3: Based on the same inventive concept as the PCB micro-defect closed-loop detection method in the foregoing examples, the present invention also provides a PCB micro-defect closed-loop detection device, the device including a base 1, a detection stage 2, a positioning and clamping mechanism, a light-shielding and illumination mechanism, and an imaging adjustment mechanism. The testing stage 2 is mounted on the base 1, and the testing stage 2 is provided with a T-slot 3; The positioning and clamping mechanism includes a fixed right-angle reference block 4, a movable positioning block 5, a locking knob 6, a T-shaped slider 7, and a downward-pressing quick clamp 9. The fixed right-angle reference block 4 is set at the right-angle position of the detection stage 2 and forms a first positioning reference and a second positioning reference that are perpendicular to each other. The movable positioning block 5 slides with the T-shaped slider 7 and the T-shaped groove 3 and is positioned by the locking knob 6 so that the PCB board 8 abuts against the fixed right-angle reference block 4 in two directions. A pressure-type quick clamp 9 is set on the inspection stage 2. The clamping end of the pressure-type quick clamp 9 is equipped with a rubber pressure head 10, which is used to press the positioned PCB board 8 onto the inspection stage 2. The imaging adjustment mechanism includes a vertical guide rail 12, a lifting slider 13, a horizontal camera frame 14, and a camera light source mounting plate 15. The lifting slider 13 is slidably connected to the vertical guide rail 12, the horizontal camera frame 14 is connected to the lifting slider 13, and the camera light source mounting plate 15 is disposed on the horizontal camera frame 14 so as to adjust the height of the camera light source mounting plate 15 relative to the PCB board 8 by means of the lifting slider 13. The light-shielding illumination mechanism includes a light-shielding plate 11 and a ring light source 18. The light-shielding plate 11 surrounds the imaging area of ​​the detection stage 2, and the ring light source 18 is set on the camera light source mounting plate 15 for illuminating the PCB board 8 in the light-shielding environment formed by the light-shielding plate 11. The camera light source mounting plate 15 is also provided with an industrial camera 16 and a lens 17 connected to the industrial camera 16. The industrial camera 16, lens 17 and ring light source 18 are raised and lowered synchronously with the lifting slider 13 to acquire the board surface image of the PCB board 8 under the reproducible board surface posture formed by the positioning and clamping mechanism and the controlled lighting conditions formed by the light shielding and lighting mechanism.

[0043] Specifically, such as Figure 2 As shown, this embodiment provides a PCB micro-defect detection device, including a base 1, a detection stage 2, a positioning and clamping mechanism, a light-shielding and illumination mechanism, and an imaging adjustment mechanism; wherein, the positioning and clamping mechanism is used to establish a unified board surface reference for the PCB board 8 and maintain its positioning state; the light-shielding and illumination mechanism is used to form a stable board surface illumination environment; the imaging adjustment mechanism is used to adjust the imaging height of the industrial camera 16, lens 17, and ring light source 18 relative to the PCB board 8; The testing stage 2 is fixed on the base 1, and several T-slots 3 are formed on its surface; the fixed right-angle reference block 4 is set at the right angle position of the testing stage 2, and the two mutually perpendicular reference surfaces form the X-direction positioning reference and the Y-direction positioning reference respectively; the movable positioning block 5 is installed in the T-slot 3 through the T-slide block 7 and can move along the T-slot 3; the locking knob 6 is used to fix the adjustment position of the movable positioning block 5; PCB boards 8 of different sizes can be reference positioned by abutting the fixed right-angle reference blocks 4 on both adjacent sides, and then limited from another direction by the movable positioning block 5; A downward-pressing quick clamp 9 is installed on the inspection stage 2, and a rubber pressure head 10 is provided at the pressing end. After the PCB board 8 is positioned in the plane, the downward-pressing quick clamp 9 drives the rubber pressure head 10 to press against the PCB board 8, so that the PCB board 8 fits against the inspection stage 2 and suppresses board surface warping, thereby maintaining the board surface posture determined by the fixed right-angle reference block 4 and the movable positioning block 5. The rubber pressure head 10 adopts a flexible contact form to reduce damage to the PCB board surface during the pressing process. The light shield 11 is arranged around the imaging area of ​​the detection stage 2 to block stray light from the environment; the ring light source 18 is located above the PCB board 8 to provide uniform illumination to the imaging area; the light shield 11 and the ring light source 18 together form a controlled lighting environment to reduce the impact of external light changes and board surface reflections on the stability of image acquisition. A vertical guide rail 12 is set along the height direction, a lifting slider 13 is slidably connected to the vertical guide rail 12, a horizontal camera frame 14 is connected to the lifting slider 13, and a camera light source mounting plate 15 is mounted on the horizontal camera frame 14; an industrial camera 16, a lens 17, and a ring light source 18 are mounted on the camera light source mounting plate 15, so that the three can rise and fall synchronously with the lifting slider 13; by adjusting the position of the lifting slider 13 on the vertical guide rail 12, the height of the industrial camera 16 and the ring light source 18 relative to the PCB board 8 can be changed, and the corresponding position is maintained after reaching the predetermined imaging height; During operation, the PCB board 8 is placed on the inspection stage 2, and positioning and clamping are completed by the fixed right-angle reference block 4, the movable positioning block 5, and the downward quick clamp 9. Then, the lifting slider 13 is adjusted so that the industrial camera 16, lens 17, and ring light source 18 reach the imaging height adapted to the PCB board 8. The ring light source 18 is turned on in the light-shielding environment formed by the light shield 11, and the industrial camera 16 acquires the board surface image of the PCB board 8. The fixed right-angle reference block 4 ensures that the PCB board 8 has a reproducible position and angle when repeatedly clamped, the imaging adjustment mechanism ensures that the camera height can be adjusted and locked, and the light shielding and illumination mechanism ensures that the imaging lighting conditions can be stably reproduced, thereby providing consistent image input for subsequent board surface image registration and micro-defect detection. In other embodiments, the rubber indenter 10 can be replaced with a polyurethane indenter or a flexible contact structure such as a vacuum suction cup; without changing the unified positioning reference and the adjustable imaging structure function, vacuum adsorption, pin holes or visual positioning marks can also be used to achieve reproducible positioning of the PCB board 8.

