Defect detection method and system for power card cap welds

Abstract

This invention relates to the field of defect detection technology, and discloses a method and system for detecting defects in wave solder joints on computing power cards. The method includes: acquiring a high-resolution image sequence of the wave solder joint area on a computing power card PCBA assembly; optimizing the high-resolution image sequence to obtain a denoised and enhanced image sequence of the wave solder joint area; performing region segmentation on the denoised and enhanced image sequence to obtain a preprocessed solder joint sequence of the wave solder joint area; extracting features from the preprocessed solder joint sequence to obtain the three-dimensional morphological features of the wave solder joint area; performing defect detection on the three-dimensional morphological features of the wave solder joint area to obtain the defect type of the wave solder joint area; and spatially mapping the layout diagram of the computing power card PCBA assembly based on the defect type and the location information of the wave solder joint area to obtain a defect distribution heatmap of the wave solder joint area. This invention provides a complete data foundation for intelligent defect judgment and significantly improves the accuracy and reliability of defect identification.

Description

Technical Field

[0001] This invention relates to the field of defect detection technology, and in particular to a method and system for detecting defects in wave solder joints of computing power cards. Background Technology

[0002] In the field of defect detection, computing cards, as core components of high-performance computing, typically integrate a large number of high-density, high-power chips in their PCBA components, requiring extremely high reliability in soldering quality. Traditional manual visual inspection methods are not only inefficient and labor-intensive, but also susceptible to subjective factors such as personnel experience and fatigue, resulting in poor consistency and a high rate of missed detections. These methods can no longer meet the stringent quality control requirements of modern electronic manufacturing.

[0003] Existing two-dimensional automated optical inspection (AOI) technology analyzes the features of solder joints by acquiring surface images. While it can effectively detect surface defects such as misalignment and missing solder, it is essentially based on two-dimensional image analysis of color and texture, making it difficult to accurately identify three-dimensional morphological defects related to height and volume. To overcome the limitations of two-dimensional inspection, the industry has gradually introduced three-dimensional inspection technology. Technologies such as structured light three-dimensional reconstruction have greatly improved inspection efficiency. However, traditional application solutions often focus on extracting single height information or using three-dimensional data only for simple threshold judgment, lacking in-depth mining and fusion analysis of the geometric features of the solder joint surface. Summary of the Invention

[0004] This invention provides a method and system for detecting defects in wave solder joints of computing cards to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method for detecting defects in wave solder joints on computing power cards, comprising:

[0006] S1. Acquire high-resolution image sequences of the wave solder joint area on the PCBA assembly of the computing power card;

[0007] S2. Perform image optimization on the high-resolution image sequence to obtain a denoised and enhanced image sequence of the wave peak solder joint region;

[0008] S3. Perform region segmentation on the denoised and enhanced image sequence to obtain the preprocessed solder joint sequence of the peak solder joint region;

[0009] S4. Extract features from the preprocessed solder joint sequence to obtain the three-dimensional morphological features of the wave peak solder joint region;

[0010] S5. Perform defect detection on the three-dimensional morphological features of the wave solder joint area to obtain the defect type of the wave solder joint area;

[0011] S6. Based on the defect type and the location information of the wave solder joint area, perform spatial mapping on the layout diagram of the computing card PCBA component to obtain a defect distribution heat map of the wave solder joint area.

[0012] In a preferred embodiment, the acquisition of high-resolution image sequences of the wave solder joint area on the computing power card PCBA assembly includes:

[0013] Obtain an image of the internal morphological features of the wave solder joint area in the computing card PCBA assembly;

[0014] Obtain the deformed grating image sequence of the wave crest solder joint region;

[0015] Pixel fusion is performed on the internal morphology feature image and the deformed grating image sequence to obtain a high-resolution image sequence of the wave crest solder joint region.

[0016] In a preferred embodiment, the step of image optimization of the high-resolution image sequence to obtain a denoised and enhanced image sequence of the wave crest solder joint region includes:

[0017] Image enhancement is performed on the high-resolution image sequence to obtain a contrast-enhanced image sequence of the wave peak solder joint region;

[0018] The contrast-enhanced image sequence is subjected to noise removal to obtain the denoised enhanced image sequence of the wave peak solder joint region.

[0019] In a preferred embodiment, the step of performing region segmentation on the denoised and enhanced image sequence to obtain the preprocessed solder joint sequence of the peak solder joint region includes:

[0020] The denoised and enhanced image sequence is binarized based on a predetermined color threshold range to obtain a preliminary binary mask sequence for the wave peak solder joint region.

[0021] The initial binary mask sequence is subjected to dilatational erosion to obtain the optimized binary mask sequence for the wave crest solder joint region;

[0022] Based on the optimized binary mask sequence, solder joint targets are extracted from the denoised and enhanced image sequence to obtain the preprocessed solder joint sequence of the peak solder joint region.

[0023] In a preferred embodiment, the step of extracting features from the preprocessed solder joint sequence to obtain the three-dimensional morphological features of the wave peak solder joint region includes:

[0024] Based on the preprocessed solder joint sequence, the relative height difference of different regions in the wave crest solder joint region is analyzed to obtain the height distribution data of the wave crest solder joint region.

[0025] Based on the height distribution data, the surface curvature state of different regions in the wave crest solder joint area is analyzed to obtain the curvature distribution data of the wave crest solder joint area.

[0026] Based on the height distribution data and the curvature distribution data, spatial volume analysis is performed on the wave crest solder joint region to obtain the volume quantification data of the wave crest solder joint region.

[0027] The height distribution data, the curvature distribution data, and the volume quantization data are integrated into the three-dimensional morphological features of the wave crest solder joint region.

[0028] In a preferred embodiment, the defect detection of the three-dimensional morphological features of the wave solder joint region to obtain the defect type of the wave solder joint region includes:

[0029] Based on predetermined quantitative indicators, a defect confidence analysis is performed on the three-dimensional morphological features to obtain the confidence level of the defect type in the wave crest solder joint region.

[0030] The type of defect in the wave solder joint region is determined based on the confidence level.

[0031] In a preferred embodiment, the predetermined quantification index includes:

[0032] A range analysis was performed on the height distribution characteristics of the three-dimensional morphology features to obtain the height deviation ratio index of the wave crest solder joint region.

[0033] The volume quantization features of the three-dimensional morphology are compared with standard values ​​to obtain the volume fill ratio index of the wave crest solder joint region.

[0034] An abnormal region proportion analysis is performed on the curvature distribution characteristics of the three-dimensional morphology features to obtain the curvature abnormality proportion index of the wave crest solder joint region.

[0035] A safety spacing analysis is performed on the solder joint spacing characteristics of the three-dimensional morphology features to obtain the spacing ratio index of the wave crest solder joint region.

[0036] The height deviation ratio, volume filling ratio, curvature anomaly ratio, and spacing ratio are integrated to obtain a predetermined quantitative index for the wave crest solder joint region.

[0037] In a preferred embodiment, the formula for calculating the confidence level of the defect type in the wave solder joint region is as follows:

[0038]

[0039] in, The solder joint belongs to the first Confidence level for each defect type For logical functions, To indicate the first The aforementioned quantitative indicators, For the first The first defect type corresponds to the first The preset threshold of the aforementioned quantitative indicator, For the first The aforementioned quantitative indicators are used to determine the first... The weighting coefficients for the aforementioned defect types, This indicates a weighted summation of all the aforementioned quantitative indicators.

