Welding Quality Detection Method and System for High-Order Computing Power Autopilot FPC Sensing Module

Abstract

The present invention relates to a welding quality detection method and system for a high-order computing power autonomous driving FPC sensing module, including the following steps: extracting features from the multi-dimensional spectral image set to obtain a welding feature vector of the welding area, and constructing a welding feature map based on the welding feature vector; performing dynamic feature mapping on the welding feature map through a map transformation network to obtain a welding feature descriptor; performing topological structure analysis on the welding area based on the welding feature descriptor to obtain a welding association graph; inputting the welding association graph into a preset region growing algorithm for quality detection and grade division to obtain a welding quality assessment report, solving the technical problem that traditional methods usually rely on manual inspection, which is not only inefficient, but also easily affected by the experience and subjective judgment of operators and cannot meet the requirements of modern manufacturing for automation and precision.

Description

Technical Field

[0001] The present invention relates to the technical field of FPC sensing modules, and particularly to a welding quality detection method and system for a high-order computing power autonomous driving FPC sensing module. Background Art

[0002] With the rapid development of autonomous driving technology, the requirements for vehicle perception systems are also increasing day by day. As an important component connecting sensors and control systems, FPC (Flexible Printed Circuit) sensing modules play a crucial role in autonomous vehicles. These modules need to maintain high reliability in complex environments, so the welding quality in their manufacturing process is of vital importance. However, traditional welding quality detection methods, such as visual inspection and X-ray inspection, although capable of identifying obvious defects, are ineffective for detecting tiny or hidden welding problems. In addition, traditional methods usually rely on manual inspection, which is not only inefficient but also easily affected by the experience and subjective judgment of operators, and cannot meet the requirements of modern manufacturing for automation and precision.

[0003] Driven by the demand for high-order computing power, autonomous driving technology is evolving towards a more intelligent direction, which poses higher requirements for FPC sensing modules. To ensure the stability and safety of autonomous driving systems, more advanced and reliable welding quality detection means must be introduced. The development of existing multispectral imaging technology and convolutional neural networks has brought new opportunities to this field. By using multispectral imaging, more information can be obtained than traditional methods, and the application of deep learning algorithms makes it possible to automatically analyze and understand this complex data. However, how to effectively combine these two advanced technologies and apply them to the welding quality assessment in the actual production environment is still an urgent problem to be solved.

[0004] In response to the above challenges, a welding quality detection method for a high-order computing power autonomous driving FPC sensing module based on the combination of multispectral imaging and deep learning is proposed. The method aims to overcome the limitations of the existing technology and provide an efficient, accurate and automated solution. It uses multi-angle scanning imaging to capture the multi-dimensional features of the welding area, and uses a convolutional neural network for feature extraction to construct a detailed welding feature map. Further, through the map transformation network, dynamic feature mapping is realized, and the region growing algorithm is used for quality detection and grading, and finally a welding quality assessment report is generated. This method not only improves the accuracy of detection, but also greatly shortens the detection time, which helps to improve the efficiency and product quality of the entire manufacturing process. Summary of the Invention

[0005] The main objective of the present invention is to provide a method and system for detecting the welding quality of a high-order computing power autonomous driving FPC sensing module, which solves the technical problem that traditional methods usually rely on manual inspection, which is not only inefficient but also easily affected by the experience and subjective judgment of operators, and cannot meet the requirements of modern manufacturing for automation and precision.

[0006] To achieve the above objective, the present invention provides a method for detecting the welding quality of a high-order computing power autonomous driving FPC sensing module, including the following steps:

[0007] Perform multi-angle scanning imaging on the welding area of the target FPC sensing module through a preset multi-spectral imaging system to obtain a multi-dimensional spectral image set of the welding area;

[0008] Extract features from the multi-dimensional spectral image set through a preset convolutional neural network to obtain a welding feature vector of the welding area, and construct a welding feature map based on the welding feature vector;

[0009] Perform dynamic feature mapping on the welding feature map through a map transformation network to obtain a welding feature descriptor;

[0010] Perform topological structure analysis on the welding area based on the welding feature descriptor to obtain a welding association graph;

[0011] Input the welding association graph into a preset region growing algorithm for quality detection and grade classification to obtain a welding quality assessment report.

[0012] Further, the step of performing multi-angle scanning imaging on the welding area of the target FPC sensing module through a preset multi-spectral imaging system to obtain a multi-dimensional spectral image set of the welding area includes:

[0013] Perform full-spectrum scanning on the welding area through a preset multi-spectral imaging system to obtain an original spectral data sequence, where the original spectral data sequence includes welding information of visible spectrum, near-infrared spectrum, and ultraviolet spectrum;

[0014] Separate the bands of the original spectral data sequence to obtain a multi-band spectral feature map, and perform spectral enhancement processing on the multi-band spectral feature map to obtain an enhanced spectral feature map;

[0015] Perform three-dimensional reconstruction based on the enhanced spectral feature map to obtain a welding three-dimensional spectral map;

[0016] Perform spatial registration on the welding three-dimensional spectral map based on a spectral registration network to obtain a registered spectral data set, and perform feature fusion on the registered spectral data set through a spectral feature fusion algorithm to obtain a fused spectral feature map;

[0017] Perform endmember analysis on the fused spectral feature map through a spectral unmixing algorithm to obtain a distribution map of welding material components, and perform geometric correction on the distribution map of welding material components to obtain a corrected component map; wherein, the distribution map of welding material components includes solder composition, substrate characteristics, and interface structure information;

[0018] Perform dimensionality reduction and denoising processing on the corrected component map to obtain a multi-dimensional spectral image set of the welding area.

[0019] Furthermore, the method for extracting features of the multi-dimensional spectral image set through a preset convolutional neural network to obtain a welding feature vector of the welding area and constructing a welding feature map based on the welding feature vector includes:

[0020] Extract spatial-spectral features from the multi-dimensional spectral image set through a preset convolutional neural network to obtain a spatial-spectral feature map; wherein, the spatial-spectral feature map includes spatial texture information and spectral composition information of the welding area;

[0021] Perform multi-scale feature fusion on the spatial-spectral feature map through a pyramid pooling module to obtain a multi-scale feature representation, and perform local binary pattern encoding on the multi-scale feature representation to obtain an LBP feature map;

[0022] Perform global context modeling on the LBP feature map based on a Transformer network to obtain a global context feature vector;

[0023] Perform feature compression on the global context feature vector through local sensitive hashing encoding technology to obtain a sparse feature vector, and use the sparse feature vector as the welding feature vector of the welding area.

[0024] Perform relationship modeling based on the sparse feature vector to obtain a welding feature relationship map, and construct a welding feature map based on the welding feature relationship map; wherein, the welding feature map includes welding strength distribution data, welding depth distribution data, and welding surface flatness data.

[0025] Furthermore, the method for performing dynamic feature mapping on the welding feature map through a map transformation network to obtain a welding feature descriptor includes:

[0026] Perform multi-resolution decomposition on the welding feature map through a preset map transformation network to obtain a set of welding feature sub-maps;

[0027] Perform temporal feature analysis on the set of welding feature sub-maps based on a bidirectional long short-term memory network to obtain a temporal feature sequence, and perform structural analysis on the temporal feature sequence through a graph attention network to obtain a welding dynamic feature map;

[0028] Perform Fourier spectrum analysis on the welding dynamic feature map to obtain the frequency-domain feature spectrum;

[0029] Extract key features from the frequency-domain feature spectrum through an adaptive feature selection algorithm to obtain a welding feature description vector;

[0030] Perform topological structure encoding on the welding feature description vector based on a graph isomorphism network to obtain a structured feature encoding;

[0031] Perform feature transformation and mapping on the structured feature encoding through a graph spectrum mapping algorithm to obtain a welding feature descriptor; wherein, the welding feature descriptor includes the interface bonding state, void distribution density, and metal interconnect strength.

[0032] Further, perform topological structure analysis on the welding area based on the welding feature descriptor to obtain a welding correlation graph, including:

[0033] Perform local structure deconstruction on the welding feature descriptor through a preset welding distribution analysis algorithm to obtain a local feature distribution map of the welding area, and perform feature connection analysis on the local feature distribution map to obtain welding feature local connection data; wherein, the welding feature local connection data includes the position information, connection weight information, and node type information of welding nodes;

[0034] Perform global network modeling on the welding area based on the welding feature local connection data to obtain a global topological structure map of the welding area, and perform feature mapping on the global topological structure map to obtain an embedded topological feature vector;

[0035] Perform dynamic adjacency relationship inference on the embedded topological feature vector through a topological optimization network to obtain dynamic feature correlation data between welding areas, and perform feature aggregation on the dynamic feature correlation data to obtain a welding clustering feature distribution map, which includes spatial correlation information of welding population structures and defect features;

[0036] Perform geometric structure decomposition on the welding clustering feature distribution map through a geometric correlation analysis algorithm to obtain a geometric correlation map of the welding area, and perform high-order feature mapping on the geometric correlation map to obtain a welding geometric feature description map;

[0037] Perform multi-layer relationship modeling on the dynamic characteristics of the welding area based on the welding geometric feature description map to obtain a welding dynamic feature map;

[0038] Perform feature joint encoding on the welding dynamic feature map and the geometric correlation map through a high-order feature fusion algorithm to obtain a welding correlation graph.