[0044] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A closed-loop detection method for minute defects in PCBs, characterized in that, The method includes: The PCB under test is positioned and pressed based on a preset positioning reference, and the first board surface image is acquired according to the first imaging parameters. The first plate image is registered and differencing with a pre-stored reference image to obtain candidate defect regions; The candidate defect region is used as a spatial prior to extract features from the first plate image. The high-frequency and low-frequency components in the extracted features are adaptively weighted and residually fused to obtain defect enhancement features. Based on the defect enhancement features, defect detection is performed on the PCB under test to obtain a first defect detection result; Based on the first defect detection result and the image quality parameters of the first board image, the first imaging parameters are adjusted, and the board image is re-acquired according to the adjusted first imaging parameters for defect detection until the preset output conditions are met and the defect detection result is output.

2. The PCB micro-defect closed-loop detection method according to claim 1, characterized in that, The preset positioning reference includes a fixed right-angle reference block that defines the orientation of the PCB board surface under test along a first direction and a second direction, and an adjustable positioning block. The PCB board under test is held in the orientation determined by the preset positioning reference by a clamping mechanism.

3. The PCB micro-defect closed-loop detection method according to claim 1, characterized in that, Acquire an image of the first plate surface according to the first imaging parameters, including: The first imaging parameters are determined based on the thickness and surface reflectivity of the PCB under test, and the imaging distance and lighting conditions between the camera and the PCB under test are adjusted based on the first imaging parameters. In an imaging environment that suppresses stray ambient light, the first board surface image is acquired according to the adjusted first imaging parameters so that the PCB under test of the same board type has reproducible imaging conditions.

4. The PCB micro-defect closed-loop detection method according to claim 1, characterized in that, The first plate image is registered and differencing with the pre-stored reference image, including: Based on the reproducible board pose formed by the preset positioning reference, the first board image and the pre-stored reference image are sub-pixel registered to obtain a registered image. The registered image and the pre-stored reference image are subjected to differential processing to obtain the candidate defect region that represents the suspected defect location; The first plate image is subjected to illumination normalization and geometric correction based on the preset positioning reference, and the candidate defect region is mapped to the corrected first plate image to form the spatial prior for feature extraction.

5. The PCB micro-defect closed-loop detection method according to claim 1, characterized in that, Feature extraction is performed on the first plate image based on candidate defect regions, including: Multi-layer feature extraction and feature fusion are performed on the first plate image through the backbone network and neck network of the target detection network; After the feature fusion module of the backbone network and the neck network, the salience of each feature position relative to the neighborhood features is determined based on the parameterless attention mechanism, and the corresponding features are weighted according to the salience. The spatial prior is fused with the weighted features to enhance the micro-defect features corresponding to the candidate defect region.

