Processing quality monitoring system and method for precision part with complex structure

Abstract

The invention provides a processing quality monitoring system and method for a precision part with a complex structure. The method comprises the following steps: generating a fusion feature map through a multispectral image sequence after a target precision part is processed; identifying spatial distribution features of the suspected processing defect area by fusing texture features in the feature map and spectral reflection characteristic distribution in combination with a preset processing defect knowledge map; controlling a three-dimensional line scanning module to perform three-dimensional point cloud reconstruction on the suspected processing defect area based on the spatial distribution characteristics to generate three-dimensional shape data; extracting a morphology deviation field of the suspected machining defect area according to the three-dimensional morphology data and a reference CAD model of the target precision part; and processing quality classification is carried out on the target precision part based on the local surface instability features in the morphology deviation field and the multispectral feature vectors in the fusion feature map. According to the technical scheme provided by the invention, the problem of monitoring efficiency reduction caused by full-field scanning of point cloud data in precision part quality monitoring can be solved.

Description

Technical Field

[0001] This application relates to the field of machining quality monitoring technology, and more specifically, to a machining quality monitoring system and method for precision parts with complex structures. Background Technology

[0002] The current processing and manufacturing industry is upgrading towards high precision and intelligence. Processing quality monitoring has become a core link in ensuring product performance and production efficiency. Traditional manual sampling inspection has problems such as low efficiency, limited coverage, and susceptibility to subjective factors, making it difficult to meet the real-time and accuracy requirements of modern production lines. With the popularization of Industry 4.0 technology, technologies such as machine vision, sensors, and big data analysis are gradually being integrated into the monitoring process, providing stable and reliable quality assurance for high-precision and mass production.

[0003] In existing machining quality monitoring, the core principles of data acquisition, analysis and judgment, and result feedback are mainly used. First, physical signal acquisition is carried out, using sensors to capture key data in real time during the machining process. Next, the acquired key data is processed and analyzed to identify machining anomalies. Finally, the results of machining anomalies are fed back to achieve machining quality monitoring. However, in machining quality monitoring of precision parts with complex structures, although 3D full-field scanning can obtain complete morphological information of precision parts, it generates hundreds of millions or even billions of point cloud data when dealing with precision parts with complex shapes and structures (i.e., complex curved surfaces and fine features) and extremely high precision requirements. This results in subsequent point cloud registration and quality defect identification failing to meet the efficiency requirements of quality monitoring. Therefore, how to solve the problem of reduced monitoring efficiency caused by full-field scanning point cloud data in precision parts quality monitoring has become a difficult problem for the industry. Summary of the Invention

[0004] This application provides a machining quality monitoring system and method for precision parts with complex structures, which can solve the problem of decreased monitoring efficiency caused by full-field scanning point cloud data in the quality monitoring of precision parts.

[0005] In a first aspect, this application provides a method for monitoring the machining quality of precision parts with complex structures, comprising the following steps:

[0006] A multispectral image sequence of the target precision part after processing is acquired, and feature fusion is performed on the multispectral image sequence to generate a fused feature map characterizing the surface features of the target precision part;

[0007] Extract the texture features and spectral reflectance distribution from the fused feature map, and identify the spatial distribution features of suspected processing defect areas by combining the texture features and the spectral reflectance distribution with a preset processing defect knowledge graph;

[0008] Based on the spatial distribution characteristics of the suspected processing defect area, the three-dimensional line scanning module is controlled to perform local three-dimensional point cloud reconstruction of the suspected processing defect area to generate sub-pixel level three-dimensional shape data.

[0009] The three-dimensional topography data is non-rigidly registered with the reference CAD model of the target precision part, and the three-dimensional spatial deviation vector of each point cloud data point in the three-dimensional topography data relative to the reference CAD model is determined, thereby generating the topography deviation field of the suspected machining defect area.

[0010] The machining quality of the target precision parts is classified based on the local surface instability features of the suspected machining defect area in the morphological deviation field and the multispectral feature vector in the fused feature map.

[0011] In some embodiments, performing feature fusion on the multispectral image sequence to generate a fused feature map characterizing the surface features of the target precision part specifically includes:

[0012] Spatial registration is performed on the monitoring images of each band in the multispectral image sequence to obtain a registered image sequence;

[0013] Principal component analysis algorithm is used to perform feature dimensionality reduction on each registered image in the registered image sequence to obtain the dimensionality reduction feature matrix corresponding to each registered image;

[0014] A fused feature map representing the surface features of the target precision part is generated based on all the reduced-dimensional feature matrices.

[0015] In some embodiments, extracting the texture features and spectral reflectance distribution from the fused feature map specifically includes:

[0016] The fused feature map is subjected to multi-directional filtering to generate a multi-directional filtered response map;

[0017] Texture features of the fused feature map are extracted based on the multi-directional filter response map;

[0018] The fused feature map is subjected to spectral channel separation to obtain the gray value matrix of each spectral channel;

[0019] The spectral reflectance vector of the corresponding pixel is calculated based on the gray value matrix of each spectral channel, and the spectral reflectance characteristic distribution of the fused feature map is constructed based on all the spectral reflectance vectors.

[0020] In some embodiments, identifying the spatial distribution features of suspected processing defect areas by combining the texture features and the spectral reflectance distribution with a preset processing defect knowledge graph specifically includes:

[0021] The texture features and the distribution of spectral reflectance characteristics are normalized to generate a normalized feature vector.

[0022] The normalized feature vector is matched with the defect feature vectors of various processing defect types in the preset processing defect knowledge graph to obtain a feature similarity set;

[0023] Based on the feature similarity set, the suspected machining defect areas of the target precision part are extracted;

[0024] Morphological filtering and connected component analysis are performed on the suspected processing defect areas to determine the spatial distribution characteristics of the suspected processing defect areas.

[0025] In some embodiments, based on the spatial distribution characteristics of the suspected processing defect area, controlling the 3D line scanning module to perform local 3D point cloud reconstruction of the suspected processing defect area to generate sub-pixel-level 3D topography data specifically includes:

[0026] Based on the spatial distribution characteristics of the suspected processing defect area, calculate the scanning path and parameter configuration of the three-dimensional line scan module;

[0027] Based on the scanning path and the parameter configuration, the three-dimensional line scanning module is controlled to perform layered scanning on the suspected processing defect area to obtain sub-pixel level depth images.