[0039] Further, inputting the welding correlation diagram into a preset region growing algorithm for quality inspection and grading to obtain a welding quality assessment report, including:

[0040] Locating seed points of the welding correlation diagram through a region growing seed selection algorithm to obtain a set of seed points for the welding region, and calculating region growing factor data for the set of seed points for the welding region to obtain region growing characteristic data; wherein, the region growing characteristic data includes welding connection strength, welding boundary gradient, and welding region morphological characteristics;

[0041] Based on the region growing characteristic data, adaptively expanding the welding region to obtain a welding region growth diagram, and extracting boundary characteristics of the welding region growth diagram through a region boundary tracking algorithm to obtain a set of welding region boundary characteristics;

[0042] Performing quality inspection and grading on the set of welding region boundary characteristics through a preset quality assessment criterion to obtain a welding quality grade distribution diagram;

[0043] Performing deep feature analysis on the welding quality grade distribution diagram based on a deep feature analysis network to obtain a welding quality feature vector, and performing hierarchical evaluation on the welding quality feature vector through a multi-level evaluation algorithm to obtain welding quality evaluation data; wherein, the welding quality evaluation data includes welding strength score, welding depth score, and welding surface score;

[0044] Performing structured processing and content optimization on the welding quality evaluation data through an evaluation report generation algorithm to obtain a welding quality assessment report; wherein, the welding quality assessment report includes quality grade distribution statistics, defect type analysis, and process parameter suggestions.

[0045] Further, the adaptively expanding the welding region based on the region growing characteristic data to obtain a welding region growth diagram includes:

[0046] Performing dynamic threshold calculation on the region growing characteristic data through a preset region growing threshold adaptive algorithm to obtain a set of region growing thresholds, and locally optimizing the region growing thresholds based on the welding region morphological characteristics to obtain an optimized set of growing thresholds;

[0047] Performing multi-directional expansion calculation on the welding region based on the optimized set of growing thresholds to obtain a region expansion direction diagram, and performing connected domain marking on the expansion direction diagram through a region connectivity analysis algorithm to obtain a welding region connectivity marking diagram;

[0048] Performing growth constraint calculation on the welding region connectivity marking diagram through a preset region growing constraint algorithm to obtain region growing constraint data,

[0049] Perform multi-region competitive growth on the region growth constraint data based on a region competition mechanism to obtain a region competition growth map, and perform similar region merging on the region competition growth map through a region merging algorithm to obtain a welding region growth map.

[0050] The present invention also provides a welding quality detection system for a high-order computing power autonomous driving FPC sensing module, including:

[0051] An imaging module, configured to perform multi-angle scanning imaging on the welding region of the target FPC sensing module through a preset multi-spectral imaging system to obtain a multi-dimensional spectral image set of the welding region;

[0052] An extraction module, configured to perform feature extraction on the multi-dimensional spectral image set through a preset convolutional neural network to obtain a welding feature vector of the welding region, and construct a welding feature map based on the welding feature vector;

[0053] A mapping module, configured to perform dynamic feature mapping on the welding feature map through a map transformation network to obtain a welding feature descriptor;

[0054] An analysis module, configured to perform topological structure analysis on the welding region based on the welding feature descriptor to obtain a welding association graph;

[0055] A division module, configured to input the welding association graph into a preset region growth algorithm for quality level division to obtain a welding quality assessment report.

[0056] The present invention also provides a computer device, including a memory and a processor, where a computer program is stored in the memory, and when the processor executes the computer program, the steps of the method described in any one of the above are implemented.

[0057] The present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described in any one of the above are implemented.

[0058] The welding quality detection method for the high-order computing power autonomous driving FPC sensing module provided by the present invention includes the following steps: performing multi-angle scanning imaging on the welding area of the target FPC sensing module through a preset multi-spectral imaging system to obtain a multi-dimensional spectral image set of the welding area; extracting features from the multi-dimensional spectral image set through a preset convolutional neural network to obtain a welding feature vector of the welding area, and constructing a welding feature map based on the welding feature vector; performing dynamic feature mapping on the welding feature map through a map transformation network to obtain a welding feature descriptor; performing topological structure analysis on the welding area based on the welding feature descriptor to obtain a welding association graph; inputting the welding association graph into a preset region growing algorithm for quality detection and grade division to obtain a welding quality assessment report, which solves the technical problem that traditional methods usually rely on manual inspection, which is not only inefficient but also easily affected by the experience and subjective judgment of operators and cannot meet the requirements of modern manufacturing for automation and precision, and realizes topological structure analysis based on the welding feature descriptor, enabling the detection system to deeply understand the spatial relationship and connectivity between welding points. This in-depth structured analysis can help discover tiny defects or potential problems that are difficult to detect by traditional methods, ensuring a comprehensive evaluation of welding quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is a schematic diagram of the steps of the welding quality detection method for the high-order computing power autonomous driving FPC sensing module in an embodiment of the present invention;

[0060] Figure 2 is a structural block diagram of the welding quality detection system for the high-order computing power autonomous driving FPC sensing module in an embodiment of the present invention;

[0061] Figure 3 is a structural schematic diagram of a computer device in an embodiment of the present invention.

[0062] The implementation, functional features, and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] In order to make the object, technical solution, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0064] As Figure 1 shown, Figure 1 is a schematic diagram of the steps of a welding quality detection method for a high-order computing power autonomous driving FPC sensing module in an embodiment of the present invention;

[0065] In an embodiment of the present invention, a method for detecting the welding quality of a high-order computing power autonomous driving FPC sensing module is provided, including the following steps:

[0066] Step S1, performing multi-angle scanning imaging on the welding area of the target FPC sensing module through a preset multi-spectral imaging system to obtain a multi-dimensional spectral image set of the welding area.

[0067] Specifically, in the process of realizing multi-angle scanning imaging of the welding area of the target FPC sensing module, the preset multi-spectral imaging system plays a crucial role. This process first involves selecting and installing suitable multi-spectral imaging devices, which can emit and capture light in different wavelength ranges, thereby forming a series of images representing different spectral information. For example, on the production line of the autonomous driving FPC sensing module, in order to ensure the welding quality, engineers will set up a multi-spectral imaging system that can scan the welding area from multiple different angles at a specific time point. When the multi-spectral imaging system is started, it will gradually scan the welding area according to the predetermined angles and sequences. Each scan will generate an image reflecting the spectral characteristics of the welding area at that angle. Due to the adoption of multi-spectral technology, each image is not just an ordinary black-and-white or color photo, but contains the reflectivity or transmittance information of the welding area under different spectra. Thus, by integrating multiple images from different angles, a more complete and detailed multi-dimensional spectral image set of the welding area can be constructed. For example, in practical applications, if there are fine cracks or insufficient soldering at a certain welding point, then in the images under certain specific spectra, it will show characteristics different from those of normal welding points, and these differences may be ignored in conventional visual inspections. In this process, the accuracy and stability of the system are crucial for obtaining high-quality multi-dimensional spectral images. Therefore, engineers need to carefully calibrate the multi-spectral imaging system to ensure that each scanning angle can provide clear and distortion-free images. At the same time, the influence of ambient light conditions needs to be considered, because external light sources may interfere with the imaging effect. To solve this problem, the scanning can be carried out in a closed environment, or filters can be used to shield unnecessary light. The finally obtained multi-dimensional spectral image set can not only reflect the morphological characteristics of the welding surface, but also reveal potential changes in material properties, providing a solid data basis for subsequent feature extraction and quality assessment. This process is a key step in the entire method for detecting the welding quality of the high-order computing power autonomous driving FPC sensing module, directly related to whether welding defects can be accurately identified, and thus ensuring the reliable operation of the sensing module in a complex driving environment.

[0068] Step S2, extract features from the multi-dimensional spectral image set through a preset convolutional neural network to obtain a welding feature vector of the welding area, and construct a welding feature map based on the welding feature vector.