[0040] In a preferred embodiment, the step of spatially mapping the layout diagram of the computing card PCBA component based on the defect type and the location information of the wave solder joint area to obtain a defect distribution heatmap of the wave solder joint area includes:

[0041] Obtain the coordinates of the defect location in the wave solder joint region;

[0042] Based on the defect type, the defect location coordinates are classified and aggregated to obtain the aggregated defect location coordinates of the wave solder joint region;

[0043] Based on the coordinates of the aggregated defects, a defect density distribution analysis is performed on the layout diagram of the computing card PCBA component to obtain the defect density value of the wave solder joint area.

[0044] Color rendering is applied to the defect density value to obtain a heat map of defect distribution in the wave crest solder joint region.

[0045] To address the above problems, this invention also provides a wave solder joint defect detection system for computing power cards, the system comprising:

[0046] Image acquisition module: used to acquire high-resolution image sequences of the wave solder joint area on the computing card PCBA assembly;

[0047] Image processing module: used to optimize the high-resolution image sequence to obtain a denoised and enhanced image sequence of the wave peak solder joint region;

[0048] Region segmentation module: used to perform region segmentation on the denoised and enhanced image sequence to obtain the preprocessed solder joint sequence of the peak solder joint region;

[0049] 3D feature analysis module: used to extract features from the preprocessed solder joint sequence to obtain the 3D morphological features of the wave peak solder joint region;

[0050] Identification and classification module: used to perform defect detection on the three-dimensional morphological features of the wave solder joint area to obtain the defect type of the wave solder joint area;

[0051] Visualization output module: Based on the defect type and the location information of the wave solder joint area, it performs spatial mapping on the layout diagram of the computing card PCBA component to obtain a heat map of defect distribution in the wave solder joint area.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. This invention establishes a confidence-based defect decision-making model and a spatial mapping visualization mechanism, enabling quantifiable evaluation of detection results and process traceability. An adaptive weighted algorithm is used to fuse multi-dimensional quantitative indicators, and the probabilistic confidence level of defect types is output through a sigmoid function. This supports automatic judgment under high confidence conditions and also includes a manual review mechanism for suspected defects, improving detection efficiency while ensuring the rigor of decision-making. The defect distribution heatmap generated based on kernel density estimation can intuitively present the aggregation pattern of different defect types on the PCBA board, providing data support for optimizing process parameters such as wave soldering temperature, flux usage, and guide rail speed.

[0054] 2. This invention overcomes the limitations of traditional two-dimensional inspection methods by employing multimodal optical acquisition and three-dimensional topography reconstruction technology. By combining structured light coded projection with phase-shifting three-dimensional reconstruction, it achieves precise quantification of the surface height, curvature, and volume parameters of wave crest solder joints, effectively solving the technical challenge of manual visual inspection and traditional AOI systems failing to identify three-dimensional defects such as cold solder joints and insufficient solder. By extracting multi-dimensional feature vectors such as the height distribution matrix, curvature gradient map, and volume quantization values, a state space capable of comprehensively characterizing solder joint quality is constructed, providing a complete data foundation for intelligent defect judgment. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating a method for detecting defects in wave solder joints on a computing card according to an embodiment of the present invention.

[0056] Figure 2 This is a functional block diagram of a computing power card wave solder joint defect detection system provided in an embodiment of the present invention;

[0057] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0058] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0059] This application provides a method for detecting defects in wave solder joints on computing power cards. The execution subject of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for detecting defects in wave solder joints on computing power cards can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0060] Reference Figure 1 The diagram shown is a flowchart illustrating a method for detecting defects in wave solder joints on a computing card according to an embodiment of the present invention. In this embodiment, the method for detecting defects in wave solder joints on a computing card includes:

[0061] S1. Acquire high-resolution image sequences of the wave solder joint area on the PCBA assembly of the computing power card;

[0062] In this embodiment of the invention, the acquisition of high-resolution image sequences of the wave solder joint area on the computing power card PCBA component includes:

[0063] Obtain an image of the internal morphological features of the wave solder joint area in the computing card PCBA assembly;

[0064] Obtain the deformed grating image sequence of the wave crest solder joint region;

[0065] Pixel fusion is performed on the internal morphology feature image and the deformed grating image sequence to obtain a high-resolution image sequence of the wave crest solder joint region.

[0066] Specifically, internal morphological features are image features that reflect the internal material structure and microstructure of the solder joint, mainly including: lattice structure features, porosity defect features, bonding interface features, and material uniformity features.

[0067] Furthermore, the internal morphological features of the solder joint area in the computing card PCBA assembly are obtained using near-infrared spectral imaging technology. A near-infrared LED array light source with a wavelength of 850nm-950nm is used to uniformly illuminate the solder joint area. This wavelength band can effectively penetrate the oxide layer on the solder surface and reduce surface specular reflection interference.

[0068] Furthermore, a high-resolution CMOS industrial camera (12 megapixels) equipped with a narrow-band infrared filter was used to capture reflection images of the solder joints in this specific spectral band. These images clearly reveal the internal lattice structure, porosity defects, and microscopic morphological features of the solder-pad interface within the solder joints, providing crucial material and texture information for subsequent defect analysis.

[0069] In summary, images of internal morphology acquired through near-infrared spectroscopy imaging can effectively reveal the internal structure and hidden defects of solder joints, compensating for the limitations of visible light imaging in detecting internal defects and providing crucial material and texture information for a comprehensive assessment of solder joint quality. Analyzing these features allows for the early detection of potential reliability issues, enabling a holistic evaluation of solder joint quality.

[0070] Specifically, the deformable grating image sequence is a series of deformable images formed by projecting a pre-designed grating pattern onto the surface of the wave crest solder joint using structured light projection technology. The three-dimensional morphology of the solder joint surface, such as height undulations and surface curvature, geometrically modulates the grating stripes.

[0071] Furthermore, a sequence of deformed grating images of the solder joint region at the peak is obtained using structured light phase measurement technology. A digital micromirror device (DMD) projector is used, employing a four-step phase-shifting method to project a set of phase-encoded sinusoidal grating patterns onto the same solder joint region, with phase shifts of 0, π / 2, π, and 3π / 2, respectively. A synchronously triggered high-speed industrial camera, with a resolution of 5 megapixels and a frame rate of 150fps, precisely acquires a sequence of striped images showing geometric deformation after modulation of the three-dimensional morphology of the solder joint surface with microsecond-level exposure times.

[0072] This sequence fully records the phase modulation information of the grating fringes on the weld joint surface, providing the necessary phase data foundation for high-precision 3D reconstruction.

[0073] Furthermore, the four-step phase-shifting method involves projecting four sinusoidal grating patterns with phase shifts of 0, π / 2, π, and 3π / 2 in sequence, and using phase difference calculations to eliminate interference from background light and surface reflectivity differences, thereby demodulating the true phase distribution of the solder joint surface.

[0074] In summary, the deformable grating image sequence obtained through structured light phase measurement technology provides accurate phase modulation information, laying the data foundation for subsequent three-dimensional topography reconstruction based on the phase shift method and ensuring that the height measurement accuracy reaches the micrometer level.