[0028] Distortion correction and coordinate transformation are performed on the depth image to generate initial 3D point cloud data;

[0029] The initial 3D point cloud data is denoised to generate subpixel-level 3D shape data.

[0030] In some embodiments, non-rigid registration of the three-dimensional topography data with the reference CAD model of the target precision part specifically includes:

[0031] An initial corresponding point pair is constructed based on the three-dimensional topographic data and the benchmark CAD model;

[0032] The initial corresponding point pairs are rigidly registered using the iterative nearest point algorithm to obtain the initial registration transformation matrix;

[0033] Construct a non-rigid transformation model based on the initial registration transformation matrix;

[0034] Based on the aforementioned non-rigid transformation model, the three-dimensional topography data is iteratively optimized with the goal of minimizing the point cloud distance error, thereby completing the non-rigid registration of the three-dimensional topography data with the benchmark CAD model.

[0035] In some embodiments, a multispectral image sequence of the target precision part after processing is acquired using a multispectral camera.

[0036] Secondly, this application provides a machining quality monitoring system for precision parts with complex structures, used to execute a machining quality monitoring method for precision parts with complex structures. The system includes:

[0037] The acquisition module is used to acquire a multispectral image sequence of the target precision part after processing, and to perform feature fusion on the multispectral image sequence to generate a fused feature map characterizing the surface features of the target precision part;

[0038] The processing module is used to extract the texture features and spectral reflectance distribution from the fused feature map, and to identify the spatial distribution features of suspected processing defect areas by combining the texture features and the spectral reflectance distribution with a preset processing defect knowledge graph.

[0039] The processing module is also used to control the three-dimensional line scanning module to perform local three-dimensional point cloud reconstruction of the suspected processing defect area based on the spatial distribution characteristics of the suspected processing defect area, and generate sub-pixel level three-dimensional shape data.

[0040] The processing module is further configured to perform non-rigid registration of the three-dimensional topography data with the reference CAD model of the target precision part, and determine the three-dimensional spatial deviation vector of each point cloud data point in the three-dimensional topography data relative to the reference CAD model, thereby generating the topography deviation field of the suspected machining defect area.

[0041] The execution module is used to classify the machining quality of the target precision parts based on the local surface instability features of the suspected machining defect area in the topographic deviation field and the multispectral feature vector in the fused feature map.

[0042] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the above-described method for monitoring the machining quality of precision parts with complex structures.

[0043] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for monitoring the machining quality of precision parts with complex structures.

[0044] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0045] The processing quality monitoring system and method for complex-structured precision parts provided in this application first acquires a multispectral image sequence of the target precision part after processing, and performs feature fusion on the multispectral image sequence to generate a fused feature map characterizing the surface features of the target precision part; secondly, it extracts the texture features and spectral reflectance distribution from the fused feature map, and identifies the spatial distribution features of suspected processing defect areas by combining the texture features and spectral reflectance distribution with a preset processing defect knowledge graph; further, based on the spatial distribution features of the suspected processing defect areas, it controls a three-dimensional line scanning module to perform local three-dimensional point cloud reconstruction of the suspected processing defect areas, generating sub-pixel-level three-dimensional topography data; then, it performs non-rigid registration between the three-dimensional topography data and the reference CAD model of the target precision part, and determines the three-dimensional spatial deviation vector of each point cloud data point in the three-dimensional topography data relative to the reference CAD model, thereby generating a topography deviation field of the suspected processing defect areas; finally, it classifies the processing quality of the target precision part based on the local surface instability features of the suspected processing defect areas in the topography deviation field and the multispectral feature vector in the fused feature map.

[0046] Therefore, this application can solve the problem of decreased monitoring efficiency caused by full-field scanning point cloud data in precision parts quality monitoring. First, by acquiring multispectral image sequences of the target precision parts after processing and performing feature fusion to generate a fused feature map, it can effectively integrate surface information from different spectral bands, providing a more comprehensive surface feature basis for subsequent defect identification. Second, by extracting texture features and spectral reflectance distribution from the fused feature map and combining it with a processing defect knowledge graph to identify the spatial distribution features of suspected processing defect areas, it can accurately locate areas with potential defects from two dimensions: texture structure and material reflectance characteristics, reducing missed detections and false detections. Furthermore, based on the spatial distribution features of suspected areas, the three-dimensional line scanning module is controlled to perform local three-dimensional point cloud reconstruction to generate sub-pixel-level three-dimensional topography data, which can maintain the quality of the parts. This method ensures high precision while reducing redundant scanning, improving 3D reconstruction efficiency, and avoiding the problem that point cloud registration and quality defect identification caused by full-field 3D scanning cannot meet the efficiency requirements of quality monitoring. Then, the 3D topography data is non-rigidly registered with the benchmark CAD model to generate a topography deviation field, which can quantitatively characterize the geometric deviation between suspected areas and the design model, providing accurate 3D quantitative basis for defect assessment. Finally, based on the local surface instability characteristics of the topography deviation field and the multispectral feature vectors of the fused feature map, processing quality classification is performed, which can comprehensively consider geometric morphology and material properties to achieve accurate grading of the quality of processed parts, ensuring the comprehensiveness and accuracy of quality assessment. In summary, the technical solution provided in this application can solve the problem of decreased monitoring efficiency caused by full-field scanning point cloud data in precision parts quality monitoring. Attached Figure Description

[0047] Figure 1This is an exemplary flowchart of a machining quality monitoring method for precision parts with complex structures, according to some embodiments of this application;

[0048] Figure 2 This is an exemplary flowchart illustrating the determination of spatial distribution characteristics according to some embodiments of this application;

[0049] Figure 3 This is a schematic diagram of the machining quality monitoring system for precision parts with complex structures, as shown in some embodiments of this application;

[0050] Figure 4 This is a schematic diagram of the structure of a computer device for implementing a method for monitoring the machining quality of precision parts with complex structures, according to some embodiments of this application. Detailed Implementation

[0051] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] refer to Figure 1 This figure is an exemplary flowchart of a machining quality monitoring method for precision parts with complex structures, according to some embodiments of this application. The figure mainly includes the following steps:

[0053] In step S101, a multispectral image sequence of the target precision part after processing is obtained, and feature fusion is performed on the multispectral image sequence to generate a fused feature map characterizing the surface features of the target precision part.