[0069] Specifically, in the process of extracting features from the multi-dimensional spectral image set, the preset convolutional neural network (CNN) plays a core role. This process automatically identifies and extracts representative features from the multi-dimensional spectral image set of the welding area, converts these features into a welding feature vector, and then constructs a welding feature map. Specifically, when the multi-spectral imaging system completes the multi-angle scanning of the welding area of the FPC sensing module, the obtained multi-dimensional spectral image set is input into the already trained convolutional neural network. The convolutional neural network first processes these image data through a series of convolutional layers. In each convolutional layer, the network uses multiple small filters to perform convolutional operations with different parts of the image to capture local patterns and structural information in the image. As the data is passed layer by layer, the network can gradually learn feature representations from simple to complex, such as from low-level features like edges and textures to higher-level geometric shape or material property features. For the application scenario of the autonomous driving FPC sensing module, this means that the CNN can identify the morphology of the solder joints, the distribution of the solder, and possible defects such as cracks or pores. Once the convolutional neural network completes the feature extraction of the multi-dimensional spectral image set, it outputs a high-dimensional feature vector, and this welding feature vector contains all the extracted feature information. To make this information more intuitive and easy to analyze, a welding feature map will be constructed based on this welding feature vector next. The process of constructing the map involves mapping the values in the feature vector into a two-dimensional or three-dimensional space to form a map that can visually display the features of the welding area. For example, in practical applications, engineers can quickly locate problems by viewing the welding feature map and evaluate whether the welding quality meets the standards. In addition, to ensure that the convolutional neural network can accurately complete the feature extraction task, it must be fully trained in advance. This usually requires a large amount of labeled data as input to let the network learn how to distinguish between qualified and unqualified welding situations. By continuously adjusting the network parameters to minimize the prediction error, finally the CNN can also give a reliable welding feature vector when facing a new multi-dimensional spectral image set. The whole process not only improves the detection efficiency but also greatly enhances the degree of automation, providing a scientific basis for subsequent quality assessment. In this way, whether it is daily production monitoring or the new product R & D stage, this method can be relied on to ensure the consistency and reliability of the welding quality of the FPC sensing module, thus ensuring the stable operation of the autonomous driving system.

[0070] Step S3, perform dynamic feature mapping on the welding feature map through a map transformation network to obtain a welding feature descriptor.

[0071] Specifically, after the construction of the welding feature map is completed, the next step is to perform dynamic feature mapping on the welding feature map through a map transformation network. This step aims to further process and analyze the extracted feature information to obtain more abstract and descriptive welding feature descriptors. Specifically, when the convolutional neural network generates welding feature vectors and constructs corresponding welding feature maps, although these maps intuitively show the feature distribution of the welding area, they are still mainly static, image-level representations. To extract deeper time-series changes or other dynamic characteristics from these static features, a map transformation network is introduced to perform dynamic feature mapping. This network can capture the change patterns that may occur during the welding process, such as subtle differences caused by fluctuations in welding process parameters or the mutual influence between different welding points. In the application scenario of the autonomous driving FPC sensing module, this means that the map transformation network can identify the trend of solder joints evolving over time and the dynamic behavior during the melting and cooling processes of the solder. For example, if a certain solder joint shows uneven shrinkage during the cooling process, this phenomenon may not be directly reflected in the static welding feature map, but through dynamic feature mapping, this non-linear change can be transformed into an easily understandable feature descriptor. The working principle of the map transformation network involves a series of complex mathematical operations and model conversions. It uses transformation mechanisms in deep learning technologies, such as self-attention mechanisms or recurrent neural networks (RNNs), to simulate and learn the spatio-temporal relationships in the welding feature map. In this way, the network can automatically adjust its internal structure to adapt to the feature expression requirements in different welding situations. In practical applications, engineers can train the map transformation network to learn how to extract the most representative dynamic features from the welding feature map and encode them as welding feature descriptors. These descriptors not only contain the spatial information in the original image but also incorporate the change laws in the time dimension, thus providing a more comprehensive basis for welding quality assessment. In addition, to ensure the effectiveness and reliability of the map transformation network, a large amount of labeled data is required for supervised learning. This includes multi-dimensional spectral image sets collected under various welding conditions and their corresponding high-quality welding feature descriptors. By continuously optimizing the network parameters, the map transformation network can accurately perform dynamic feature mapping when faced with new welding feature maps. The finally obtained welding feature descriptors not only contribute to subsequent topological structure analysis and quality inspection but can also be used as feedback information to improve the welding process and increase production efficiency. In this way, during the manufacturing process of the entire high-order computing power autonomous driving FPC sensing module, through the refined management and real-time monitoring of welding quality, the reliability and safety of the product can be effectively guaranteed, thus promoting the development of autonomous driving technology.

[0072] Step S4: Based on the welding feature descriptors, perform a topological structure analysis on the welding area to obtain a welding connection graph.

[0073] Specifically, after the construction of the welding feature descriptors is completed, next, perform a topological structure analysis on the welding area based on the welding feature descriptors. This process aims to deeply analyze the spatial relationships and connection patterns between welding points to obtain deeper welding quality information. Through topological structure analysis, it is possible to identify how each welding point in the welding area is interconnected and the impact of these connections on the stability and reliability of the entire welding structure. For example, in the manufacturing process of high-order computing power autonomous driving FPC (Flexible Printed Circuit) sensing modules, the welding quality is directly related to whether the electrical connection between electronic components and the circuit board is firm and reliable. Therefore, accurately understanding the topological structure of the welding area is crucial. When the spectral transformation network completes the dynamic feature mapping and generates the welding feature descriptors, the next step is to use these descriptors for a more refined topological structure analysis. This step involves taking the welding feature descriptors as input data and calculating geometric parameters such as the relative positions, distances, and angles between different welding points through specific algorithms. Based on these parameters, a mathematical model that can comprehensively reflect the internal structure of the welding area is established - namely, the welding connection graph. The welding connection graph not only records the specific coordinates of each welding point, but more importantly, it depicts the connection paths and strength distributions between the welding points, enabling engineers to intuitively see which parts are key connection points and which areas may have potential weak links. In practical applications, for example, for an autonomous driving FPC sensing module, assume that a certain circuit requires particularly stable signal transmission. Then, through topological structure analysis, all the welding points involved in this circuit can be identified, and further check whether there are any abnormal connections or weak areas in its welding connection graph that may affect signal integrity. If it is detected that the connection density around a certain welding point is low or there are unexpected long-distance connections, this may indicate problems during the welding process, such as insufficient solder or poor welding caused by overheating. At this time, corresponding improvement measures can be taken based on the detailed information provided by the welding connection graph to ensure that all welding points meet the best quality standards, thereby guaranteeing the performance and safety of the final product. In addition, to achieve efficient and accurate topological structure analysis, machine learning techniques are usually combined to assist the analysis process. By training classifiers or clustering algorithms, the system can automatically learn the distinguishing features between normal and abnormal welding patterns from a large number of welding feature descriptors, thereby improving the understanding ability of complex welding structures. This can not only speed up the analysis process but also reduce errors caused by human judgment. In summary, through a comprehensive and detailed topological structure analysis of the welding area and obtaining an accurate welding connection graph, it provides a solid foundation for subsequent quality control and also points the way for optimizing the welding process flow.

[0074] Step S5: Input the welding correlation diagram into a preset region growing algorithm for quality inspection and grading to obtain a welding quality assessment report.

[0075] Specifically, after completing the topological structure analysis of the welding area and obtaining the welding correlation graph, the next step is to input the welding correlation graph into a preset region growing algorithm for quality inspection and grading. This process is to further evaluate the quality of the welding, ensure that each welding point meets the established standards, and thus guarantee the safety and reliability of the entire welding structure. Specifically, the region growing algorithm is an image processing technique that starts from one or more seed points and gradually adds adjacent pixels with similar attributes (such as gray value, color, texture, etc.) to the same region until all eligible pixels are included. In this application scenario, we are concerned with the connection strength and stability between welding points. Therefore, the algorithm will use the information provided by the welding correlation graph, take the key welding points as seed points, and then determine which welding points should be classified into the same quality level according to the set rules and thresholds, such as the distance, angle between welding points, and solder distribution. Taking the manufacturing of high-order computing power autonomous driving FPC sensing modules as an example, assume that we have obtained a detailed welding correlation graph of the welding area of this module, which records the positions of each welding point and their mutual relationships. Now, we input this graph into a pre-configured region growing algorithm. The algorithm will first select some representative welding points as initial seed points, and these points may be the key connection positions determined through topological structure analysis. As the algorithm runs, it will check the welding points around each seed point and judge whether these welding points should be merged with the seed point into a larger region according to their physical properties (such as solder thickness, contact area, etc.) and electrical performance (such as resistance value). If it is found that a certain welding point does not meet the expected quality standard, such as having a dry joint or excessive solder, then this point will not be added to the current region but will be marked for subsequent review. When all welding points have been evaluated, the region growing algorithm will generate a detailed welding quality assessment report. This report not only includes the specific quality ratings of each welding point but may also point out the problem areas that need special attention. For engineers, this is a very valuable reference material because it can help them quickly locate potential quality hazards and take corresponding corrective measures. For example, in the case of the autonomous driving FPC sensing module, if some welding points are rated as low grade, this may mean that the signal transmission efficiency in this area will be affected, thereby affecting the performance of the entire system. At this time, engineers can adjust the welding process parameters or replace the problematic components according to the suggestions in the report to ensure that the final product can meet the strict quality requirements. In addition, by continuously applying this process, manufacturers can also continuously optimize the welding process, improve production efficiency while also improving product quality. In short, by combining the welding correlation graph with the region growing algorithm, effective monitoring and management of welding quality can be achieved, providing guarantee for the reliable operation of complex electronic devices.