[0075] In summary, the four-step phase-shifting method can effectively suppress the impact of illumination fluctuations and sensor noise on phase extraction through redundant calculation of multi-frame data, and is especially suitable for electromagnetic interference scenarios in industrial environments.

[0076] Specifically, a high-resolution image sequence is an image sequence that combines high spatial resolution and rich information dimensions, formed by fusing near-infrared internal topography images with deformable grating image sequences.

[0077] Furthermore, the two types of image data are fused using multi-source image fusion technology. The internal morphology feature image and the deformed grating image are spatially registered with sub-pixel precision to eliminate geometric deviations caused by differences in imaging perspective.

[0078] The registered images are then fused in the transform domain to preserve the material texture details in the near-infrared image while enhancing the phase feature edges in the deformable grating image. The result is a high-resolution image sequence with rich texture details and accurate phase information, achieving a resolution of 2448×2048 pixels and a bit depth of 12 bits.

[0079] In summary, by performing pixel-level fusion processing on the two types of image data, the complementary advantages of material texture information and phase geometry information are achieved, generating a high-resolution image sequence with more complete information dimensions.

[0080] In summary, the acquired high-resolution multimodal image sequences provide high-quality data input for subsequent image optimization, region segmentation, and 3D feature extraction, which is the primary guarantee for the entire detection method to achieve high-precision defect identification.

[0081] In summary, this step, through the synergistic application of multimodal optical imaging technology, not only obtained the geometric morphology information of the solder joint surface, but also acquired the internal material characteristics, providing a rich and accurate data foundation for a comprehensive evaluation of solder joint quality.

[0082] S2. Perform image optimization on the high-resolution image sequence to obtain a denoised and enhanced image sequence of the wave peak solder joint region;

[0083] In this embodiment of the invention, the step of optimizing the high-resolution image sequence to obtain a denoised and enhanced image sequence of the wave peak solder joint region includes:

[0084] Image enhancement is performed on the high-resolution image sequence to obtain a contrast-enhanced image sequence of the wave peak solder joint region;

[0085] The contrast-enhanced image sequence is subjected to noise removal to obtain the denoised enhanced image sequence of the wave peak solder joint region.

[0086] Specifically, the contrast-enhanced image sequence is an image sequence obtained by optimizing a high-resolution image sequence through local adaptive contrast enhancement technology.

[0087] Furthermore, histogram equalization is employed, dividing the image into 8×8 pixel local regions. A histogram is calculated within each region, and a clipping limit of 0.03 is set, effectively suppressing local over-enhancement. Bilinear interpolation is used to eliminate block artifacts, significantly improving the contrast difference between solder joints and the PCBA background. This results in a marked enhancement of texture details in solder joint outlines, solder distribution, and potential defect areas, generating a contrast-optimized image sequence.

[0088] In summary, the adaptive contrast enhancement technology effectively solves the problem of insufficient local contrast caused by uneven lighting and surface reflection, significantly improving the clarity of the boundary between the solder joint and the background, and laying the foundation for subsequent region segmentation.

[0089] Specifically, the denoising and enhancement image sequence is a high-quality image sequence obtained by eliminating various noise interferences through a hybrid filtering strategy based on the contrast enhancement image sequence.

[0090] Furthermore, noise reduction is performed on the contrast-enhanced image sequence using a hybrid filtering strategy. A two-stage filtering scheme is employed: firstly, a 5×5 pixel window value filter is used to effectively eliminate salt-and-pepper noise and impulse interference in the image.

[0091] Subsequently, a Gaussian low-pass filter with a standard deviation of 8 is applied for smoothing, suppressing Gaussian noise while preserving solder joint edges and detailed features. This hybrid filtering scheme demonstrates excellent performance in improving the signal-to-noise ratio (SNR), increasing the SNR of the processed image sequence by more than 15 dB, providing high-quality input for subsequent accurate segmentation.

[0092] Furthermore, the processing results were validated using an image quality assessment mechanism. Peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) were used as quantitative indicators to evaluate the quality of the denoised and enhanced image sequence. The PSNR value was ensured to be greater than 40 dB and the SSIM value greater than 0.95 to guarantee that the optimized image maintains structural information while possessing excellent visual quality.

[0093] In summary, the innovative application of the hybrid filtering strategy effectively suppressed various types of image noise while successfully preserving the key edge and texture features of the solder joints, avoiding the detail loss problem caused by traditional single filtering methods.

[0094] In summary, through rigorous image quality assessment, the optimized image sequence is ensured to meet both visual discrimination requirements and the quantitative standards for subsequent algorithm processing, thus achieving a balance between subjective quality and objective indicators.

[0095] In summary, the high-quality denoised and enhanced image sequence obtained through this image optimization process provides a reliable data foundation for subsequent region segmentation and feature extraction, directly affecting the accuracy and reliability of the entire detection system.

[0096] In summary, this step, through advanced image processing technology and rigorous quality control, significantly improves the quality of the original image, providing clear, complete, and reliable image data for subsequent processing steps. It is a crucial link in the entire detection process.

[0097] S3. Perform region segmentation on the denoised and enhanced image sequence to obtain the preprocessed solder joint sequence of the peak solder joint region;

[0098] In this embodiment of the invention, the step of performing region segmentation on the denoised and enhanced image sequence to obtain the preprocessed solder joint sequence of the peak solder joint region includes:

[0099] The denoised and enhanced image sequence is binarized based on a predetermined color threshold range to obtain a preliminary binary mask sequence for the wave peak solder joint region.

[0100] The initial binary mask sequence is subjected to dilatational erosion to obtain the optimized binary mask sequence for the wave crest solder joint region;

[0101] Based on the optimized binary mask sequence, solder joint targets are extracted from the denoised and enhanced image sequence to obtain the preprocessed solder joint sequence of the peak solder joint region.

[0102] Specifically, the predetermined color threshold range is the essential difference between the solder joint and the PCBA substrate in a specific color space, and is a pre-set quantitative range used to distinguish the solder joint target from the background area.

[0103] Specifically, after the initial binary mask sequence is used to denoise and enhance the image sequence, the resulting image sequence contains only black and white pixels. White pixels represent areas identified as solder joint targets, while black pixels represent background areas, including the substrate, circuitry, and noise.

[0104] Furthermore, the denoised and enhanced image sequence is binarized using color space conversion and threshold segmentation techniques. The image is converted from the RGB color space to the HSV color space, effectively decoupling brightness and color information. Based on the significant differences in hue characteristics between solder joints and the PCBA substrate, a precise threshold range is set as H∈[0,50], S∈[50,255], V∈[50,255]. This range can effectively distinguish between silver-white solder joints, green pads, and dark backgrounds. Dynamic binarization of the brightness component V generates a preliminary binary mask sequence, where the solder joint area is the foreground, set to white with a pixel value of 255, and the background area is the background, set to black with a pixel value of 0. The initial segmentation accuracy can reach over 90%.

[0105] In summary, by using HSV color space conversion and adaptive threshold segmentation, the effects of lighting changes and background interference on the segmentation results were effectively overcome, achieving preliminary and accurate separation of the solder joint area and providing a reliable binary template for subsequent processing.

[0106] Specifically, the optimized binary mask sequence is obtained by performing morphological closing operations on the initial binary mask sequence, followed by dilation, erosion, and bridging separation, resulting in a binary mask sequence with smooth boundaries, complete interior, and independent solder joints.