[0054] In specific implementation, a multispectral image sequence of the target precision part after processing can be obtained through a multispectral camera. In addition, in other embodiments, a multispectral image sequence of the target precision part after processing can also be obtained through other spectral sensors. This is not limited here. The multispectral image sequence is obtained by combining multiple band monitoring images. The band monitoring images refer to the visualized images that present monitoring information related to the processing quality of precision parts under different spectral bands, which are the core basis for subsequent quality analysis and judgment.

[0055] In some embodiments, feature fusion of the multispectral image sequence to generate a fused feature map characterizing the surface features of the target precision part is specifically achieved through the following steps:

[0056] Spatial registration is performed on the monitoring images of each band in the multispectral image sequence to obtain a registered image sequence;

[0057] Principal component analysis algorithm is used to perform feature dimensionality reduction on each registered image in the registered image sequence to obtain the dimensionality reduction feature matrix corresponding to each registered image;

[0058] A fused feature map representing the surface features of the target precision part is generated based on all the reduced-dimensional feature matrices.

[0059] In specific implementation, firstly, the band image with the highest signal-to-noise ratio in the multispectral image sequence is selected as the reference image. For the remaining band monitoring images, an 8×8 pixel template window is set, and the cross-correlation coefficient between the remaining band monitoring images and the reference image in the template window area is calculated. The spatial offset corresponding to the maximum value of the cross-correlation coefficient is determined. Based on the spatial offset, the monitoring images of each band are translated and corrected (that is, the spatial position of each pixel in the band monitoring image is added to the spatial offset), thus obtaining all the registered images. All the registered images are combined in their original order to obtain the registered image sequence. The details are not elaborated here. Here, the registered image refers to the registered band monitoring image, and spatial registration refers to the process of correcting the spatial position deviation between different images to ensure that pixels corresponding to the same target area in the image are at the same coordinate position. Secondly, the covariance matrix of each registered image in the registered image sequence is calculated, and the covariance matrix is ​​processed using the Jacobi iteration method. Eigenvalue decomposition is performed to obtain eigenvalues ​​and corresponding eigenvectors. After sorting the eigenvalues ​​from largest to smallest, the top 3 eigenvectors (i.e., retaining more than 95% of the original data information) are selected as principal component directions. The two-dimensional data matrix corresponding to the registered image is multiplied by the selected 3 eigenvectors to obtain the dimensionality-reduced feature matrix corresponding to each registered image. The two-dimensional data matrix refers to the pixel matrix of the registered image expanded row by row, and the dimensionality-reduced feature matrix refers to the dimensionality-reduced registered image. Finally, the maximum and minimum pixel values ​​at the same spatial location are extracted from the dimensionality-reduced feature matrices corresponding to all registered images. The original pixel values ​​are mapped to the grayscale range of 0-255 by combining the maximum and minimum pixel values ​​with a linear stretching formula, and the pixel values ​​at the original spatial locations are replaced to obtain a fused feature map representing the surface features of the target precision part. The linear stretching formula is: normalized grayscale value = (original element value - minimum pixel value) / (maximum pixel value - minimum pixel value) × 255).

[0060] It should be noted that the fusion feature map in this application refers to a single image that integrates the effective information of each band of a multispectral image. By determining the fusion feature map, the problem of high analysis complexity caused by the dispersion of multi-band information is solved. Furthermore, feature enhancement makes the key features of the part surface clearer and more identifiable, providing a unified, efficient and accurate visual and data foundation for subsequent core analysis steps such as precision part surface defect identification, texture feature extraction, and processing quality assessment.

[0061] In step S102, the texture features and spectral reflectance distribution in the fused feature map are extracted, and the spatial distribution features of suspected processing defect areas are identified by combining the texture features and the spectral reflectance distribution with a preset processing defect knowledge graph.

[0062] In some embodiments, the extraction of texture features and spectral reflectance distribution from the fused feature map is specifically achieved through the following steps:

[0063] The fused feature map is subjected to multi-directional filtering to generate a multi-directional filtered response map;

[0064] Texture features of the fused feature map are extracted based on the multi-directional filter response map;

[0065] The fused feature map is subjected to spectral channel separation to obtain the gray value matrix of each spectral channel;

[0066] The spectral reflectance vector of the corresponding pixel is calculated based on the gray value matrix of each spectral channel, and the spectral reflectance characteristic distribution of the fused feature map is constructed based on all the spectral reflectance vectors.

[0067] In specific implementation, firstly, when performing multi-directional filtering on the fused feature map, a Gabor filter bank (containing filters in the 0°, 45°, 90°, and 135° directions) is selected. The Gabor filter in each direction is convolved with the fused feature map to obtain a multi-directional filtering response map. This multi-directional filtering response map refers to the image after filtering the fused feature map through filters in different directions. Secondly, when extracting texture features based on the multi-directional filtering response map, the gray-level co-occurrence matrix method is used to calculate the second-order angular moment, contrast, and entropy statistics at different directions and distances. These statistics are then combined to form the texture features of the fused feature map. These texture features refer to features that characterize the spatial distribution and combination patterns of pixel gray values ​​in the image. Finally, when performing spectral channel separation on the fused feature map, based on the original data of the multispectral imaging system… Spectral band segmentation involves separating the pixel grayscale values ​​corresponding to different spectral bands in the fused feature map, forming a two-dimensional matrix (i.e., the grayscale value matrix of each spectral channel) that corresponds one-to-one with each original spectral band. The grayscale value matrix of each spectral channel refers to a two-dimensional data matrix containing only the grayscale values ​​of all pixels under a single spectral band. Finally, when calculating the spectral reflectance vector of the corresponding pixel based on the grayscale value matrix of each spectral channel, the reflectance conversion coefficient of each spectral channel is first obtained by calibration using an existing standard whiteboard. Then, the grayscale value of each pixel in each spectral channel is multiplied by the conversion coefficient of the corresponding spectral channel to obtain the reflectance value of the corresponding pixel in each band. The vector formed by combining these reflectance values ​​is used to obtain the spectral reflectance vector of each pixel. The spectral reflectance vectors of all pixels are arranged according to their corresponding spatial positions to form the spectral reflectance characteristic distribution of the fused feature map.