[0076] In a specific embodiment, the welding area of the target FPC sensing module is scanned and imaged from multiple angles by a preset multispectral imaging system to obtain a multi-dimensional spectral image set of the welding area, including:

[0077] The welding area is scanned with the full spectrum by a preset multispectral imaging system to obtain an original spectral data sequence, where the original spectral data sequence includes welding information of visible spectrum, near-infrared spectrum and ultraviolet spectrum;

[0078] The original spectral data sequence is separated into bands to obtain a multi-band spectral feature map, and the multi-band spectral feature map is subjected to spectral enhancement processing to obtain an enhanced spectral feature map;

[0079] Three-dimensional reconstruction is performed based on the enhanced spectral feature map to obtain a welded three-dimensional spectral map;

[0080] Spatial registration is performed on the welded three-dimensional spectral map based on a spectral registration network to obtain a registered spectral data set, and feature fusion is performed on the registered spectral data set through a spectral feature fusion algorithm to obtain a fused spectral feature map;

[0081] Endmember analysis is performed on the fused spectral feature map through a spectral unmixing algorithm to obtain a welding material component distribution map; and geometric correction is performed on the welding material component distribution map to obtain a corrected component map; where the welding material component distribution map includes solder composition, substrate characteristics and interface structure information;

[0082] Dimensionality reduction and denoising processing are performed on the corrected component map to obtain the multi-dimensional spectral image set of the welding area.

[0083] Specifically, during the process of multi-angle scanning and imaging of the welding area of the target FPC sensing module, a series of complex steps are achieved through a preset multi-spectral imaging system to ensure that the most detailed and accurate information of the welding area can be captured. This process begins with full-spectrum scanning, that is, using the multi-spectral imaging system to comprehensively cover the welding area and obtain the original spectral data sequence in the range from visible spectrum, near-infrared spectrum to ultraviolet spectrum. These original spectral data not only contain the color information of the welding surface, but also record the ability of the material to reflect or transmit light of different wavelengths, providing a rich basis for subsequent analysis. Specifically, in the application scenario of the autonomous driving FPC sensing module, when engineers need to evaluate the welding quality of a new batch, they will start the multi-spectral imaging system, which is equipped with a variety of filters and sensors and can complete the full-spectrum scanning of the welding points in a short time. Each scan will generate a large amount of original spectral data, which are organized into an ordered data sequence, and each data point represents the reflectivity or transmittance of a certain part of the welding area at a specific wavelength. For example, for the interface between the solder and the substrate, different materials may show similar color characteristics in the visible spectrum range, but significant differences are shown in the near-infrared or ultraviolet spectrum, which helps to identify potential problems such as poor soldering or uneven solder distribution. To further process these original spectral data sequences, the next step is to separate the bands of the original spectral data sequence, aiming to decompose the entire spectral range into multiple independent bands, thereby obtaining a multi-band spectral feature map. In this process, the image of each band can highlight the welding characteristics within a specific wavelength range. Then, in order to enhance the key features in the image, spectral enhancement processing will be performed on these multi-band spectral feature maps. By adjusting the contrast, brightness or other parameters, the subtle differences in the welding area become more obvious. For example, after spectral enhancement processing, tiny cracks or pores that were originally difficult to detect may become clearly visible, which is crucial for ensuring welding quality. Based on the enhanced spectral feature map, the next step is the 3D reconstruction stage, where the goal is to construct a three-dimensional stereo spectral map of the welding. This step involves using computer vision technology to infer the spatial structure of the welding area based on the two-dimensional spectral feature maps from various perspectives. 3D reconstruction can not only provide a planar view of the welding points, but also show their height changes, helping engineers to more intuitively understand the morphological characteristics of the welding. For example, in some cases, the height difference of the welding points may be caused by spatter generated during the welding process, and 3D reconstruction can accurately reveal this microscopic structure so as to take appropriate corrective measures. Once the three-dimensional stereo spectral map of the welding is obtained, spatial registration needs to be performed on it to ensure that the data from different perspectives or time points can be accurately aligned. This is achieved through a spectral registration network, which can automatically identify and correct the deviations caused by changes in imaging conditions (such as camera position, angle, illumination, etc.).The registered spectral dataset lays the foundation for the next step of feature fusion. On this basis, through the spectral feature fusion algorithm, information from different bands is integrated together to generate a comprehensive fused spectral feature map. This step enhances the overall information content of the image while reducing redundant data, making the welding features more concentrated and prominent. To gain a deeper understanding of the composition and distribution of the welding materials, the spectral unmixing algorithm is then used to perform endmember analysis on the fused spectral feature map, resulting in a distribution map of the welding material components. This process is similar to spectral analysis in chemical analysis, where by identifying different spectral signatures, the main components in the welding materials and their spatial distributions can be determined. For example, solder composition, substrate characteristics, and interface structure information can all be clearly shown in this map. However, due to the inevitable introduction of some geometric distortions during the actual imaging process, geometric correction is also required for the distribution map of the welding material components to ensure the authenticity and accuracy of the final result. The corrected component map can provide engineers with first-hand information about the welding quality, helping them quickly locate problems. Finally, for the convenience of subsequent quality inspection and analysis, dimensionality reduction and denoising processing are required for the corrected component map to obtain a multi-dimensional spectral image set. For example, during the manufacturing process of the FPC sensing module for autonomous driving, the multi-dimensional spectral image set after dimensionality reduction and denoising processing can be used in an automated detection system to quickly screen qualified and unqualified products, greatly improving production efficiency and product quality control levels. In summary, through a series of carefully designed operations such as full-spectrum scanning, band separation, spectral enhancement, three-dimensional reconstruction, spatial registration, feature fusion, endmember analysis, geometric correction, and dimensionality reduction and denoising of the welding area, a comprehensive assessment of the welding quality of the FPC sensing module is achieved. This method not only improves the detection accuracy but also provides strong support for optimizing the welding process and improving product reliability, ensuring the stable operation of the autonomous driving system.

[0084] In a specific embodiment, the feature extraction of the multi-dimensional spectral image set through a preset convolutional neural network to obtain a welding feature vector of the welding area and the construction of a welding feature map based on the welding feature vector includes:

[0085] Performing spatio-spectral feature extraction on the multi-dimensional spectral image set through a preset convolutional neural network to obtain a spatio-spectral feature map; wherein, the spatio-spectral feature map includes spatial texture information and spectral component information of the welding area;

[0086] Performing multi-scale feature fusion on the spatio-spectral feature map through a pyramid pooling module to obtain a multi-scale feature representation, and performing local binary pattern encoding on the multi-scale feature representation to obtain an LBP feature map;

[0087] Perform global context modeling on the LBP feature map based on the Transformer network to obtain a global context feature vector;

[0088] Feature compress the global context feature vector through local sensitive hashing encoding technology to obtain a sparse feature vector, and use the sparse feature vector as the welding feature vector of the welding area;

[0089] Perform relationship modeling based on the sparse feature vector to obtain a welding feature relationship graph, and construct a welding feature map based on the welding feature relationship graph; wherein, the welding feature map includes welding strength distribution data, welding depth distribution data, and welding surface flatness data.