[0107] Furthermore, the initial binary mask sequence is optimized using morphological closing operations. A 3×3 rectangular structuring element is used, and an expansion operation is first performed to fill the small holes inside the solder joints caused by uneven lighting and to connect the weak edges interrupted by noise.

[0108] Subsequently, an erosion operation is performed to restore the original size of the solder joint and smooth irregular edges. This closing operation effectively eliminates noise points with an area smaller than 10 pixels, solves the over-segmentation and under-segmentation problems in the initial segmentation, and significantly improves the integrity and boundary smoothness of the solder joint area.

[0109] In summary, morphological closing operations significantly improved the quality of the binary mask, solved the problems of edge burrs and internal voids in the initial segmentation, and made the morphological features of the solder joint area more complete and clear.

[0110] Specifically, the preprocessed solder joint sequence is a standardized image block sequence extracted from the denoised and enhanced image based on the optimized binary mask sequence, containing independent solder joint targets and spatial information.

[0111] Furthermore, solder joint target extraction is achieved through mask application and connected component analysis techniques. The optimized binary mask is subjected to a pixel-by-pixel logical AND operation with the original denoised and enhanced image sequence to accurately extract the image content of each solder joint region. An 8-adjacency-based connected component labeling algorithm is used to identify and separate independent solder joints in each binary mask frame. Simultaneously, the geometric features such as area, perimeter, and centroid of each connected component are calculated, automatically filtering out false targets with abnormal areas, such as dust and reflective points. Finally, an image block sequence containing only complete solder joint targets is generated, with each image block having a uniform size of 64×64 pixels, ensuring data standardization for subsequent processing.

[0112] In summary, by applying masks and analyzing connected components, we were able to accurately extract individual solder joint targets from complex backgrounds and effectively filter out non-target interference, ensuring the data purity for subsequent feature extraction and target analysis.

[0113] In summary, the preprocessed weld point sequence obtained after this region segmentation process provides high-quality and standardized input for subsequent 3D feature extraction and defect identification, directly affecting the detection accuracy of the entire system.

[0114] In summary, this step, through the organic combination of color space analysis, morphological processing, and connected component analysis, achieves precise segmentation and extraction of the solder joint area, laying a solid foundation for subsequent feature analysis and defect detection. It is a crucial link in the entire detection process.

[0115] S4. Extract features from the preprocessed solder joint sequence to obtain the three-dimensional morphological features of the wave peak solder joint region;

[0116] In this embodiment of the invention, the step of extracting features from the preprocessed solder joint sequence to obtain the three-dimensional morphological features of the wave peak solder joint region includes:

[0117] Based on the preprocessed solder joint sequence, the relative height difference of different regions in the wave crest solder joint region is analyzed to obtain the height distribution data of the wave crest solder joint region.

[0118] Based on the height distribution data, the surface curvature state of different regions in the wave crest solder joint area is analyzed to obtain the curvature distribution data of the wave crest solder joint area.

[0119] Based on the height distribution data and the curvature distribution data, spatial volume analysis is performed on the wave crest solder joint region to obtain the volume quantification data of the wave crest solder joint region.

[0120] The height distribution data, the curvature distribution data, and the volume quantization data are integrated into the three-dimensional morphological features of the wave crest solder joint region.

[0121] Specifically, the height distribution data is obtained by using structured light phase-height mapping technology to convert the phase information corresponding to the solder joints after preprocessing into three-dimensional relative height quantization data with the PCBA pads as the reference surface.

[0122] Furthermore, using phase-shifting phase-height mapping technology, the structured light modulation phase information corresponding to the preprocessed solder joint sequence is converted into high-precision three-dimensional relative height data. This process first performs precise phase calculation on each independent solder joint region based on the acquired multi-frame deformable grating image sequence: using phase difference calculation to effectively separate and eliminate ambient light and background noise interference, a preliminary periodic wrapping phase reflecting the surface undulations of the solder joint is obtained.

[0123] Furthermore, taking the continuity of phase distribution and gradient consistency as key quality evaluation criteria, a phase unrolling algorithm is used to eliminate jump points in the truncated phase, reconstructing a continuous and monotonically changing absolute phase field, thereby fully restoring the accurate phase value of each pixel.

[0124] Furthermore, based on this, by using pre-calibrated vision system parameters, including the inherent parameters of the imaging and projection units and their spatial pose relationships, and strictly adhering to the principles of optical triangulation, a mapping model from absolute phase values ​​to three-dimensional spatial coordinates is established. This transforms the two-dimensional phase distribution into three-dimensional height information for each point on the solder joint surface. Ultimately, a height distribution data matrix accurately reflects the actual shape of the solder joint, possessing micron-level vertical resolution. This matrix precisely characterizes the height differences of various locations on the solder joint surface relative to the reference plane, providing a reliable data foundation for subsequent morphology analysis and defect identification.

[0125] In summary, the height distribution data obtained by phase-shifting 3D reconstruction technology accurately characterizes the macroscopic morphology of the weld joints, providing a quantitative basis for detecting highly correlated defects such as bridging and warping. The measurement accuracy is far higher than that of traditional laser scanning.

[0126] Specifically, the curvature distribution data is quantitative data describing the degree of curvature at each point on the weld joint surface, obtained through differential geometric operations based on the height distribution data.

[0127] Furthermore, the curvature distribution data of the wave crest solder joint region is obtained through differential geometric operations. Based on the obtained height distribution data and differential geometric principles, the Gaussian curvature of each point on the solder joint surface is obtained. Through curvature value analysis, features such as concave and convex regions and edge transition zones on the solder joint surface can be accurately identified.

[0128] In summary, the curvature distribution data obtained through differential geometric operations provides a detailed description of the microscopic geometric characteristics of the solder joint surface, effectively identifying local morphological anomalies such as solder balls and solder spikes, and significantly improving the detection sensitivity of surface defects.

[0129] Specifically, the volume quantization data is the total volume quantization value of the actual solder filling of the solder joint, calculated using voxel integration technology with the solder pad as the reference plane.

[0130] Furthermore, volumetric quantization data of the wave-crest solder joint region is obtained through voxel integration. Using the solder pad surface as a reference plane, the solder joint region is discretized into a regular voxel mesh. Based on height and curvature distribution data, volume integration is performed on each voxel unit, with the integration region covering the entire projected area of ​​the solder joint, thus obtaining the volumetric quantization data of the wave-crest solder joint region. This effectively detects volume-related defects such as cold solder joints and insufficient solder.

[0131] In summary, the volume quantification data calculated using the voxel integration method contains three key parameters: actual volume, theoretical volume, and volume deviation rate. These parameters directly reflect whether the solder filling is sufficient. It objectively reflects the solder filling status and provides direct evidence for judging process defects such as cold solder joints and insufficient solder.

[0132] Specifically, the three-dimensional morphological features are standardized feature vectors with strong discriminative power, which are integrated from height distribution data, curvature distribution data, and volume quantization data through multi-feature fusion and dimensionality reduction techniques.

[0133] Furthermore, a complete three-dimensional morphology feature vector is constructed using multi-feature fusion technology. Height distribution data, curvature distribution data, and volume quantization data are standardized to eliminate the influence of dimensions. Principal component analysis (PCA) is used to reduce the dimensionality of high-dimensional features, retaining more than 95% of the feature variance. The final generated feature vector contains 12 of the most discriminative feature parameters, including height range, average curvature, and volume ratio, forming a complete digital description of the three-dimensional morphology of the weld point.

