A mold precision fitting machining quality on-line detection system and method

Abstract

This invention discloses an online inspection system and method for the machining quality of precision mold parts, relating to the fields of industrial online inspection and precision manufacturing quality control. It includes an optical data acquisition module based on a line laser profile measurement device. During the machining process of precision mold parts, it simultaneously acquires geometric contour optical data and surface imaging optical data of the same inspection area of ​​the precision mold part to be inspected, and associates the two types of optical data with unified acquisition timing information. In this invention, by simultaneously acquiring geometric contour optical data and surface imaging optical data during machining and unifying the acquisition timing information for the two types of data, the different optical inspection results remain consistent in both the temporal and spatial dimensions, providing a reliable data foundation for subsequent multi-source data collaborative analysis, thereby avoiding the information bias caused by a single inspection method.

Description

Technical Field

[0001] This invention relates to the field of industrial online inspection and precision manufacturing quality control technology, and in particular to an online inspection system and method for the processing quality of precision mold parts. Background Technology

[0002] Precision mold components typically undergo multiple stages during manufacturing, including roughing, semi-finishing, and finishing. The final machining quality directly impacts the mold assembly accuracy, service life, and consistency of the molded products. During the finishing stage, the geometric contour accuracy and surface condition of precision mold components are often simultaneously affected by various factors such as tool wear, machining vibration, material inhomogeneity, and environmental disturbances. This can easily lead to small but cumulative quality deviations. Therefore, online quality inspection of precision mold components during processing, and timely identification of potential quality anomalies, is a crucial technical step in ensuring processing stability and finished product consistency.

[0003] Current mold processing quality inspection methods often rely on a single optical approach or a single feature dimension for judgment. For example, they may rely solely on line laser contour measurement to analyze geometric shape changes or on industrial vision imaging to analyze surface texture and edge conditions. Since geometric contour changes and surface imaging changes differ in spatial distribution, scale characteristics, and change mechanisms, a single data source cannot fully reflect the true processing quality. In actual processing, minute geometric changes and surface anomalies are not necessarily synchronized, and vibration, noise, and sampling instability can easily introduce discrete points. Without effective alignment, suppression, and weighting mechanisms, the reliability of fusion analysis is low. Existing multi-source fusion schemes often use fixed scales or global analysis, which are difficult to adapt to the differences in the density of changes in different regions, leading to unstable consistency assessments and affecting the accuracy of online quality judgment. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing an online inspection system and method for the machining quality of precision mold parts.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: an online inspection system for the processing quality of precision mold parts, comprising: an optical data acquisition module, based on a line laser profile measurement device, which simultaneously acquires geometric profile optical data and surface imaging optical data of the same detection area of ​​the precision mold part to be inspected during the processing of the precision mold part, and associates the two types of optical data with unified acquisition timing information; a profile edge change data processing module, which performs profile continuity analysis on the geometric profile optical data to form profile change data characterizing the profile change trend, and marks candidate positions with significant profile changes in the profile change data; performs edge stability analysis on the surface imaging optical data to form edge change data characterizing the continuous state of the surface edges, and marks candidate areas with abnormal edge concentration in the edge change data; and a feature alignment position information module. The information generation module performs spatial alignment processing on candidate positions in the contour change data and candidate regions in the edge change data to generate feature alignment position information data for associating geometric contour changes and surface edge changes. The local enhancement data generation module performs local reconstruction processing on the geometric contour optical data and surface imaging optical data respectively under the constraint of the feature alignment position information data to generate local contour enhancement data and local imaging enhancement data for the corresponding feature regions. The consistency evaluation data generation module generates consistency evaluation data reflecting the degree of correlation between geometric changes and surface changes based on the consistency relationship between the local contour enhancement data and the local imaging enhancement data. The quality judgment module performs joint judgment processing on the geometric contour optical data and surface imaging optical data based on the consistency evaluation data and outputs the quality judgment results of the corresponding detection areas of the precision mold parts.

[0006] As a further description of the above technical solution: In the machining process of precision mold parts, the optical data acquisition module uses a line laser contour measuring device to project a laser beam onto the detection area of ​​the precision mold part to form a preliminary signal of geometric contour optical data. The reflected laser beam is received by an optical sensor, and the reflected image is acquired, forming the original geometric contour signal data. Noise filtering is performed on the original geometric contour signal data to remove noise caused by environmental factors or equipment errors, resulting in processed geometric contour signal data. The processed geometric contour signal data is then converted into two-dimensional or three-dimensional spatial coordinate data to form the final geometric contour optical data, representing the changes in the surface shape and geometric features of the precision mold part. The optical data acquisition module utilizes a line laser wheel... The profile measurement device acquires images of the area to be inspected, obtaining preliminary signals of surface imaging optical data. An imaging sensor then acquires image data of the area to be inspected, generating raw surface imaging signal data, which reflects the surface texture and edge features of the precision mold parts. Distortion correction processing is performed on the raw surface imaging signal data to eliminate image distortion caused by lens distortion or other factors. Brightness normalization processing adjusts the brightness distribution of the image to adapt to imaging requirements under different ambient lighting conditions. The distortion-corrected and brightness-normalized image data is then converted into two-dimensional image data, forming the final surface imaging optical data, representing the surface condition and detailed features of the precision mold parts.