[0068] It should be noted that the spectral reflectance distribution in this application refers to a dataset describing the reflectance variation of each pixel in the fused feature map under different spectral bands. The spectral reflectance distribution can reflect the differences in material composition in different areas of the part surface (such as the different spectral responses of different metal platings, oxide layers, or contaminants), and can be used to identify whether there are problems such as uneven material, local oxidation, or oil residue on the part surface. On the other hand, texture features focus on the microstructure of the part surface (such as the direction, density, and uniformity of the processing lines, as well as texture anomalies caused by defects such as micro-scratches and dents), and can accurately capture information such as whether the surface processing quality meets the standards and whether there is physical damage. The combination of the two avoids the limitations of relying solely on spectral data to judge structural defects or relying solely on texture features to distinguish material problems, and can form a comprehensive characterization of the part surface state, providing more reliable decision support for subsequent practical applications such as part surface quality grading, defect type determination, and processing technology optimization.

[0069] In some embodiments, reference Figure 2 As shown in the figure, this is an exemplary flowchart of determining spatial distribution features according to some embodiments of this application. In this embodiment, the spatial distribution features of suspected processing defect areas can be identified by combining the texture features and the spectral reflectance distribution with a preset processing defect knowledge graph through the following steps:

[0070] In step S1021, the texture features and the spectral reflectance distribution are normalized to generate a normalized feature vector.

[0071] In step S1022, the normalized feature vector is matched with the defect feature vectors of various processing defect types in the preset processing defect knowledge graph to obtain a feature similarity set;

[0072] In step S1023, suspected machining defect areas of the target precision part are extracted based on the feature similarity set;

[0073] In step S1024, morphological filtering and connected component analysis are performed on the suspected processing defect region to determine the spatial distribution characteristics of the suspected processing defect region.

[0074] In specific implementation, firstly, when normalizing the texture features and the spectral reflectance distribution, the min-max normalization method is used to map each statistical quantity (such as the second moment of the angle, contrast, and entropy) in the texture features and the reflectance values ​​of each band in the spectral reflectance distribution to the [0,1] interval, respectively, to obtain dimensionless normalized features. All normalized features are combined to form a normalized feature vector, which is the vector composed of the normalized texture features and the spectral reflectance distribution. Secondly, the Euclidean distance algorithm is used to calculate the feature similarity between the normalized feature vector and the defect feature vectors of various processing defects in the preset processing defect knowledge graph, to obtain a feature similarity set. The various defects include, but are not limited to, scratches, dents, and uneven materials. The defect feature vectors of various defects in the processing defect knowledge graph are specifically set based on simulation of historical normalized feature vectors using machine learning, and are not limited here. The feature similarity set includes feature similarities corresponding to different processing defect types. The feature similarity refers to the degree of similarity between the defect features detected in the current precision part and various defect features. Then, based on the feature similarity set, the suspected processing defect regions of the target precision part are extracted. That is, the processing defect type corresponding to the maximum feature similarity in the feature similarity set is extracted, the identification pixel interval corresponding to the processing defect type is called, the pixel value of each pixel in the aforementioned fused feature map is compared with the identification pixel interval, and the pixels falling into the identification pixel interval are taken as suspected processing defect points. Then, the region formed by all suspected processing defect points is taken as the suspected processing defect region of the target precision part. The identification pixel interval is specifically set by expert knowledge or simulation experiments according to the corresponding processing defect type, which is not limited. The candidate processing defect region refers to the set of pixel regions that are initially determined to have processing defects. Finally, when performing morphological filtering and connected component analysis on the candidate processing defect region, morphological opening operation is first used to remove noise points in the candidate processing defect region, and then the connected region corresponding to the candidate processing defect region is identified by the 8-neighbor connected component labeling algorithm. The spatial distribution characteristics of the suspected processing defect region are constituted by the center position and shape factor (which can be represented by the area of ​​the connected region).

[0075] It should be noted that the spatial distribution characteristics of suspected processing defect areas in this application refer to the information on the distribution of suspected processing defect areas in the image space. By determining the spatial distribution characteristics of suspected processing defect areas, the specific areas on the surface of precision parts that may have defects can be accurately located, avoiding the waste of efficiency caused by subsequent inspection only targeting the entire part. It also provides a clear target range for subsequent local 3D point cloud reconstruction, eliminating the need for redundant scanning of the entire surface of the part. Only the suspected areas pointed to by the spatial distribution characteristics need to be focused on, which greatly improves the efficiency and accuracy of 3D reconstruction and ensures the continuity and efficiency of the entire processing quality monitoring process from feature recognition to precise positioning and then to depth analysis.

[0076] In step S103, based on the spatial distribution characteristics of the suspected processing defect area, the three-dimensional line scanning module is controlled to perform local three-dimensional point cloud reconstruction of the suspected processing defect area to generate sub-pixel level three-dimensional shape data.

[0077] In some embodiments, based on the spatial distribution characteristics of the suspected processing defect area, controlling the 3D line scanning module to perform local 3D point cloud reconstruction of the suspected processing defect area to generate sub-pixel level 3D topography data is specifically implemented through the following steps:

[0078] Based on the spatial distribution characteristics of the suspected processing defect area, calculate the scanning path and parameter configuration of the three-dimensional line scan module;

[0079] Based on the scanning path and the parameter configuration, the three-dimensional line scanning module is controlled to perform layered scanning on the suspected processing defect area to obtain sub-pixel level depth images.

[0080] Distortion correction and coordinate transformation are performed on the depth image to generate initial 3D point cloud data;

[0081] The initial 3D point cloud data is denoised to generate subpixel-level 3D shape data.