[0134] In summary, by using 3D reconstruction and geometric calculations, 2D image information is transformed into quantifiable 3D morphological feature parameters such as height, curvature, and volume, thus realizing a digital description of the weld quality.

[0135] In summary, the three-dimensional morphological feature vector constructed through multi-feature fusion comprehensively characterizes the geometric properties of the weld joint, forming a comprehensive quality assessment system and providing 120-dimensional feature input for subsequent intelligent classification.

[0136] In summary, this step, through advanced 3D reconstruction and geometric analysis techniques, achieves accurate conversion from 2D images to 3D morphological features, providing rich and reliable quantitative data for weld joint quality assessment, which is the core technological breakthrough of this invention.

[0137] S5. Perform defect detection on the three-dimensional morphological features of the wave solder joint area to obtain the defect type of the wave solder joint area;

[0138] In this embodiment of the invention, the defect detection of the three-dimensional morphological features of the wave solder joint region to obtain the defect type of the wave solder joint region includes:

[0139] Based on predetermined quantitative indicators, a defect confidence analysis is performed on the three-dimensional morphological features to obtain the confidence level of the defect type in the wave crest solder joint region.

[0140] The type of defect in the wave solder joint region is determined based on the confidence level.

[0141] The predetermined quantitative indicators for judgment include:

[0142] A range analysis was performed on the height distribution characteristics of the three-dimensional morphology features to obtain the height deviation ratio index of the wave crest solder joint region.

[0143] The volume quantization features of the three-dimensional morphology are compared with standard values ​​to obtain the volume fill ratio index of the wave crest solder joint region.

[0144] An abnormal region proportion analysis is performed on the curvature distribution characteristics of the three-dimensional morphology features to obtain the curvature abnormality proportion index of the wave crest solder joint region.

[0145] A safety spacing analysis is performed on the solder joint spacing characteristics of the three-dimensional morphology features to obtain the spacing ratio index of the wave crest solder joint region.

[0146] The height deviation ratio, volume filling ratio, curvature anomaly ratio, and spacing ratio are integrated to obtain a predetermined quantitative index for the wave crest solder joint region.

[0147] The formula for calculating the confidence level of the defect type in the wave solder joint region is as follows:

[0148]

[0149] in, The solder joint belongs to the first Confidence level for each defect type For logical functions, To indicate the first The aforementioned quantitative indicators, For the first The first defect type corresponds to the first The preset threshold of the aforementioned quantitative indicator, For the first The aforementioned quantitative indicators are used to determine the first... The weighting coefficients for the aforementioned defect types, This indicates a weighted summation of all the aforementioned quantitative indicators.

[0150] Specifically, the predetermined quantitative indicators are a set of multi-dimensional quantitative parameters designed for different defect types based on three-dimensional morphological features and can be directly used for defect judgment. These include height deviation ratio, volume filling ratio, curvature anomaly ratio, and spacing ratio.

[0151] Specifically, the confidence level of the defect type is the probability value calculated by the logistic regression model that the solder joint belongs to a certain type of defect or is acceptable, with a value ranging from [0,1]. The closer the value is to 1, the higher the probability of that defect type.

[0152] Furthermore, based on predetermined quantitative indicators, the confidence level of the defect type is calculated using a weighted fusion and logistic regression model. The Sigmoid function is used as the activation function for logistic regression to fuse multiple quantitative indicators into a comprehensive confidence score.

[0153] In general, the confidence calculation of defect judgment takes multi-dimensional quantitative indicators based on three-dimensional morphological features as the core input. First, through predetermined judgment quantitative indicators such as height deviation ratio, volume filling ratio, curvature anomaly ratio, and spacing ratio, the defect-related features of the weld point are accurately extracted from different dimensions.

[0154] After weighted fusion of these multi-dimensional indicators, they are input into a logistic regression model with the Sigmoid function as the activation function, and finally the confidence scores of defect types in the range of [0,1] are calculated.

[0155] In summary, this confidence level directly reflects the probability that a solder joint belongs to a certain type of defect or is acceptable. The closer the value is to 1, the higher the reliability of the corresponding defect type determination. The whole process realizes a systematic transformation from feature quantification to probabilistic determination, providing a quantitative basis for the accurate identification of defect types.

[0156] Specifically, the defect types in the solder joint area are divided into 5 types of core defects and 1 type of acceptable condition, including:

[0157] Bridging: The spacing between adjacent solder joints is too small, with a spacing ratio of <0.5, which poses a risk of short circuit.

[0158] Cold solder joint / insufficient solder: Volume fill ratio <0.8, insufficient solder leads to unreliable connection;

[0159] Solder balls: Abnormal curvature accounts for >20%, and isolated protruding solder balls exist on the surface;

[0160] Offset: Height deviation ratio > 1.5 and solder joint center deviates from pad center > 30%;

[0161] Pass: All indicators are within the normal range, with the highest confidence level and >0.6.

[0162] Furthermore, the defect type of the wave crest solder joint region is determined based on the confidence score.

[0163] Furthermore, if the confidence level of a certain defect type is greater than 0.9, it is directly identified as that defect. For example, if the confidence level of bridging is 0.93, it is identified as bridging.

[0164] If the confidence level of all defect types is less than 0.6, and the confidence level of the qualified status is the highest, then it is judged as qualified.

[0165] If the confidence level is between 0.6 and 0.9, activate the expert rule base to assist in decision-making.

[0166] If it still cannot be determined, mark it as a suspected defect and submit it for manual review.

[0167] In summary, for the defect determination of the wave solder joint area of ​​the computing power card PCBA component, a classification system of 1 type of qualified status + 5 types of core defects was first defined. The confidence score is used as the core judgment basis, and the hierarchical judgment logic is executed according to different confidence intervals, forming a standard, logically hierarchical, and automated and manual verification solder joint defect type judgment scheme.

[0168] Specifically, the height deviation ratio is a quantitative parameter reflecting the abnormality of the surface flatness of the weld joint, which is the ratio of the actual height range to the standard height range.

[0169] Furthermore, the height deviation ratio is derived from statistical analysis of the height distribution data on the solder joint surface. The highest and lowest points in the entire solder joint area are located, and the height difference between them is calculated. This difference is then compared with a standard height difference reference value determined beforehand through statistical analysis of a large number of qualified solder joints, ultimately yielding the height deviation ratio. This index is primarily used to quantitatively evaluate the overall flatness consistency of the solder joint.

[0170] In summary, the height deviation ratio index can accurately quantify the overall flatness consistency of the weld joint, effectively identify warping and flattening defects caused by excessive height differences, and provide an intuitive and reliable basis for flatness evaluation for subsequent defect judgment.

[0171] Specifically, the volume fill ratio is a quantitative parameter that reflects the adequacy of solder filling; it is the ratio of the actual volume to the theoretical volume.

[0172] Furthermore, the volume fill ratio index is obtained through volume calculation and comparative analysis of the solder joint's three-dimensional spatial data. First, based on the surface's three-dimensional coordinate data, a numerical integration method is used to reconstruct the actual three-dimensional volume model of the solder joint, and the actual volume occupied by the solder is calculated. This actual volume value is then compared with the standard theoretical volume value determined based on the pad design dimensions and process specifications, and their relative ratio is calculated. This index directly reflects the sufficiency of solder filling.