6. The PCB micro-defect closed-loop detection method according to claim 1, characterized in that, Adaptive weighting and residual fusion of high-frequency and low-frequency components, including: The low-frequency components are obtained by performing low-frequency extraction on the input features obtained from the feature extraction. The high-frequency component is determined based on the difference between the input feature and the low-frequency component; Frequency enhancement features are obtained by adaptively weighting the high-frequency components and the low-frequency components using learnable weights. The frequency enhancement feature is residually fused with the input feature to obtain the defect enhancement feature.

7. The PCB micro-defect closed-loop detection method according to claim 1, characterized in that, When there is a domain difference between the first plate image and the training image used to train the defect detection model, including: Unsupervised image domain transformation is performed on the first board image to obtain a transformed image with a corresponding image style to the training image; The transformed image is used to replace the first board image for feature extraction, so that the defect detection model can perform defect detection on PCB images of different board types or different image sources.

8. The PCB micro-defect closed-loop detection method according to claim 1, characterized in that, Adjusting the first imaging parameters based on the first defect detection result and the image quality parameters of the first plate surface image, including: The detection confidence level is determined based on the first defect detection result, and the image quality parameters characterizing the current imaging state are determined based on the first plate image. The detection confidence and the image quality parameters are compared with their respective preset thresholds, and the first panel image is determined not to meet the preset output conditions when at least one of them does not meet the corresponding preset threshold. Based on the detection confidence level and the image quality parameters, at least one of the first imaging parameters, namely camera height and light source intensity, is adjusted, and the board surface image is re-acquired according to the adjusted first imaging parameters for defect detection until the preset output conditions are met.

9. A closed-loop detection system for minute defects in PCBs, characterized in that, The system includes: The positioning and pressing imaging module positions and presses the PCB under test based on a preset positioning reference, and acquires the first board surface image according to the first imaging parameters; The registration and difference processing module registers and differs the first plate image with the pre-stored reference image to obtain candidate defect regions; The prior feature enhancement module uses the candidate defect region as a spatial prior to extract features from the first plate image, and performs adaptive weighting and residual fusion on the high-frequency and low-frequency components in the extracted features to obtain defect enhancement features. The defect detection and identification module performs defect detection on the PCB under test based on defect enhancement features and obtains the first defect detection result. The closed-loop imaging adjustment module adjusts the first imaging parameters based on the first defect detection result and the image quality parameters of the first plate image, and re-acquires the plate image according to the adjusted first imaging parameters to perform defect detection until the preset output conditions are met and then outputs the defect detection result.

10. A closed-loop detection device for minute defects in PCBs, characterized in that, The device includes a base (1), a detection stage (2), a positioning and clamping mechanism, a light-shielding and illumination mechanism, and an imaging adjustment mechanism. The testing platform (2) is disposed on the base (1), and a T-slot (3) is provided on the testing platform (2). The positioning and clamping mechanism includes a fixed right-angle reference block (4), a movable positioning block (5), a locking knob (6), a T-shaped slider (7), and a downward-pressing quick clamp (9). The fixed right-angle reference block (4) is set at the right angle position of the detection platform (2) and forms a first positioning reference and a second positioning reference that are perpendicular to each other. The movable positioning block (5) slides with the T-shaped slider (7) and the T-shaped groove (3) and is positioned by the locking knob (6) so that the PCB board (8) abuts against the fixed right-angle reference block (4) in two directions. The pressing quick clamp (9) is set on the detection platform (2), and the pressing end of the pressing quick clamp (9) is provided with a rubber pressure head (10) for pressing the positioned PCB board (8) onto the detection platform (2). The imaging adjustment mechanism includes a vertical guide rail (12), a lifting slider (13), a horizontal camera frame (14), and a camera light source mounting plate (15). The lifting slider (13) is slidably connected to the vertical guide rail (12), the horizontal camera frame (14) is connected to the lifting slider (13), and the camera light source mounting plate (15) is disposed on the horizontal camera frame (14) to adjust the height of the camera light source mounting plate (15) relative to the PCB board (8) by means of the lifting slider (13). The light-shielding illumination mechanism includes a light-shielding plate (11) and a ring light source (18). The light-shielding plate (11) surrounds the imaging area of ​​the detection stage (2), and the ring light source (18) is disposed on the camera light source mounting plate (15) for illuminating the PCB board (8) under the light-shielding environment formed by the light-shielding plate (11). The camera light source mounting plate (15) is also provided with an industrial camera (16) and a lens (17) connected to the industrial camera (16). The industrial camera (16), the lens (17) and the ring light source (18) move up and down synchronously with the lifting slider (13) to acquire the board surface image of the PCB board (8) under the conditions of the reproducible board surface posture formed by the positioning and pressing mechanism and the controlled lighting formed by the light shielding and lighting mechanism.

Images