[0090] Specifically, in order to achieve an in-depth analysis of the welding area of the FPC sensing module, the method mentioned in the claims introduces a pre-set convolutional neural network (CNN) to extract features from a multi-dimensional spectral image set. This process is completed by inputting the multi-dimensional spectral image set into a pre-trained convolutional neural network, which can recognize and learn complex patterns in the images, thus obtaining an empty-spectrum feature map containing spatial texture information and spectral component information of the welding area. In this process, the multi-layer structure of the convolutional neural network enables it to gradually abstract high-level features (such as shape and texture) from low-level features (such as edges and corners), which is crucial for understanding the quality of the welding. For example, in the process of manufacturing autonomous driving sensors, engineers may use this method to ensure that each welding point achieves optimal performance. When a multi-dimensional spectral image set is fed into the convolutional neural network, the network will automatically process these images and extract the features that are most critical for welding quality assessment. For instance, if the welding points of a certain batch look normal under visible light but show abnormal reflectivity changes in the spectral images at a specific wavelength, this may indicate that the solder is not fully fused or there are other quality issues. The convolutional neural network has the ability to capture such subtle changes and encode them as part of the empty-spectrum feature map. Then, in order to further enhance the feature representation, a pyramid pooling module is adopted in the method to perform multi-scale feature fusion on the empty-spectrum feature map. This is because welding defects may occur at different scales, from macroscopic shape irregularities to microscopic material distribution differences, so a mechanism is needed to integrate this cross-scale information. The pyramid pooling module samples the feature map at multiple scales and combines the results to obtain a richer and more comprehensive multi-scale feature representation. Then, local binary pattern (LBP) encoding is applied to these multi-scale feature representations to generate an LBP feature map. LBP is an effective method for describing the local texture of an image. It can highlight the details of the welding surface and reveal potential problems even at very small scales. Based on the obtained LBP feature map, a Transformer network is then used to perform global context modeling. The Transformer was originally designed as a model architecture for natural language processing, but it is also applicable to computer vision tasks, especially when considering the relationships between different parts of an image. In the application scenario of welding quality detection, the Transformer can be used to capture the global context information within the welding area, that is, the interaction between the welding point and its surrounding environment. For example, whether a welding point is correctly bonded to the surrounding substrate or whether there is a risk of cracks due to stress concentration can be more accurately evaluated through global context modeling. Finally, this step will generate a global context feature vector, which condenses the comprehensive information about the welding area.However, such feature vectors can be very large, containing too much data, which is not conducive to subsequent rapid processing or storage. Therefore, the method also proposes to use local sensitive hashing (LSH) coding technology to compress the global context feature vector to generate a sparse feature vector. LSH coding is a technology that can reduce the data dimension while maintaining similarity. It enables even high-dimensional features to be effectively compressed into a sparse representation while retaining the main characteristics of the original features. The obtained sparse feature vector is not only easy to process but also can still accurately represent the key attributes of the welding area and can be used as a welding feature vector for the next step of analysis. Finally, based on the sparse feature vector, relationship modeling is performed to construct a welding feature map. This map is not just a static image but a dynamic relationship network, where each node represents a welding feature and the edges represent the associations between them. For example, welding strength distribution data, welding depth distribution data, and welding surface flatness data are all integrated into this map to form a complete description of welding quality. In this way, engineers can obtain a comprehensive perspective on the health status of the welding points, so as to take appropriate measures to improve the production process or solve the discovered problems, ensuring that each FPC sensing module can meet strict performance standards and guaranteeing the safety and reliability of the autonomous driving system.

[0091] In a specific embodiment, the dynamic feature mapping of the welding feature map by the map transformation network to obtain a welding feature descriptor includes:

[0092] Performing multi-resolution decomposition on the welding feature map through a preset map transformation network to obtain a set of welding feature sub-maps;

[0093] Performing temporal feature analysis on the set of welding feature sub-maps based on a bidirectional long short-term memory network to obtain a temporal feature sequence, and performing structural analysis on the temporal feature sequence through a graph attention network to obtain a welding dynamic feature map;

[0094] Performing Fourier spectrum analysis on the welding dynamic feature map to obtain a frequency domain feature spectrum;

[0095] Extracting key features from the frequency domain feature spectrum through an adaptive feature selection algorithm to obtain a welding feature description vector;

[0096] Performing topological structure coding on the welding feature description vector based on a graph isomorphism network to obtain a structured feature code;

[0097] Performing feature transformation and mapping on the structured feature code through a map mapping algorithm to obtain a welding feature descriptor; wherein, the welding feature descriptor includes interface bonding state, void distribution density, and metal interconnection strength.

[0098] Specifically, to deeply analyze the welding quality, a method of dynamically feature mapping the welding feature map through a spectral transformation network is proposed in the claims to obtain a more detailed and accurate welding feature descriptor. This process starts with inputting the welding feature map into a preset spectral transformation network, which first performs multi-resolution decomposition to obtain a series of welding feature sub-map sets. These sub-map sets not only represent the welding features at different scales but also provide a basis for subsequent temporal feature analysis. For example, in the manufacturing process of FPC (Flexible Printed Circuit) sensing modules, engineers may use this method to ensure that each welding point meets the optimal quality standard. When a welding feature map is fed into the spectral transformation network, the network decomposes it into multiple sub-maps according to different resolution levels. Each sub-map focuses on features at a specific scale, from the macroscopic overall shape to the microscopic surface texture, ensuring that all levels of welding features are fully considered. This step is crucial for identifying subtle defects that may be overlooked at a single scale, such as tiny voids or poor interface bonding. Next, to understand the changes in welding features over time, a Bidirectional Long Short-Term Memory network (BiLSTM) is introduced in the method to perform temporal feature analysis on the welding feature sub-map sets. BiLSTM is a neural network architecture particularly suitable for processing sequence data. It can capture the evolution pattern of welding features during the manufacturing process and generate a temporal feature sequence. At the same time, to further reveal the internal connections between these temporal features, a Graph Attention Network (GAT) is used to perform a structured analysis on the temporal feature sequence to obtain a welding dynamic feature map. The Graph Attention Network allows the model to focus on the most relevant feature nodes and their connections, thus better simulating the complex relationships within the welding area. For example, during the welding process, stages such as the melting, cooling, and solidification of the solder will all affect the final welding quality, and GAT can help us more precisely understand and predict the impact of these dynamic changes. Based on the obtained welding dynamic feature map, a Fourier spectral analysis is then performed on it. The Fourier transform is a technique for converting a signal from the time domain to the frequency domain. In this way, frequency domain feature spectra can be extracted from the welding dynamic feature map. This transformation enables us to examine welding features from the perspective of frequency, helping to discover periodic or quasi-periodic behaviors that are not easily detectable in the time domain. For example, if there are certain regular vibrations or temperature fluctuations during the welding process, these phenomena will appear as peaks at specific frequencies in the frequency domain feature spectra. Such information is very valuable for diagnosing welding problems because it can help engineers identify potential problem sources and take corresponding preventive measures. Subsequently, to refine the obtained frequency domain feature spectra, an adaptive feature selection algorithm is used to extract key features from the frequency domain feature spectra, and then a welding feature description vector is obtained. The adaptive feature selection algorithm aims to find the key features that can best represent the welding quality and remove redundant or irrelevant information.This means that the final welding feature description vector will only contain those attributes that are crucial for evaluating welding quality, such as the interface bonding state, void distribution density, and metal interconnect strength. Doing so can not only improve the efficiency of subsequent analysis but also enhance the reliability of the results. Finally, to fully represent the information in the welding feature description vector, the method proposes using a Graph Isomorphism Network (GIN) to encode the topological structure of the welding feature description vector, obtaining a structured feature encoding. A Graph Isomorphism Network is a deep learning model for learning graph-structured data that can encode node features while preserving the topological properties of the graph. In this way, even in the face of complex welding feature graphs, the internal structural characteristics can be accurately captured. Then, through a graph mapping algorithm, feature transformation and mapping are performed on the structured feature encoding, and finally, a welding feature descriptor is obtained. The welding feature descriptor is a highly condensed representation form that comprehensively reflects various key attributes of the welding area, including but not limited to the interface bonding state, void distribution density, and metal interconnect strength. This not only provides a solid foundation for the quantitative evaluation of welding quality but also provides valuable guidance for optimizing welding process parameters, ensuring the reliability and performance of the FPC sensing module, especially in application scenarios such as autonomous driving sensors where high safety requirements are imposed.

[0099] In a specific embodiment, the topological structure analysis of the welding area based on the welding feature descriptor to obtain a welding correlation graph includes:

[0100] Performing local structure deconstruction on the welding feature descriptor through a preset welding distribution analysis algorithm to obtain a local feature distribution map of the welding area, and performing feature connection analysis on the local feature distribution map to obtain welding feature local connection data; wherein the welding feature local connection data includes the position information, connection weight information, and node type information of the welding nodes;

[0101] Based on the welding feature local connection data, global network modeling of the welding area is performed to obtain a global topological structure map of the welding area, and feature mapping is performed on the global topological structure map to obtain an embedded topological feature vector;

[0102] Performing dynamic adjacency relationship inference on the embedded topological feature vector through a topological optimization network to obtain dynamic feature correlation data between welding areas, and performing feature aggregation on the dynamic feature correlation data to obtain a welding clustering feature distribution map, which includes spatial correlation information of welding population structures and defect features;

[0103] Performing geometric structure decomposition on the welding clustering feature distribution map through a geometric correlation analysis algorithm to obtain a geometric correlation map of the welding area, and performing high-order feature mapping on the geometric correlation map to obtain a welding geometric feature description map;

[0104] Based on the welding geometric feature description diagram, a multi-layer relationship model of the dynamic characteristics of the welding area is established to obtain a welding dynamic feature diagram;

[0105] Through a high-order feature fusion algorithm, feature joint encoding is performed on the welding dynamic feature diagram and the geometric association spectrum to obtain a welding association diagram.