[0173] In summary, the volume fill ratio directly reflects whether the solder fill is sufficient, providing a precise quantitative basis for judging defects related to fill volume, such as cold solder joints, insufficient solder, or solder overflow.

[0174] Specifically, the curvature anomaly ratio index is a quantitative parameter reflecting the microscopic morphological anomalies on the surface of the solder joint. It is the ratio of the area of ​​the region where the curvature value exceeds the threshold to the total surface area of ​​the solder joint.

[0175] Furthermore, the curvature anomaly percentage index is generated based on the statistical results of the solder joint surface curvature distribution data. First, an empirical curvature threshold is set to identify surface anomalies. Then, the curvature data of all regions on the solder joint surface is traversed to identify all abnormal regions whose curvature values ​​exceed the threshold, and the percentage of the cumulative area of ​​these abnormal regions relative to the total surface area of ​​the solder joint is calculated. This index is used to identify irregular geometric shapes such as localized warping and abrupt changes on the solder joint surface.

[0176] In summary, the curvature anomaly ratio index can accurately capture microscopic morphological problems such as local warping and abrupt edge changes on the solder joint surface, providing an intuitive and quantitative basis for identifying surface morphology-related defects such as solder balls and spikes.

[0177] Specifically, the spacing ratio is a quantitative parameter that reflects the short-circuit risk of adjacent solder joints. It is the ratio between the actual minimum spacing and the standard safe spacing.

[0178] Furthermore, the spacing ratio index is calculated by measuring the distance between adjacent weld points and combining it with design specifications. First, the minimum spacing distance between the edge of the weld point and its surrounding adjacent weld points is measured in three-dimensional space. Then, this measured minimum distance is compared with the minimum safe spacing standard value set according to electrical safety standards and process requirements to obtain the corresponding ratio result. This index is mainly used for early warning of bridging risks.

[0179] In summary, the spacing ratio index can directly reflect whether the spacing between adjacent weld points meets the safety requirements, and is the core quantitative basis for early warning of bridging defects and assessment of short-circuit risks.

[0180] In summary, by accurately calculating multi-dimensional quantitative indicators, a digital indicator system for comprehensively evaluating solder joint quality has been established, providing reliable input data for intelligent judgment. The indicator coverage is far higher than that of traditional methods.

[0181] Specifically, the confidence level is calculated using the following formula:

[0182]

[0183] Furthermore, among them, The solder joint belongs to the first Confidence level for each defect type The corresponding issues are bridging, cold solder joints, solder balls, misalignment, and acceptance.

[0184] Furthermore, For logical functions, It can map any real number z to the interval [0,1], which exactly matches the range of probability / confidence. It has specific nonlinear characteristics, is sensitive to intermediate values, and saturates with extreme values. It conforms to the logic of defect judgment - when the indicator is significantly abnormal, it can be clearly judged, and when it is slightly abnormal, it retains ambiguity.

[0185] Furthermore, To indicate the first The aforementioned quantitative indicators, For each of the following indicators—height deviation ratio, volume filling ratio, curvature anomaly ratio, and spacing ratio—each is converted into a dimensionless parameter by using actual / standard values ​​to ensure that the weighted summation of different indicators is meaningful.

[0186] Furthermore, For the first The first defect type corresponds to the first The preset threshold of the aforementioned quantitative indicator is determined through sample statistics and expert calibration.

[0187] Furthermore, for example:

[0188] Collect 10,000 labeled samples, including various defects and qualified solder joints;

[0189] For each defect type Statistics on its indicators Distribution characteristics, such as the spacing ratio of bridging defects;

[0190] The initial threshold is determined using the Youden exponent maximization method to maximize sensitivity and specificity;

[0191] Fine-tuning is done by incorporating the experience of process experts, such as taking into account minor differences in solder joints between different batches.

[0192] When bridging defects of When the spacing ratio is less than 0.8, it tends to be judged as a bridge.

[0193] Furthermore, For the first The aforementioned quantitative indicators are used to determine the first... The weight coefficients for the aforementioned defect types are learned through training samples.

[0194] Furthermore, the training sample learning steps are as follows: construct a training set containing N samples, each sample having 4 indicators and a true label; define the loss function using cross-entropy loss; then minimize the loss function using stochastic gradient descent; iteratively update the weights; and finally adjust the hyperparameters using the validation set to ensure the model's generalization ability.

[0195] Furthermore, This indicates a weighted summation of all the aforementioned quantitative indicators.

[0196] Furthermore, the weighted summation process involves calculating the difference between each indicator and its corresponding threshold, weighting these differences, summing the amplified deviations of key indicators to obtain the total score, and then substituting these values ​​into the input... The confidence level is obtained from the logical function.

[0197] Furthermore, the model can output the probability distribution of solder joints belonging to various categories such as bridging, cold solder joint, solder ball, offset, and qualified.

[0198] Furthermore, the final defect type is determined through a multi-level decision-making mechanism. A three-level decision rule is established: when the confidence level of a certain defect type... When the confidence level is >0.9, it is directly classified as this type of defect; when all confidence levels are >0.9, it is classified as this type of defect. When the confidence level is less than 0.6, the solder joint is considered acceptable. When the confidence level is between 0.6 and 0.9, a multi-rule comprehensive judgment is initiated, combining expert rule bases for auxiliary decision-making. For example, high curvature + low volume = solder ball defect. If it still cannot be determined, it is marked as a suspected defect and handed over to manual review. This mechanism ensures automation efficiency while keeping the false judgment rate within a controllable range.

[0199] Furthermore, the model parameters are optimized through a feedback learning mechanism. The system records the differences between the manual review results and the model's judgment, and updates the weight coefficients periodically. and threshold The stochastic gradient descent method is used to adjust the model parameters, enabling the model to continuously adapt to changes in the production process and achieve a continuous improvement in detection accuracy.

[0200] In summary, by using a logistic regression model and a weighted fusion algorithm, comprehensive analysis and decision-making based on multiple features are achieved, overcoming the limitations of single-feature judgment and making the defect identification accuracy far higher than that of traditional defect identification methods.

[0201] In summary, the introduction of a multi-level decision-making mechanism significantly reduces the risk of misjudgment while ensuring detection efficiency, and effectively improves the handling of complex and difficult cases.

[0202] In summary, the implementation of the feedback learning mechanism enables the system to continuously optimize, adapt to changes in production processes, and maintain a high detection accuracy rate over the long term.

[0203] In summary, this step, through the organic combination of intelligent decision-making models and multi-level judgment mechanisms, achieves accurate and reliable identification of solder joint defects, providing a scientific basis for quality control and embodying the core value of this invention.

[0204] S6. Based on the defect type and the location information of the wave solder joint area, perform spatial mapping on the layout diagram of the computing card PCBA component to obtain a defect distribution heat map of the wave solder joint area.

[0205] In this embodiment of the invention, the step of spatially mapping the layout diagram of the computing card PCBA component based on the defect type and the location information of the wave solder joint area to obtain a defect distribution heatmap of the wave solder joint area includes:

[0206] Obtain the coordinates of the defect location in the wave solder joint region;

[0207] Based on the defect type, the defect location coordinates are classified and aggregated to obtain the aggregated defect location coordinates of the wave solder joint region;

[0208] Based on the coordinates of the aggregated defects, a defect density distribution analysis is performed on the layout diagram of the computing card PCBA component to obtain the defect density value of the wave solder joint area.