[0007] As a further description of the above technical solution: The contour edge change data processing module performs acquisition timing consistency verification and contour coordinate continuity correction on the geometric contour optical data to form corrected contour data. It then performs adjacent contour difference processing and change trend aggregation processing on the corrected contour data according to the acquisition timing to form contour change data characterizing the contour change trend. The module also performs local change intensity marking processing on the contour change data, marking candidate locations with significant contour changes. Finally, it performs acquisition timing consistency verification and edge region extraction processing on the surface imaging optical data to form edge candidate data. The module then performs edge continuity verification and adjacent edge difference processing on the edge candidate data to form edge difference data. It performs anomaly clustering calculation processing on the edge difference data to form edge change data characterizing the continuous state of the surface edges. Finally, it performs anomaly clustering marking processing on the edge change data, marking candidate regions with abnormally concentrated edges.

[0008] As a further description of the above technical solution: The feature alignment position information generation module reads candidate positions with significant contour changes from the contour change data, extracts a set of candidate contour feature points based on the spatial coordinates of the candidate positions in the geometric contour optical data, and retains the point sequence index information corresponding to the acquisition time sequence information in the set of candidate contour feature points. It also reads candidate regions with concentrated edge anomalies from the edge change data, extracts a set of candidate edge feature points based on the pixel coordinates of the candidate regions in the surface imaging optical data, and converts the set of candidate edge feature points into a spatial coordinate expression consistent with the geometric contour optical data, forming a set of imaging candidate spatial points. Coordinate scale consistency correction is performed on the set of candidate contour feature points and the set of imaging candidate spatial points to form a set of contour points to be aligned and a set of imaging points to be aligned. Initial pose alignment is then performed based on the centroid positions of the two types of point sets to form initial alignment pose data. Under the constraints of the initial alignment pose data, iterative nearest-neighbor alignment is performed on the set of contour points to be aligned and the set of imaging points to be aligned. During the iteration process, point pair matching data is generated based on the nearest neighbor matching rule, pose update data is calculated based on the point pair matching data, and the pose update data is updated based on the pose update data. Alignment pose data is generated, and matching residual data is calculated after each iteration. During the iterative nearest-point alignment process, an iterative update adjustment coefficient is introduced. This coefficient is generated based on the change relationship between the matching residual data of two adjacent iterations. The update amplitude of the pose update data is adjusted using this coefficient, resulting in adjusted pose update data. A noise suppression weight coefficient is also introduced during the iterative nearest-point alignment process. This coefficient is generated based on the local dispersion of the point-pair matching data. Weighted residual calculation is performed on the point-pair matching data using this coefficient, resulting in weighted matching residual data. The matching residual data is then updated based on this weighted matching residual data. An iteration termination flag is generated based on the change relationship between the alignment pose data of two adjacent iterations. The iterative nearest-point alignment process terminates when the alignment pose data indicates that the alignment pose data has entered a stable state. Based on the alignment pose data corresponding to the terminated iteration, spatial position alignment mapping is performed between the candidate positions in the contour change data and the candidate regions in the edge change data, generating feature alignment position information data for associating geometric contour changes and surface edge changes.

[0009] As a further description of the above technical solution: The local enhancement data generation module reads feature alignment position information data. Based on the spatial alignment mapping relationship recorded in the feature alignment position information data, it determines the local processing regions corresponding to candidate positions with significant contour changes in the geometric contour optical data and forms contour local processing region index data. Based on the contour local processing region index data, it extracts the original contour sub-data of the corresponding region from the geometric contour optical data, and performs contour continuity preservation processing on the original contour sub-data according to the acquisition time sequence to form contour continuous sub-data. Local morphological compensation processing is performed on the contour continuous sub-data to weaken abnormal fluctuations caused by sparse sampling, local discretization, or noise interference, forming local contour enhancement data to characterize real geometric changes. Based on the spatial alignment mapping relationship recorded in the feature alignment position information data... The system identifies local processing regions corresponding to candidate regions with concentrated edge anomalies in the surface imaging optical data and forms local processing region index data. Based on the local processing region index data, the original imaging sub-data of the corresponding regions is extracted from the surface imaging optical data, and brightness distribution consistency correction processing is performed on the original imaging sub-data to form brightness-corrected imaging sub-data. Local contrast preservation processing is performed on the brightness-corrected imaging sub-data to enhance grayscale differences in edge variation regions and suppress background interference in non-edge regions, forming local imaging enhancement data for characterizing surface detail changes. The local contour enhancement data and the local imaging enhancement data are associated and organized according to the spatial correspondence in the feature alignment position information data to form a data alignment enhancement data set for subsequent consistency evaluation.