[0082] In specific implementation, firstly, when calculating the scanning path and parameter configuration of the 3D line scanning module based on the spatial distribution characteristics of the suspected processing defect area, the center position in the spatial distribution characteristics is extracted as the scanning center of the 3D line scanning module. The processing defect point farthest from the scanning center in the suspected processing defect area is obtained, and the distance between the scanning center and the processing defect point is used as the scanning radius of the 3D line scanning module, thereby obtaining the scanning path of the 3D line scanning module. Then, the parameter configuration of the 3D line scanning module is set according to the sub-pixel level accuracy requirements. There is no limitation here. The parameter configuration includes scanning resolution (i.e., the distance between adjacent scanning lines), scanning speed (i.e., the acquisition frequency), and laser power (i.e., the laser power used to adapt to the reflection characteristics of the surface material of the processed part), forming the scanning path and parameter configuration of the 3D line scanning module. The scanning path refers to the path planning that guides the 3D line scanning module to complete the scanning of the target area, and the parameter configuration refers to the set of scanning parameters used to guide the 3D line scanning module to complete the scanning of the target area. Secondly, based on the scanning path and parameter configuration, the 3D line scanning module is controlled to perform layered scanning of the suspected processing defect area to obtain a sub-pixel level depth image. The depth image refers to the sub-pixel level depth image of the suspected processing defect area. The image is obtained through level recognition. Then, when generating initial 3D point cloud data by performing distortion correction and coordinate transformation on the depth image, the Zhang calibration method (calculating camera intrinsic parameters by pre-capturing images of a calibration board) is first used to correct the distortion of the depth image, eliminating depth errors caused by camera lens distortion. Then, based on the extrinsic parameters of the 3D line scan module (the positional relationship between the module and the world coordinate system), the two-dimensional coordinates (u, v) of each pixel in the corrected depth image are combined with the depth value z to convert it into three-dimensional coordinates (X, Y, Z) in the world coordinate system, thus obtaining the initial 3D point cloud data. The initial 3D point cloud data refers to the data after distortion correction. After correction and coordinate transformation, an initial point cloud set containing the surface three-dimensional coordinate information of the suspected processing defect area is generated. Finally, existing filters are used to denoise the initial three-dimensional point cloud data to generate sub-pixel level three-dimensional shape data. For example, a statistical filter can be used to denoise the initial three-dimensional point cloud data. This is not limited here. The statistical filter first calculates the average distance between each point cloud data point and several neighboring points in its neighborhood. Then, a standard deviation threshold is set according to the average distance of all points. Noise points with an average distance exceeding the threshold are removed, and valid point cloud data is retained. The denoised point cloud data is the sub-pixel level three-dimensional shape data.

[0083] It should be noted that the three-dimensional topography data in this application refers to point cloud data that reflects the three-dimensional morphology of the surface of the suspected machining defect area. The determination of the three-dimensional topography data transforms the suspected machining defect area from the planar features of the two-dimensional image into three-dimensional quantitative information with precise spatial dimensions. This not only overcomes the limitation that two-dimensional images cannot accurately represent the three-dimensional morphology of defects (such as the depth of depression, the height of protrusion, and the cross-sectional contour of scratches), but also provides a high-precision three-dimensional data foundation for subsequent registration with the benchmark CAD model and deviation calculation. This ensures that the evaluation of the machining quality of precision parts is upgraded from "qualitative judgment" to "quantitative analysis", effectively improving the accuracy and reliability of precision parts machining quality monitoring.

[0084] In step S104, the three-dimensional topography data is non-rigidly registered with the reference CAD model of the target precision part, and the three-dimensional spatial deviation vector of each point cloud data point in the three-dimensional topography data relative to the reference CAD model is determined, thereby generating the topography deviation field of the suspected machining defect area.

[0085] In some embodiments, the non-rigid registration of the three-dimensional topography data with the reference CAD model of the target precision part is specifically achieved through the following steps:

[0086] An initial corresponding point pair is constructed based on the three-dimensional topographic data and the benchmark CAD model;

[0087] The initial corresponding point pairs are rigidly registered using the iterative nearest point algorithm to obtain the initial registration transformation matrix;

[0088] Construct a non-rigid transformation model based on the initial registration transformation matrix;

[0089] Based on the aforementioned non-rigid transformation model, the three-dimensional topography data is iteratively optimized with the goal of minimizing the point cloud distance error, thereby completing the non-rigid registration of the three-dimensional topography data with the benchmark CAD model.

[0090] It should be noted that the reference CAD model in this application refers to a computer-aided design (CAD) model used as a reference standard in engineering scenarios such as three-dimensional geometric design, part machining quality inspection, or point cloud registration. It is a digital model built based on the part's design drawings, technical parameters, or theoretical geometric features, containing the part's precise three-dimensional geometric information (such as surface contours, dimensional parameters, feature structures, etc.) and used to compare with the actual acquired three-dimensional topographic data (such as point cloud data obtained through scanning) to evaluate the deviation between the actual part and the design target.

[0091] In specific implementation, firstly, when constructing initial corresponding point pairs based on 3D topographic data and a benchmark CAD model, a certain number of representative points are selected from the 3D topographic data point cloud using a uniform sampling algorithm. Then, the surface of the benchmark CAD model is meshed to obtain a corresponding number of model points. By calculating the spatial Euclidean distance between the representative points of the 3D topographic data and the model points, points with a distance less than a set threshold are paired to form initial corresponding point pairs. These initial corresponding point pairs refer to the set of point-to-point matching relationships initially established between the 3D topographic data and the benchmark CAD model. Secondly, an iterative nearest-point algorithm is used. When performing rigid registration on the initial corresponding point pairs, the rotation matrix and translation vector that minimize the spatial positional deviation between the two points are calculated based on the corresponding point pairs (the specific values ​​can be determined using the singular value decomposition method, which will not be elaborated here). The 3D topography data is then spatially transformed according to the rotation matrix and translation vector. The nearest neighbor points of the transformed 3D topography data points on the surface of the reference CAD model are then searched again to update the corresponding point pairs until the deviation of the corresponding point pairs converges. The obtained rotation matrix and translation vector are then combined to form the initial registration transformation matrix. The initial registration transformation matrix refers to the matrix that initially aligns the 3D topography data to the reference model through rigid transformation. The mathematical matrix of the CAD model coordinate system; then, when constructing a non-rigid transformation model based on the initial registration transformation matrix, the three-dimensional topography data after initial registration is used as the basis. An interpolation method based on radial basis functions is employed, using feature points (such as curvature points) in the three-dimensional topography data as control points. A spatial deformation function (specifically a radial basis function) is constructed based on the deviation between the control points and the corresponding points in the reference CAD model. This allows non-control points to be flexibly adjusted according to this spatial deformation function, forming a non-rigid transformation model. This non-rigid transformation model refers to a model capable of locally and flexibly adjusting the three-dimensional topography data relative to the reference CAD model. The mathematical model of rigid deformation can handle the non-rigid deformation of parts caused by processing. Finally, when iteratively optimizing based on the non-rigid transformation model with the goal of minimizing the point cloud distance error, the three-dimensional topography data is first substituted into the non-rigid transformation model for deformation. The shortest distance between the deformed data point cloud and the corresponding model point on the surface of the reference CAD model (i.e., the point cloud distance error) is calculated. Then, the point cloud distance error is reduced by adjusting the weight coefficient of the radial basis function until the point cloud distance error reaches the minimum value. At this time, the three-dimensional topography data and the reference CAD model achieve precise geometric fit, and the non-rigid registration is completed.