[0209] Color rendering is applied to the defect density value to obtain a heat map of defect distribution in the wave crest solder joint region.

[0210] Specifically, the defect location coordinates are the key bridge connecting the microscopic image detection results with the macroscopic physical space of the PCBA. In essence, they accurately map the defective solder joints identified in the image to their physical locations on the actual PCBA board, which is a prerequisite for subsequent defect aggregation, density analysis, and heat map rendering.

[0211] Furthermore, the defect location coordinates of the wave solder joint area are obtained through coordinate calibration and transformation techniques. Based on pre-completed camera-projector system calibration parameters, a mapping relationship between the image pixel coordinate system and the PCBA physical coordinate system is established. An affine transformation matrix is ​​used to accurately transform the pixel coordinates of each defect solder joint in the image to millimeter coordinates in the PCBA design drawing. This is achieved through calibration using a checkerboard calibration board, with reprojection errors controlled within ±0.05mm. This process ensures that each defect can be accurately located in its actual position in the PCBA layout drawing.

[0212] In summary, through precise coordinate transformation and calibration techniques, accurate mapping from image space to physical space was achieved, ensuring that each defect can be located in its actual position with a positioning accuracy of ±0.1mm.

[0213] Specifically, aggregated defect location coordinates are the core coordinates and related spatial parameters that represent the clustered areas of defects of the same type after spatial clustering of the location coordinates.

[0214] Furthermore, the coordinates of aggregated defects are obtained through defect classification and spatial aggregation. Based on the defect types output by S5, such as bridging, cold solder joints, solder balls, and offsets, the defect coordinates are subjected to hierarchical clustering. The coordinates of defects of the same type are aggregated, and the aggregation center coordinates and their influence ranges for each type of defect are generated after clustering. At the same time, statistical information such as the number and density distribution of defects in each cluster region is collected to provide data support for process analysis.

[0215] In summary, through defect classification and spatial clustering analysis, the spatial patterns and clustering characteristics of defect distribution were revealed, providing an important basis for diagnosing process problems, with a clustering accuracy rate of over 90%.

[0216] Specifically, the layout diagram of the computing power card PCBA component is a digital drawing generated based on the computing power card PCBA design file and matched 1:1 with the actual physical size. Its core value is to establish the correspondence between image detection defects and the actual physical location of the PCBA, providing accurate spatial reference for subsequent defect location and distribution analysis.

[0217] Specifically, the defect density value is the number of defects per unit area within a specific region of the PCBA layout diagram, which is a quantitative value of defect aggregation intensity obtained by kernel density estimation.

[0218] Specifically, the defect distribution heatmap is a color-density correspondence visualization chart formed by overlaying the defect density values ​​onto the PCBA layout diagram through color gradient mapping.

[0219] Furthermore, for the location data of each type of defect, a two-dimensional density estimation is performed using a Gaussian kernel function to generate a defect density distribution map.

[0220] Furthermore, the defect density value is calculated as follows: the PCBA partial pattern is divided into uniform grid cells of 0.5mm×0.5mm, and each grid cell is a density calculation cell; all contribution values ​​calculated by the Gaussian kernel function are accumulated for each grid cell to obtain the density value of the cell, and the defect density value is mapped to a dimensionless exponent for easy subsequent color rendering.

[0221] Furthermore, the HSV color space is used for color mapping, mapping density values ​​to a predetermined color gradient: low-density areas are mapped to green, medium-density areas to yellow, and high-density areas to red. This heatmap layer is then overlaid on the PCB layout to generate a defect distribution heatmap with a gradient transparency effect. Legends, color bars, and statistical information annotations are also added, providing a complete visualization and analysis interface.

[0222] Furthermore, the system automatically identifies high-density abnormal areas on the heat map and performs correlation analysis in conjunction with PCB layout characteristics. It generates a process improvement suggestion report; for example, identifying a banded distribution of bridging defects along the guide rail direction may indicate that the wave soldering nozzle needs adjustment; cold solder joint defects accumulating around large components may reflect problems such as insufficient preheating temperature.

[0223] In summary, kernel density estimation technology enables quantitative analysis and visualization of defect distribution, clearly displaying areas of dense defects and significantly reducing density estimation errors.

[0224] In summary, through professional color mapping and visualization rendering, an intuitive and easy-to-understand heat map is generated, enabling process engineers to quickly identify problem areas and effectively improve decision-making efficiency.

[0225] In summary, this step, through the integrated application of spatial analysis and visualization technologies, transforms microscopic defect detection results into a macroscopic view of process quality, providing a scientific basis and data support for production process improvement and realizing the ultimate value transformation of detection results.

[0226] like Figure 2 The diagram shown is a functional block diagram of a computing power card wave solder joint defect detection system provided in an embodiment of the present invention.

[0227] The computing power card wave solder joint defect detection system 100 of the present invention can be installed in an electronic device. Depending on the functions implemented, the computing power card wave solder joint defect detection system 100 may include an image acquisition module 101, an image processing module 102, a region segmentation module 103, a three-dimensional feature analysis module 104, a recognition and classification module 105, and a visualization output module 106. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0228] In this embodiment, the functions of each module / unit are as follows:

[0229] The image acquisition module 101 is used to acquire high-resolution image sequences of the wave solder joint area on the computing card PCBA assembly.

[0230] The image processing module 102 is used to perform image optimization on the high-resolution image sequence to obtain a denoised and enhanced image sequence of the wave peak solder joint region.

[0231] The region segmentation module 103 is used to perform region segmentation on the denoised and enhanced image sequence to obtain a preprocessed solder joint sequence of the peak solder joint region.

[0232] The three-dimensional feature analysis module 104 is used to extract features from the preprocessed solder joint sequence to obtain the three-dimensional morphological features of the wave peak solder joint region.

[0233] The identification and classification module 105 is used to perform defect detection on the three-dimensional morphological features of the wave solder joint area to obtain the defect type of the wave solder joint area.

[0234] The visualization output module 106 is used to spatially map the layout diagram of the computing card PCBA component based on the defect type and the location information of the wave solder joint area to obtain a defect distribution heat map of the wave solder joint area.