[0106] Specifically, in order to conduct a comprehensive and detailed topological structure analysis of the welding area, a method based on welding feature descriptors is proposed in the claims. The aim is to obtain a welding correlation graph that can accurately reflect the welding quality through multi-step processing and modeling. This process begins with the local structure deconstruction of the welding feature descriptors by a preset welding distribution analysis algorithm to obtain the local feature distribution map of the welding area, and further conducts feature connection analysis on these local feature distribution maps to obtain the local connection data of welding features. In the application scenario of the FPC (Flexible Printed Circuit) sensing module, engineers can use this method to ensure that each welding point meets the best quality standards. Specifically, when the welding feature descriptor is fed into the welding distribution analysis algorithm, the algorithm will decompose the complex welding features into multiple local structures and generate a local feature distribution map. Each local feature distribution map details the position information, connection weight information, and node type information of the welding nodes, which enables engineers to clearly see the specific conditions of each welding point and its surrounding environment. For example, during the manufacturing process of autonomous driving sensors, if the solder distribution around a certain welding point is uneven or there is a situation of false soldering, these problems will be revealed in the local feature distribution map. Next, by conducting feature connection analysis on these local feature distribution maps, the connection relationship between different welding nodes, that is, the local connection data of welding features, can be determined. This step helps to reveal the physical connection strength and stability between welding points. Based on the obtained local connection data of welding features, the next step is to conduct global network modeling on the welding area to construct the global topological structure diagram of the welding area. This global topological structure diagram not only shows the direct connections between welding points but also reflects the spatial layout and overall structure of the entire welding area. To better understand and analyze these complex relationships, feature mapping is performed on the global topological structure diagram to generate an embedded topological feature vector. The embedded topological feature vector is a compact and expressive data representation form that condenses the key information about the topological structure of the welding area, making the subsequent dynamic adjacency relationship inference more efficient and accurate. Then, through the topological optimization network, dynamic adjacency relationship inference is performed on the embedded topological feature vector to obtain the dynamic feature association data between welding areas. The topological optimization network is a deep learning model specifically designed to learn and predict the relationships between nodes in graph structure data. At this stage, the network will infer the dynamic relationships between them, such as whether there is stress concentration or heat conduction paths, based on the connection patterns and feature similarities between welding points. Subsequently, feature aggregation is performed on these dynamic feature association data to generate a welding clustering feature distribution map, which includes the spatial association information of the welding group structure and defect features.For example, certain welding points may form specific aggregation patterns due to the influence of material properties or process parameters. These patterns may be related to the welding quality, and the welding clustering feature distribution map can help identify these patterns so as to take targeted improvement measures. To further understand the geometric structure of the welding area, the geometric correlation analysis algorithm is used next to decompose the welding clustering feature distribution map to obtain the geometric correlation map of the welding area. The geometric correlation map provides detailed information about the spatial position, shape, and relative arrangement of the welding points, which is crucial for evaluating the flatness of the welding surface and the solder distribution. Then, by performing high-order feature mapping on the geometric correlation map, a welding geometric feature description map is generated. This map not only contains the geometric attributes of the welding points themselves but also integrates the interaction between them and the surrounding environment, providing a basis for deeper analysis. Finally, based on the welding geometric feature description map, a multi-layer relationship model of the dynamic characteristics of the welding area is established to obtain the welding dynamic feature map. This step involves complex interactions between various factors within the welding area, such as temperature changes, mechanical stress, and material aging. These dynamic characteristics have an important impact on the reliability during long-term use. Through the high-order feature fusion algorithm, feature joint coding is performed on the welding dynamic feature map and the geometric correlation map, and finally the welding correlation map is obtained. The welding correlation map is a comprehensive representation form that integrates the geometric structure and dynamic behavior of the welding area, and can provide engineers with a comprehensive and detailed basis for welding quality assessment, helping them quickly locate potential problems and optimize the production process to ensure that each FPC sensing module can meet strict performance standards, especially in application environments such as autonomous driving systems with extremely high requirements for safety.

[0107] In a specific embodiment, inputting the welding correlation map into a preset region growing algorithm for quality detection and grade division to obtain a welding quality assessment report includes:

[0108] Locating seed points of the welding correlation map through the region growing seed selection algorithm to obtain a set of seed points for the welding area, and calculating region growing feature data for the set of seed points for the welding area, where the region growing feature data includes welding connection strength, welding boundary gradient, and welding area morphological features;

[0109] Performing adaptive region expansion on the welding area based on the region growing feature data to obtain a welding area growth map, and extracting boundary features of the welding area growth map through a region boundary tracking algorithm to obtain a set of welding area boundary features;

[0110] Performing quality detection and grade division on the set of welding area boundary features through a preset quality assessment criterion to obtain a welding quality grade distribution map;

[0111] Perform deep feature analysis on the welding quality grade distribution map using a deep feature analysis network to obtain a welding quality feature vector, and perform hierarchical evaluation on the welding quality feature vector through a multi-level evaluation algorithm to obtain welding quality evaluation data; wherein, the welding quality evaluation data includes welding strength score, welding depth score, and welding surface score;

[0112] Perform structured processing and content optimization on the welding quality evaluation data through an evaluation report generation algorithm to obtain a welding quality evaluation report; wherein, the welding quality evaluation report includes quality grade distribution statistics, defect type analysis, and process parameter suggestions.

[0113] Specifically, in order to achieve quality inspection and grading of the welding area, a method is proposed in the claims. This method inputs the welding correlation graph into a preset region growing algorithm. This process first requires localizing the seed points of the welding correlation graph through a technique called the region growing seed selection algorithm. In application scenarios such as the manufacturing of flexible printed circuit (FPC) sensing modules, engineers can use this method to ensure that each welding point meets the best quality standards. Specifically, the region growing seed selection algorithm automatically identifies and selects key positions that can represent different welding characteristics as seed points, and these seed points form the set of seed points for the welding area. Once the positions of the seed points are determined, the next step is to calculate the region growing factors for this set of seed points for the welding area, thereby obtaining region growing characteristic data, including information such as welding connection strength, welding boundary gradient, and welding area morphological characteristics. For example, in the case of key welding points in the FPC module, if they show abnormal connection strength or irregular morphological characteristics, this information will be accurately recorded. When detailed region growing characteristic data is obtained, it becomes possible to perform adaptive region expansion on the welding area based on this data. Adaptive region expansion means dynamically determining the growth direction and range according to the characteristics around each seed point, and finally forming a welding area growth map. In this process, in order to more precisely understand the edge situation of the welding area, a region boundary tracking algorithm is also used to extract the boundary characteristics of the welding area growth map, thereby obtaining a set of welding area boundary characteristics. This set of characteristics contains information about the shape, continuity of the welding boundary, and any possible defects, which is crucial for subsequent quality assessment. For example, during the manufacturing process of an autonomous driving sensor, if there are cracks or discontinuities in the boundary of a certain welding point, these problems will be clearly shown in the set of welding area boundary characteristics, enabling engineers to discover and take corrective measures in a timely manner. Subsequently, the set of welding area boundary characteristics is analyzed according to preset quality assessment criteria to perform quality inspection and grading, and a welding quality grade distribution map is generated. This step involves a series of strict standards and thresholds for judging whether the welding quality meets the expectations. The welding quality grade distribution map not only shows the quality differences between different welding points but also provides a visual reference for subsequent improvement. For example, in an FPC module, some welding points may exhibit a low quality level due to improper process parameter settings, and it can be intuitively seen from the welding quality grade distribution map which areas need special attention. In order to further deeply understand the essence of welding quality, a deep feature analysis network is then used to perform deep feature analysis on the welding quality grade distribution map, thereby obtaining a welding quality feature vector. The deep feature analysis network is a complex machine learning model that can mine hidden patterns and rules from a large amount of data.The welding quality feature vector is the result of such analysis, which comprehensively reflects various aspects of welding quality, such as welding strength score, welding depth score, and welding surface score. By applying a multi-level evaluation algorithm to the welding quality feature vector, a comprehensive grading evaluation of welding quality can be carried out to obtain welding quality evaluation data. This evaluation data is a summary of the entire welding process, which quantifies different aspects of welding quality and helps engineers make more scientific and reasonable decisions. Finally, in order to make these evaluation results more practical and readable, the welding quality evaluation data is structurally processed and content optimized through an evaluation report generation algorithm, and finally a welding quality evaluation report is formed. This report is not just a collection of numbers and charts. More importantly, it includes the statistical distribution of quality grades, defect type analysis, and process parameter suggestions for the problems found. For example, in the production line of FPC sensing modules, if the welding quality evaluation report shows that there are generally problems with insufficient welding strength in a certain batch of products, then the process parameter suggestions given in the report can guide engineers to adjust welding temperature, time, or other relevant factors, thereby improving product quality and ensuring that each FPC sensing module leaving the factory can meet strict performance requirements, especially in application environments such as autonomous driving systems where safety requirements are extremely high.