[0092] It should be noted that the non-rigid registration of the three-dimensional topography data and the benchmark CAD model in this application refers to the process of eliminating local geometric deviations caused by part processing deformation by flexibly adjusting the spatial shape of the three-dimensional topography data. Non-rigid registration can not only provide accurate positional correspondence for subsequent geometric error analysis, but also provide a reliable geometric alignment basis for part quality inspection, defect identification (such as local depressions and protrusions), and model correction in reverse engineering. This ensures that subsequent analysis or processing correction work is based on accurate shape matching, avoiding error misjudgment or correction deviation caused by rigid registration's inability to handle non-rigid deformation.

[0093] In some embodiments, determining the three-dimensional spatial deviation vector of each point cloud data point in the three-dimensional topography data relative to the reference CAD model, and then generating the topography deviation field of the suspected processing defect area, is specifically achieved through the following steps:

[0094] Spatial position mapping is performed on the non-rigidly registered 3D topography data and the benchmark CAD model to determine the nearest point on the benchmark CAD model for each point cloud data point in the 3D topography data.

[0095] Calculate the spatial coordinate difference between each point cloud data point and its nearest corresponding point to obtain the three-dimensional spatial deviation vector corresponding to each point cloud data point.

[0096] The three-dimensional spatial deviation vector of the suspected processing defect area is extracted, and the morphological deviation field of the suspected processing defect area is constructed from all the extracted three-dimensional spatial deviation vectors.

[0097] In specific implementation, firstly, when spatially mapping the non-rigidly registered 3D topography data with the reference CAD model to determine the nearest point of each point cloud data point on the reference CAD model, a k-neighbor search algorithm is used. The surface of the reference CAD model is first discretized into a triangular mesh. For each point cloud data point in the 3D topography data, the nearest triangular facet in the triangular mesh of the reference CAD model is searched. The perpendicular projection point of the point cloud data point to the triangular facet is taken as the nearest point of that point cloud data point on the reference CAD model. The nearest point refers to the point on the surface of the reference CAD model with the smallest spatial distance to the point cloud data point in the 3D topography data. Secondly, when calculating the spatial coordinate difference between each point cloud data point and its corresponding nearest point to obtain the 3D spatial deviation vector, the 3D coordinates of each point cloud data point are extracted. The coordinates (X1, Y1, Z1) and the three-dimensional coordinates (X2, Y2, Z2) of the nearest point are used to calculate the deviation values ​​in each coordinate axis direction according to the following method: X-direction deviation ΔX = X1 - X2, Y-direction deviation ΔY = Y1 - Y2, and Z-direction deviation ΔZ = Z1 - Z2. The ΔX, ΔY, and ΔZ are combined to form the three-dimensional spatial deviation vector corresponding to the point cloud data point. The three-dimensional spatial deviation vector refers to the vector describing the positional deviation of each point in the three-dimensional topography data relative to the nearest point of the reference CAD model in the three coordinate axis directions of X, Y, and Z. Finally, the three-dimensional spatial deviation vector of the suspected machining defect area is extracted, and the topography deviation field of the suspected machining defect area is constructed by all the extracted three-dimensional spatial deviation vectors. That is, the topography deviation field of the suspected machining defect area is formed by arranging all the extracted three-dimensional spatial deviation vectors according to their corresponding spatial positions.

[0098] It should be noted that, in this application, the topographic deviation field refers to the set of deviation vectors of each position in the suspected machining defect area relative to the reference CAD model. The topographic deviation field can clearly reflect the magnitude, direction and distribution pattern of the deviation of each position in the suspected machining defect area, which provides direct geometric data support for judging the severity of machining defects.

[0099] In step S105, the machining quality of the target precision part is classified based on the local surface instability features of the suspected machining defect area in the topographic deviation field and the multispectral feature vector in the fused feature map.

[0100] In some embodiments, classifying the machining quality of target precision parts based on the local surface instability features of suspected machining defect areas in the topography deviation field and the multispectral feature vectors in the fused feature map is specifically achieved through the following steps:

[0101] Local surface instability features of suspected processing defect areas are extracted from the morphological deviation field. These local surface instability features include surface roughness, curvature abrupt change rate, and indentation depth.

[0102] Extract multispectral feature vectors corresponding to suspected processing defect areas from the fused feature map. The multispectral feature vectors include the peak wavelength of spectral reflectance and band contrast.

[0103] A comprehensive feature matrix is ​​generated from the local surface instability characteristic parameters and the multispectral feature vectors;

[0104] The comprehensive feature matrix is ​​input into a pre-trained deep learning classification model, which outputs the machining quality category of the target precision part.