[0235] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0236] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0237] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0238] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0239] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0240] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for detecting defects in a cap bump of a mining card, the method comprising: The method includes: ​ S1. Acquire high-resolution image sequences of the wave solder joint area on the PCBA assembly of the computing power card; S2. Perform image optimization on the high-resolution image sequence to obtain a denoised and enhanced image sequence of the wave peak solder joint region; S3. Perform region segmentation on the denoised and enhanced image sequence to obtain the preprocessed solder joint sequence of the peak solder joint region; S4. Perform feature extraction on the preprocessed solder joint sequence to obtain the three-dimensional morphological features of the wave peak solder joint region, including: Based on the preprocessed solder joint sequence, the relative height difference of different regions in the wave crest solder joint region is analyzed to obtain the height distribution data of the wave crest solder joint region. Based on the height distribution data, the surface curvature state of different regions in the wave crest solder joint area is analyzed to obtain the curvature distribution data of the wave crest solder joint area. Based on the height distribution data and the curvature distribution data, spatial volume analysis is performed on the wave crest solder joint region to obtain the volume quantification data of the wave crest solder joint region. The height distribution data, the curvature distribution data, and the volume quantization data are integrated into the three-dimensional morphological features of the wave crest solder joint region; S5. Perform defect detection on the three-dimensional morphological features of the wave solder joint area to obtain the defect types of the wave solder joint area, including: Based on predetermined quantification indicators, a defect confidence analysis is performed on the three-dimensional morphological features to obtain the confidence level of the defect type in the wave crest solder joint region. The predetermined quantification indicators include: A range analysis was performed on the height distribution characteristics of the three-dimensional morphology features to obtain the height deviation ratio index of the wave crest solder joint region. The volume quantization features of the three-dimensional morphology are compared with standard values ​​to obtain the volume fill ratio index of the wave crest solder joint region. An abnormal region proportion analysis is performed on the curvature distribution characteristics of the three-dimensional morphology features to obtain the curvature abnormality proportion index of the wave crest solder joint region. A safety spacing analysis is performed on the solder joint spacing characteristics of the three-dimensional morphology features to obtain the spacing ratio index of the wave crest solder joint region. The height deviation ratio, volume fill ratio, curvature anomaly percentage, and spacing ratio are integrated to obtain a predetermined quantitative index for the wave solder joint region. The formula for calculating the confidence level of the defect type in the wave solder joint region is as follows: ； in, The solder joint belongs to the first Confidence level for each defect type For logical functions, To indicate the first The aforementioned quantitative indicators, For the first The first defect type corresponds to the first The preset threshold of the aforementioned quantitative indicator, For the first The aforementioned quantitative indicators are used to determine the first... The weighting coefficients for the aforementioned defect types, This indicates a weighted summation of all the aforementioned quantitative indicators. The type of defect in the wave crest solder joint region is determined based on the confidence level. S6. Based on the defect type and the location information of the wave solder joint area, perform spatial mapping on the layout diagram of the computing card PCBA component to obtain a defect distribution heat map of the wave solder joint area.

2. The method for detecting defects in wave solder joints on a computing power card as described in claim 1, characterized in that, High-resolution image sequences of the wave solder joint area on the PCBA assembly of the computing power card were acquired, including: Obtain an image of the internal morphological features of the wave solder joint area in the computing card PCBA assembly; Obtain the deformed grating image sequence of the wave crest solder joint region; Pixel fusion is performed on the internal morphology feature image and the deformed grating image sequence to obtain a high-resolution image sequence of the wave crest solder joint region.

3. The method for detecting defects in wave solder joints on a computing power card as described in claim 1, characterized in that, The step of performing region segmentation on the denoised and enhanced image sequence to obtain the preprocessed solder joint sequence of the peak solder joint region includes: The denoised and enhanced image sequence is binarized based on a predetermined color threshold range to obtain a preliminary binary mask sequence for the wave peak solder joint region. The initial binary mask sequence is subjected to dilatational erosion to obtain the optimized binary mask sequence for the wave crest solder joint region; Based on the optimized binary mask sequence, solder joint targets are extracted from the denoised and enhanced image sequence to obtain the preprocessed solder joint sequence of the peak solder joint region.

4. The method for detecting defects in wave solder joints on a computing power card as described in claim 1, characterized in that, The step of optimizing the high-resolution image sequence to obtain the denoised and enhanced image sequence of the wave crest solder joint region includes: Image enhancement is performed on the high-resolution image sequence to obtain a contrast-enhanced image sequence of the wave peak solder joint region; The contrast-enhanced image sequence is subjected to noise removal to obtain the denoised enhanced image sequence of the wave peak solder joint region.

5. The method for detecting defects in wave solder joints on a computing power card as described in claim 1, characterized in that, The step of spatially mapping the layout diagram of the computing card PCBA component based on the defect type and the location information of the wave solder joint area to obtain a defect distribution heatmap of the wave solder joint area includes: Obtain the coordinates of the defect location in the wave solder joint region; Based on the defect type, the defect location coordinates are classified and aggregated to obtain the aggregated defect location coordinates of the wave solder joint region; Based on the coordinates of the aggregated defects, a defect density distribution analysis is performed on the layout diagram of the computing card PCBA component to obtain the defect density value of the wave solder joint area. Color rendering is applied to the defect density value to obtain a heat map of defect distribution in the wave crest solder joint region.

6. A computing power card wave solder joint defect detection system, characterized in that, The system includes: Image acquisition module: used to acquire high-resolution image sequences of the wave solder joint area on the computing card PCBA assembly; Image processing module: used to optimize the high-resolution image sequence to obtain a denoised and enhanced image sequence of the wave peak solder joint region; Region segmentation module: used to perform region segmentation on the denoised and enhanced image sequence to obtain the preprocessed solder joint sequence of the peak solder joint region; 3D Feature Analysis Module: Used to extract features from the preprocessed solder joint sequence to obtain the 3D morphological features of the wave peak solder joint region, including: Based on the preprocessed solder joint sequence, the relative height difference of different regions in the wave crest solder joint region is analyzed to obtain the height distribution data of the wave crest solder joint region. Based on the height distribution data, the surface curvature state of different regions in the wave crest solder joint area is analyzed to obtain the curvature distribution data of the wave crest solder joint area. Based on the height distribution data and the curvature distribution data, spatial volume analysis is performed on the wave crest solder joint region to obtain the volume quantification data of the wave crest solder joint region. The height distribution data, the curvature distribution data, and the volume quantization data are integrated into the three-dimensional morphological features of the wave crest solder joint region; The identification and classification module is used to perform defect detection on the three-dimensional morphological features of the wave solder joint region to obtain the defect type of the wave solder joint region, including: Based on predetermined quantification indicators, a defect confidence analysis is performed on the three-dimensional morphological features to obtain the confidence level of the defect type in the wave crest solder joint region. The predetermined quantification indicators include: A range analysis was performed on the height distribution characteristics of the three-dimensional morphology features to obtain the height deviation ratio index of the wave crest solder joint region. The volume quantization features of the three-dimensional morphology are compared with standard values ​​to obtain the volume fill ratio index of the wave crest solder joint region. An abnormal region proportion analysis is performed on the curvature distribution characteristics of the three-dimensional morphology features to obtain the curvature abnormality proportion index of the wave crest solder joint region. A safety spacing analysis is performed on the solder joint spacing characteristics of the three-dimensional morphology features to obtain the spacing ratio index of the wave crest solder joint region. The height deviation ratio, volume fill ratio, curvature anomaly percentage, and spacing ratio are integrated to obtain a predetermined quantitative index for the wave solder joint region. The formula for calculating the confidence level of the defect type in the wave solder joint region is as follows: ； in, The solder joint belongs to the first Confidence level for each defect type For logical functions, To indicate the first The aforementioned quantitative indicators, For the first The first defect type corresponds to the first The preset threshold of the aforementioned quantitative indicator, For the first The aforementioned quantitative indicators are used to determine the first... The weighting coefficients for the aforementioned defect types, This indicates a weighted summation of all the aforementioned quantitative indicators. The type of defect in the wave crest solder joint region is determined based on the confidence level. Visualization output module: Based on the defect type and the location information of the wave solder joint area, it performs spatial mapping on the layout diagram of the computing card PCBA component to obtain a heat map of defect distribution in the wave solder joint area.

Images