[0114] In a specific embodiment, the adaptive region expansion of the welding region based on the region growth feature data to obtain a welding region growth map includes:

[0115] Performing dynamic threshold calculation on the region growth feature data through a preset region growth threshold self-adaptive algorithm to obtain a region growth threshold set, and locally optimizing the region growth threshold based on the welding region morphological features to obtain an optimized growth threshold set;

[0116] Performing multi-directional expansion calculation on the welding region based on the optimized growth threshold set to obtain a region expansion direction map, and performing connected domain marking on the expansion direction map through a region connectivity analysis algorithm to obtain a welding region connection marking map;

[0117] Performing growth constraint calculation on the welding region connection marking map through a preset region growth constraint algorithm to obtain region growth constraint data,

[0118] Performing multi-region competitive growth on the region growth constraint data based on a region competition mechanism to obtain a region competition growth map, and merging similar regions in the region competition growth map through a region merging algorithm to obtain a welding region growth map.

[0119] Specifically, in order to achieve adaptive region expansion of the welding area, a method based on region growth feature data is proposed in the claims. This method aims to generate a welding area growth map through a series of carefully designed steps. This process first involves performing dynamic threshold calculation on the region growth feature data by a preset region growth threshold adaptive algorithm to obtain a set of region growth thresholds. The key in this stage is to dynamically adjust the growth thresholds around each seed point according to information such as welding connection strength, welding boundary gradient, and welding area morphological features, ensuring that they can accurately reflect the changes in local welding quality. For example, during the manufacturing process of FPC (Flexible Printed Circuit) sensing modules, engineers may encounter significant quality differences between different welding points. For regions with high welding connection strength and regular morphological features, relatively loose growth thresholds can be set; while for regions that may have defects or irregular morphologies, more stringent thresholds are required to avoid misjudgment. Therefore, through the region growth threshold adaptive algorithm, the thresholds can be automatically adjusted according to specific welding characteristics, making the growth process more accurate and reliable. After determining the initial region growth thresholds, local optimization of these thresholds is also required based on the welding area morphological features to obtain an optimized set of growth thresholds. This step is to further improve the pertinence of the growth thresholds and ensure that the growth conditions around each welding point can be optimally configured. Next, multi-directional expansion calculation is performed on the welding area based on the optimized set of growth thresholds to generate a region expansion direction map. Multi-directional expansion calculation means starting from each seed point and gradually expanding along multiple possible directions until the set growth threshold is reached. During this process, the direction and distance of each expansion are recorded to form a detailed region expansion direction map. To ensure the connectivity between each expansion direction, a region connectivity analysis algorithm is used to process the expansion direction map to obtain a welding area connectivity label map. The region connectivity analysis algorithm can identify and label all interconnected regions, which can ensure that the finally generated welding area growth map not only covers all key welding points but also maintains the consistency and integrity of the overall structure. For example, in the manufacturing of autonomous driving sensors, if there are multiple connected solder joints around a certain welding point, then these solder joints will be correctly classified into the same connected domain for subsequent analysis. To control the expansion range of the welding area and prevent overgrowth or omission of important regions, next, growth constraint calculation is performed on the welding area connectivity label map through a preset region growth constraint algorithm to obtain region growth constraint data. The region growth constraint algorithm introduces a series of rules and limiting conditions, such as maximum growth radius, minimum connected area, etc., to ensure that the expansion of each welding area is within a reasonable range. These constraint conditions help to maintain the overall balance of the welding area and avoid affecting the normal evaluation of other parts due to over-expansion of some local regions.For example, when the solder distribution around a solder joint is particularly dense, without restraint, it may cause abnormal expansion in this area, masking other potential problems. Through the region growth constraint algorithm, this situation can be effectively avoided, making the evaluation of the entire welding area more scientific and reasonable. Finally, to simulate the welding situation in reality, a method based on the region competition mechanism is adopted to perform multi-region competitive growth on the region growth constraint data, obtaining a region competition growth map. The region competition mechanism mimics the competition relationship between different biological groups in nature, that is, adjacent welding regions will compete during the expansion process, and only those regions that better meet the growth conditions can continue to expand. This mechanism can better reflect the complex interactions in the actual welding process and improve the authenticity of the evaluation results. Subsequently, by merging similar regions in the region competition growth map, a welding region growth map is obtained. Similar region merging combines those connected domains with similar characteristics into larger regions, reducing redundant information and simplifying subsequent analysis work. For example, in the FPC sensing module, if the connections between several solder joints are very close and have similar characteristics, they will be merged into a larger welding region, which not only improves efficiency but also helps to more intuitively display the overall quality of the welding. In summary, through a series of steps such as dynamic threshold calculation, multi-directional expansion calculation, connectivity analysis, growth constraint calculation, and region competition growth on the region growth feature data, the adaptive expansion of the welding region is achieved, and finally a welding region growth map is obtained. This method not only improves the accuracy of welding quality evaluation but also provides a comprehensive and detailed tool for engineers to quickly locate problems and optimize the production process, ensuring that each FPC sensing module leaving the factory can meet strict quality standards, especially in application environments such as autonomous driving systems with extremely high requirements for safety.

[0120] The welding quality detection method for the high-order computing power autonomous driving FPC sensing module in the embodiments of the present invention has been described above. Next, the welding quality detection system for the high-order computing power autonomous driving FPC sensing module in the embodiments of the present invention will be described. Please refer to Figure 2 , an embodiment of the welding quality detection system for the high-order computing power autonomous driving FPC sensing module in the embodiments of the present invention includes:

[0121] An imaging module 21, configured to perform multi-angle scanning imaging on the welding area of the target FPC sensing module through a preset multi-spectral imaging system, obtaining a multi-dimensional spectral image set of the welding area;

[0122] An extraction module 22, configured to extract features from the multi-dimensional spectral image set through a preset convolutional neural network, obtaining a welding feature vector of the welding area, and constructing a welding feature map based on the welding feature vector;

[0123] A mapping module 23, configured to perform dynamic feature mapping on the welding feature map through a map transformation network to obtain a welding feature descriptor;

[0124] An analysis module 24, configured to perform topological structure analysis on the welding area based on the welding feature descriptor to obtain a welding association graph;

[0125] A partitioning module 25, configured to input the welding association graph into a preset region growing algorithm for quality grade partitioning to obtain a welding quality assessment report.

[0126] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to the description in the above method embodiment, and details are not described herein again.

[0127] Refer to Figure 3 , this embodiment of the present invention further provides a computer device, and its internal structure can be as Figure 3 shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface, and a database connected through a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program, when executed by the processor, implements the above method.

[0128] Those skilled in the art can understand that Figure 3 the structure shown in

[0129] is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0130] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium provided by the present invention and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0131] It should be noted that in this document, the terms "include", "comprise", or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that includes a series of elements includes not only those elements but also other elements not expressly listed, or elements that are inherent to such process, apparatus, article, or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, apparatus, article, or method that includes such element.

[0132] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A welding quality detection method for a high-order computing power autonomous driving FPC sensing module, characterized in that, Including the following steps: Performing multi-angle scanning imaging on the welding area of the target FPC sensing module through a preset multi-spectral imaging system to obtain a multi-dimensional spectral image set of the welding area; Performing feature extraction on the multi-dimensional spectral image set through a preset convolutional neural network to obtain a welding feature vector of the welding area, and constructing a welding feature map based on the welding feature vector; Performing dynamic feature mapping on the welding feature map through a map transformation network to obtain a welding feature descriptor; Performing topological structure analysis on the welding area based on the welding feature descriptor to obtain a welding association graph; Inputting the welding association graph into a preset region growing algorithm for quality detection and grade division to obtain a welding quality assessment report; The step of performing feature extraction on the multi-dimensional spectral image set through a preset convolutional neural network to obtain a welding feature vector of the welding area, and constructing a welding feature map based on the welding feature vector includes: Performing spatial-spectral feature extraction on the multi-dimensional spectral image set through a preset convolutional neural network to obtain a spatial-spectral feature map; wherein, the spatial-spectral feature map includes spatial texture information and spectral component information of the welding area; Performing multi-scale feature fusion on the spatial-spectral feature map through a pyramid pooling module to obtain a multi-scale feature representation, and performing local binary pattern encoding on the multi-scale feature representation to obtain an LBP feature map; Performing global context modeling on the LBP feature map based on a Transformer network to obtain a global context feature vector; Performing feature compression on the global context feature vector through a locality-sensitive hashing encoding technique to obtain a sparse feature vector, and using the sparse feature vector as the welding feature vector of the welding area; Performing relationship modeling based on the sparse feature vector to obtain a welding feature relationship graph, and constructing a welding feature map based on the welding feature relationship graph; wherein, the welding feature map includes welding strength distribution data, welding depth distribution data, and welding surface flatness data.