[0105] In specific implementation, firstly, when extracting local surface instability features (including surface roughness, curvature abrupt change rate, and depression depth) of suspected processing defect areas from the topographic deviation field, the arithmetic mean deviation method is used to calculate the arithmetic mean of the deviation values ​​of all point cloud data points in the suspected processing defect area as surface roughness. When extracting the curvature abrupt change rate, the local curvature of each point cloud data point in the suspected processing defect area is first calculated using principal component analysis. Then, the proportion of the number of times the local curvature difference between adjacent point cloud data points exceeds a set threshold to the total local curvature difference is counted as the curvature abrupt change rate. Point cloud data points in the suspected processing defect area with negative deviation values ​​(indicating a value lower than the baseline CAD model) are selected. The absolute value of the minimum deviation is taken as the indentation depth, and the surface roughness, curvature abrupt change rate, and indentation depth constitute the local surface instability features. All local surface instability features refer to the set of quantitative parameters used to describe the surface geometric anomalies of the suspected area. Secondly, when extracting the multispectral feature vector (including the peak wavelength of spectral reflectance and band contrast) corresponding to the suspected processing defect area from the fused feature map, the corresponding suspected processing defect area is extracted from the fused feature map, and the gray values ​​of each spectral channel within the suspected processing defect area are extracted. The spectral wavelength corresponding to the spectral channel with the largest gray value is taken as the peak wavelength of spectral reflectance, and the average gray value of the spectral channel corresponding to the peak wavelength of spectral reflectance and the adjacent spectral channels is calculated. The ratio of the difference in values ​​is used as the band contrast (i.e., the spectral channel corresponding to the peak wavelength of spectral reflectance is taken as the peak wavelength channel, and two directly adjacent spectral channels on the left and right sides of this peak wavelength channel are selected, respectively. The difference between the average gray value of the band monitoring image under the peak wavelength channel and the average gray value of the band monitoring image under the left adjacent channel, and the difference between the average gray value of the band monitoring image under the peak wavelength channel and the average gray value of the band monitoring image under the right adjacent channel are calculated. Then, these two differences are divided by the average gray value of the peak wavelength channel to obtain two ratios. The average of these two ratios is taken as the final band contrast. If the peak wavelength channel is located in all... If the leftmost spectral channel (without an adjacent left spectral channel) is the peak channel, then only the difference between the average gray value of the band monitoring image under the peak channel and the average gray value of the band monitoring image under the adjacent right channel is calculated and divided by the average gray value corresponding to the peak channel to obtain the band contrast. If the peak channel is located at the rightmost of all spectral channels (without an adjacent right channel), then only the difference between the average gray value of the band monitoring image under the peak channel and the average gray value of the band monitoring image under the adjacent left channel is calculated and divided by the average gray value corresponding to the peak channel to obtain the band contrast. The two are combined to form a multispectral feature vector, which is a vector describing the reflectance characteristics of the suspected area under different spectral bands.Then, the parameters of the local surface instability features (surface roughness, curvature abrupt change rate, indentation depth) and the components of the multispectral feature vector (reflection peak wavelength, band contrast) are arranged column-wise to generate a comprehensive feature matrix. This comprehensive feature matrix integrates the local surface instability features and the multispectral features. Finally, when the comprehensive feature matrix is ​​input into a pre-trained deep learning classification model to output the processing quality category, the pre-trained deep learning classification model uses a convolutional neural network (CNN). First, features are extracted from the comprehensive feature matrix through the convolutional layers of the deep learning classification model. Then, the feature dimensions are simplified through pooling layers. Finally, the extracted features are mapped to preset processing quality categories (such as acceptable, minor defects, and severe defects) through fully connected layers. The final output is the processing quality category corresponding to the target precision part. The pre-trained deep learning classification model is trained using a large amount of sample data labeled with processing quality categories (including a comprehensive feature matrix combining the local surface instability features and multispectral features of the part, and corresponding "acceptable," "minor defects," and "severe defects" category labels), which will not be elaborated further here.

[0106] It should be noted that the machining quality category of the target precision parts refers to the level of machining quality of the parts according to preset standards, such as qualified, minor defects, and serious defects.

[0107] Furthermore, in another aspect of this application, in some embodiments, this application provides a machining quality monitoring system for precision parts with complex structures, referring to... Figure 3 The figure is a schematic diagram of the machining quality monitoring system for precision parts with complex structures, according to some embodiments of this application. The machining quality monitoring system for precision parts with complex structures includes: an acquisition module 201, a processing module 202, and an execution module 203, which are described below:

[0108] The acquisition module 201 in this application is mainly used to acquire a multispectral image sequence after the target precision part has been processed, and to perform feature fusion on the multispectral image sequence to generate a fused feature map characterizing the surface features of the target precision part.

[0109] Processing module 202, in this application, is mainly used to extract the texture features and spectral reflectance distribution in the fused feature map, and to identify the spatial distribution features of suspected processing defect areas by combining the texture features and the spectral reflectance distribution with a preset processing defect knowledge graph;

[0110] The processing module 202 is also used to control the three-dimensional line scanning module to perform local three-dimensional point cloud reconstruction on the suspected processing defect area based on the spatial distribution characteristics of the suspected processing defect area, and generate sub-pixel level three-dimensional shape data.

[0111] In addition, the processing module 202 is also used to perform non-rigid registration of the three-dimensional topography data with the reference CAD model of the target precision part, and determine the three-dimensional spatial deviation vector of each point cloud data point in the three-dimensional topography data relative to the reference CAD model, thereby generating the topography deviation field of the suspected machining defect area.

[0112] The execution module 203 in this application is mainly used to classify the processing quality of the target precision parts based on the local surface instability features of the suspected processing defect area in the topographic deviation field and the multispectral feature vector in the fused feature map.

[0113] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the above-described method for monitoring the machining quality of precision parts with complex structures.

[0114] In some embodiments, reference Figure 4 The figure is a schematic diagram of the structure of a computer device for implementing a machining quality monitoring method for complex-structure precision parts, according to some embodiments of this application. The machining quality monitoring method for complex-structure precision parts in the above embodiments can... Figure 4 The computer device shown is used to implement this, and the computer device includes at least one processor 301, a communication bus 302, a memory 303, and at least one communication interface 304.

[0115] The processor 301 can be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more devices used to control the execution of the machining quality monitoring method for complex-structured precision parts in this application.

[0116] The communication bus 302 can be used to transmit information between the aforementioned components.

[0117] The memory 303 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 303 may exist independently and be connected to the processor 301 via the communication bus 302. The memory 303 may also be integrated with the processor 301.