2. The welding quality detection method of the high-order computing power automatic driving FPC sensing module according to claim 1, wherein The step of performing multi-angle scanning imaging on the welding area of the target FPC sensing module through a preset multi-spectral imaging system to obtain a multi-dimensional spectral image set of the welding area includes: Performing full-spectrum scanning on the welding area through a preset multi-spectral imaging system to obtain an original spectral data sequence, wherein the original spectral data sequence includes welding information of visible spectrum, near-infrared spectrum, and ultraviolet spectrum; Performing band separation on the original spectral data sequence to obtain a multi-band spectral feature map, and performing spectral enhancement processing on the multi-band spectral feature map to obtain an enhanced spectral feature map; Performing three-dimensional reconstruction based on the enhanced spectral feature map to obtain a welding three-dimensional spectral map; Performing spatial registration on the welding three-dimensional spectral map based on a spectral registration network to obtain a registered spectral data set, and performing feature fusion on the registered spectral data set through a spectral feature fusion algorithm to obtain a fused spectral feature map; Perform endmember analysis on the fused spectral feature map through spectral unmixing algorithm to obtain the distribution map of welding material components, and perform geometric correction on the distribution map of welding material components to obtain the corrected component map; wherein, the distribution map of welding material components includes solder composition, substrate characteristics, and interface structure information; Perform dimensionality reduction and denoising processing on the corrected component map to obtain the multi-dimensional spectral image set of the welding area.

3. The welding quality detection method for the high-order computing power autonomous driving FPC sensing module according to claim 1, wherein, The dynamic feature mapping of the welding feature map through the map transformation network to obtain the welding feature descriptor includes: Perform multi-resolution decomposition on the welding feature map through a preset map transformation network to obtain a set of welding feature sub-maps; Based on the bidirectional long short-term memory network, perform temporal feature analysis on the set of welding feature sub-maps to obtain a temporal feature sequence, and perform structured analysis on the temporal feature sequence through a graph attention network to obtain a welding dynamic feature map; Perform Fourier spectral analysis on the welding dynamic feature map to obtain a frequency domain feature spectrum; Extract key features from the frequency domain feature spectrum through an adaptive feature selection algorithm to obtain a welding feature description vector; Based on the graph isomorphism network, perform topological structure encoding on the welding feature description vector to obtain a structured feature encoding; Perform feature transformation and mapping on the structured feature encoding through a map mapping algorithm to obtain a welding feature descriptor; wherein, the welding feature descriptor includes interface bonding state, void distribution density, and metal interconnect strength.

4. The welding quality detection method of the high-order computing power automatic driving FPC sensing module according to claim 1, characterized in that The topological structure analysis of the welding area based on the welding feature descriptor to obtain a welding association graph includes: Perform local structure deconstruction on the welding feature descriptor through a preset welding distribution analysis algorithm to obtain the local feature distribution map of the welding area, and perform feature connection analysis on the local feature distribution map to obtain welding feature local connection data; wherein, the welding feature local connection data includes the position information, connection weight information, and node type information of the welding nodes; Based on the welding feature local connection data, perform global network modeling on the welding area to obtain the global topological structure map of the welding area, and perform feature mapping on the global topological structure map to obtain an embedded topological feature vector; Perform dynamic adjacency relationship inference on the embedded topological feature vector through a topological optimization network to obtain dynamic feature association data between welding areas, and perform feature aggregation on the dynamic feature association data to obtain a welding clustering feature distribution map, which includes spatial association information of welding population structure and defect characteristics; Perform geometric structure decomposition on the welding clustering feature distribution map through a geometric association analysis algorithm to obtain a geometric association map of the welding area, and perform high-order feature mapping on the geometric association map to obtain a welding geometric feature description map; Based on the welding geometric feature description map, perform multi-layer relationship modeling on the dynamic characteristics of the welding area to obtain a welding dynamic feature map; Perform feature joint encoding on the welding dynamic feature map and the geometric association map through a high-order feature fusion algorithm to obtain a welding association graph.

5. The welding quality detection method of the high-order computing power autonomous driving FPC sensing module according to claim 1, characterized in that Inputting the welding correlation diagram into a preset region growing algorithm for quality detection and grading to obtain a welding quality assessment report, including: Locating seed points of the welding correlation diagram through a region growing seed selection algorithm to obtain a set of seed points for the welding region, and calculating region growing factor for the set of seed points for the welding region to obtain region growing feature data; wherein, the region growing feature data includes welding connection strength, welding boundary gradient, and welding region morphological features; Based on the region growing feature data, adaptively expanding the welding region to obtain a welding region growth diagram, and extracting boundary features of the welding region growth diagram through a region boundary tracking algorithm to obtain a set of welding region boundary features; Performing quality detection and grading on the set of welding region boundary features through a preset quality assessment criterion to obtain a welding quality grade distribution diagram; Performing deep feature analysis on the welding quality grade distribution diagram based on a deep feature analysis network to obtain a welding quality feature vector, and performing hierarchical evaluation on the welding quality feature vector through a multi-level evaluation algorithm to obtain welding quality assessment data; wherein, the welding quality assessment data includes welding strength score, welding depth score, and welding surface score; Performing structured processing and content optimization on the welding quality assessment data through an evaluation report generation algorithm to obtain a welding quality assessment report; wherein, the welding quality assessment report includes quality grade distribution statistics, defect type analysis, and process parameter suggestions.

6. The method for detecting the welding quality of the high-order computing power autonomous driving FPC sensing module according to claim 5, wherein, The adaptively expanding the welding region based on the region growing feature data to obtain a welding region growth diagram includes: Performing dynamic threshold calculation on the region growing feature data through a preset region growing threshold adaptive algorithm to obtain a set of region growing thresholds, and locally optimizing the region growing thresholds based on the welding region morphological features to obtain an optimized set of growth thresholds; Performing multi-directional expansion calculation on the welding region based on the optimized set of growth thresholds to obtain a region expansion direction diagram, and performing connected domain marking on the expansion direction diagram through a region connectivity analysis algorithm to obtain a welding region connectivity marking diagram; Performing growth constraint calculation on the welding region connectivity marking diagram through a preset region growing constraint algorithm to obtain region growing constraint data, Performing multi-region competitive growth on the region growing constraint data based on a region competition mechanism to obtain a region competition growth diagram, and merging similar regions of the region competition growth diagram through a region merging algorithm to obtain a welding region growth diagram.

7. A welding quality detection system for a high-order computing power autonomous driving FPC sensing module, characterized in that, Including: An imaging module, configured to perform multi-angle scanning imaging on the welding region of the target FPC sensing module through a preset multi-spectral imaging system to obtain a multi-dimensional spectral image set of the welding region; An extraction module, configured to extract features of the multi-dimensional spectral image set through a preset convolutional neural network to obtain a welding feature vector of the welding region, and constructing a welding feature map based on the welding feature vector; A mapping module, configured to perform dynamic feature mapping on the welding feature map through a map transformation network to obtain a welding feature descriptor; An analysis module for performing topological structure analysis on the welding area based on the welding feature descriptor to obtain a welding association graph; A division module for inputting the welding association graph into a preset region growing algorithm for quality grade division to obtain a welding quality assessment report; The method of obtaining a welding feature vector of the welding area by performing feature extraction on the multi-dimensional spectral image set through a preset convolutional neural network and constructing a welding feature map based on the welding feature vector, includes: Performing spatial-spectral feature extraction on the multi-dimensional spectral image set through a preset convolutional neural network to obtain a spatial-spectral feature map; wherein, the spatial-spectral feature map includes spatial texture information and spectral component information of the welding area; Performing multi-scale feature fusion on the spatial-spectral feature map through a pyramid pooling module to obtain a multi-scale feature representation, and performing local binary pattern encoding on the multi-scale feature representation to obtain an LBP feature map; Performing global context modeling on the LBP feature map based on a Transformer network to obtain a global context feature vector; Performing feature compression on the global context feature vector through a locality-sensitive hashing coding technique to obtain a sparse feature vector, and using the sparse feature vector as the welding feature vector of the welding area; Performing relationship modeling based on the sparse feature vector to obtain a welding feature relationship graph, and constructing a welding feature map based on the welding feature relationship graph; wherein, the welding feature map includes welding strength distribution data, welding depth distribution data, and welding surface flatness data.

8. A computer device, comprising a memory and a processor, wherein a computer program is stored in the memory, characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images