[0118] The memory 303 stores program code for executing the solution of this application, and its execution is controlled by the processor 301. The processor 301 executes the program code stored in the memory 303. The program code may include one or more software modules. In the above embodiments, the determination of the machining quality monitoring method for complex-structured precision parts can be achieved through the processor 301 and one or more software modules in the program code in the memory 303.

[0119] Communication interface 304 uses any transceiver-like device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0120] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0121] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0122] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for monitoring the machining quality of precision parts with complex structures.

[0123] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0124] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for monitoring the machining quality of precision parts with complex structures, characterized in that, Includes the following steps: A multispectral image sequence of the target precision part after processing is acquired, and feature fusion is performed on the multispectral image sequence to generate a fused feature map characterizing the surface features of the target precision part; Extract the texture features and spectral reflectance distribution from the fused feature map, and identify the spatial distribution features of suspected processing defect areas by combining the texture features and the spectral reflectance distribution with a preset processing defect knowledge graph; Based on the spatial distribution characteristics of the suspected processing defect area, the three-dimensional line scanning module is controlled to perform local three-dimensional point cloud reconstruction of the suspected processing defect area to generate sub-pixel level three-dimensional shape data. The three-dimensional topography data is non-rigidly registered with the reference CAD model of the target precision part, and the three-dimensional spatial deviation vector of each point cloud data point in the three-dimensional topography data relative to the reference CAD model is determined, thereby generating the topography deviation field of the suspected machining defect area. The machining quality of the target precision parts is classified based on the local surface instability features of the suspected machining defect area in the morphological deviation field and the multispectral feature vector in the fused feature map.

2. The method as described in claim 1, characterized in that, The process of fusing features from the multispectral image sequence to generate a fused feature map characterizing the surface features of the target precision part specifically includes: Spatial registration is performed on the monitoring images of each band in the multispectral image sequence to obtain a registered image sequence; Principal component analysis algorithm is used to perform feature dimensionality reduction on each registered image in the registered image sequence to obtain the dimensionality reduction feature matrix corresponding to each registered image; A fused feature map representing the surface features of the target precision part is generated based on all the reduced-dimensional feature matrices.

3. The method as described in claim 1, characterized in that, Extracting the texture features and spectral reflectance distribution from the fused feature map specifically includes: The fused feature map is subjected to multi-directional filtering to generate a multi-directional filtered response map; Texture features of the fused feature map are extracted based on the multi-directional filter response map; The fused feature map is subjected to spectral channel separation to obtain the gray value matrix of each spectral channel; The spectral reflectance vector of the corresponding pixel is calculated based on the gray value matrix of each spectral channel, and the spectral reflectance characteristic distribution of the fused feature map is constructed based on all the spectral reflectance vectors.

4. The method as described in claim 1, characterized in that, The spatial distribution characteristics of suspected processing defect areas are identified by combining the texture features and the spectral reflectance distribution with a preset processing defect knowledge graph. Specifically, this includes: The texture features and the distribution of spectral reflectance characteristics are normalized to generate a normalized feature vector. The normalized feature vector is matched with the defect feature vectors of various processing defect types in the preset processing defect knowledge graph to obtain a feature similarity set; Based on the feature similarity set, the suspected machining defect areas of the target precision part are extracted; Morphological filtering and connected component analysis are performed on the suspected processing defect areas to determine the spatial distribution characteristics of the suspected processing defect areas.

5. The method as described in claim 1, characterized in that, Based on the spatial distribution characteristics of the suspected processing defect area, the 3D line scanning module is controlled to perform local 3D point cloud reconstruction of the suspected processing defect area to generate sub-pixel level 3D topography data, specifically including: Based on the spatial distribution characteristics of the suspected processing defect area, calculate the scanning path and parameter configuration of the three-dimensional line scan module; Based on the scanning path and the parameter configuration, the three-dimensional line scanning module is controlled to perform layered scanning on the suspected processing defect area to obtain sub-pixel level depth images. Distortion correction and coordinate transformation are performed on the depth image to generate initial 3D point cloud data; The initial 3D point cloud data is denoised to generate subpixel-level 3D shape data.

6. The method as described in claim 1, characterized in that, The non-rigid registration of the three-dimensional topographic data with the reference CAD model of the target precision part specifically includes: An initial corresponding point pair is constructed based on the three-dimensional topographic data and the benchmark CAD model; The initial corresponding point pairs are rigidly registered using the iterative nearest point algorithm to obtain the initial registration transformation matrix; Construct a non-rigid transformation model based on the initial registration transformation matrix; Based on the aforementioned non-rigid transformation model, the three-dimensional topography data is iteratively optimized with the goal of minimizing the point cloud distance error, thereby completing the non-rigid registration of the three-dimensional topography data with the benchmark CAD model.

7. The method as described in claim 1, characterized in that, Multispectral image sequences of precision parts after machining are obtained using a multispectral camera.

8. A machining quality monitoring system for precision parts with complex structures, used to execute the machining quality monitoring method for precision parts with complex structures as described in any one of claims 1 to 7, characterized in that, The system includes: The acquisition module is used to acquire a multispectral image sequence of the target precision part after processing, and to perform feature fusion on the multispectral image sequence to generate a fused feature map characterizing the surface features of the target precision part; The processing module is used to extract the texture features and spectral reflectance distribution from the fused feature map, and to identify the spatial distribution features of suspected processing defect areas by combining the texture features and the spectral reflectance distribution with a preset processing defect knowledge graph. The processing module is also used to control the three-dimensional line scanning module to perform local three-dimensional point cloud reconstruction of the suspected processing defect area based on the spatial distribution characteristics of the suspected processing defect area, and generate sub-pixel level three-dimensional shape data. The processing module is further configured to perform non-rigid registration of the three-dimensional topography data with the reference CAD model of the target precision part, and determine the three-dimensional spatial deviation vector of each point cloud data point in the three-dimensional topography data relative to the reference CAD model, thereby generating the topography deviation field of the suspected machining defect area. The execution module is used to classify the machining quality of the target precision parts based on the local surface instability features of the suspected machining defect area in the topographic deviation field and the multispectral feature vector in the fused feature map.

9. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the machining quality monitoring method for precision parts with complex structures as described in any one of claims 1 to 7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the machining quality monitoring method for precision parts with complex structures as described in any one of claims 1 to 7.

Images