[0010] As a further description of the above technical solution: The consistency assessment data generation module reads local contour enhancement data and local imaging enhancement data, and performs spatial pairing processing on the two types of enhancement data according to the spatial correspondence recorded in the feature alignment position information data, forming one-to-one corresponding enhancement feature data pairs. Based on the enhancement feature data pairs, contour change feature sequences representing geometric change trends in the local contour enhancement data and imaging change feature sequences representing surface edge change states in the local imaging enhancement data are extracted respectively, and acquisition timing consistency correction processing is performed on the two types of feature sequences. Under the constraint of enhancement feature data pairs, dynamic window adjustment parameters are introduced to determine the local analysis window used to perform cross-correlation analysis, and dynamic window adjustment parameters for corresponding enhancement feature data pairs are generated according to the density of changes in the spatial distribution of local contour enhancement data and local imaging enhancement data. Based on the dynamic window adjustment parameters, adaptive adjustment processing is performed on the range of the local analysis window involved in the calculation during the cross-correlation analysis. Different analysis window scales are used for regions with concentrated geometric changes and regions with concentrated edge changes. During cross-correlation analysis, local weighting parameters are introduced. Based on the significance of changes at different spatial locations in the enhanced feature data, local weighting parameters are generated for corresponding enhanced feature data pairs. These local weighting parameters are then used to weight the feature values ​​participating in the cross-correlation calculation, forming a weighted feature sequence. Based on the weighted feature sequence and the local analysis window constrained by dynamic window adjustment parameters, cross-correlation analysis is performed on the local contour enhancement data and the local imaging enhancement data to obtain local correlation results characterizing the correspondence between geometric changes and surface changes. The local correlation results corresponding to each enhanced feature data are summarized and subjected to consistency constraints to form consistency evaluation data reflecting the overall correlation between geometric changes and surface changes. The consistency evaluation data is organized and output according to the spatial index relationship in the feature alignment position information data, serving as input data for the quality judgment process.

[0011] As a further description of the above technical solution: The quality assessment module reads the consistency assessment data and, based on the spatial index relationship recorded in the feature alignment position information data, reorganizes the consistency assessment data according to the detection area to form a consistency assessment result set corresponding to the detection area. It then performs stability verification processing on the consistency assessment result set, identifying consistency assessment results that maintain a consistent trend of change in adjacent acquisition time sequences and forming a stable consistency assessment subset. Based on the stable consistency assessment subset, and combining the local contour enhancement data and local imaging enhancement data of the corresponding detection area, it performs joint correlation verification processing to confirm the validity of the correspondence between geometric changes and surface changes within the same detection area. It generates a region-level quality assessment identifier for detection areas that pass the joint correlation verification processing and generates an anomaly identification identifier for detection areas that fail. Finally, it summarizes the region-level quality assessment identifiers and anomaly identification identifiers according to the detection area index relationship to form the quality assessment result for the corresponding detection area of ​​the precision mold parts. The quality assessment result is then output as the final assessment data for the online inspection of the processing quality of the precision mold parts.

[0012] As a further description of the above technical solution: The data chain between modules includes the following data flow relationship: the geometric contour optical data, surface imaging optical data, and corresponding acquisition timing information generated by the optical data acquisition module are transmitted in parallel to the contour edge change data processing module; the contour edge change data processing module generates contour change data based on the geometric contour optical data and edge change data based on the surface imaging optical data. The contour change data contains candidate locations with significant contour changes, and the edge change data contains candidate regions with concentrated edge anomalies. The contour change data and edge change data are synchronously output to the feature alignment position information generation module; the feature alignment position information generation module uses the contour change data and edge change data as input to generate feature alignment position information data, and sends the feature alignment position information data to the local enhancement data generation module and a... The consistency evaluation data generation module and the local enhancement data generation module generate local contour enhancement data and local imaging enhancement data respectively from geometric contour optical data and surface imaging optical data under the constraint of feature alignment position information data. The local contour enhancement data and local imaging enhancement data are output as a set of data alignment enhancement data to the consistency evaluation data generation module. The consistency evaluation data generation module generates consistency evaluation data based on the data alignment enhancement data set and feature alignment position information data, and transmits the consistency evaluation data to the quality judgment module. The quality judgment module generates the quality judgment result of the corresponding inspection area of ​​the precision mold parts based on the consistency evaluation data and combined with the spatial index relationship recorded in the feature alignment position information data, realizing a complete data closed-loop transmission from the original optical data to the quality judgment result.

[0013] As a further description of the above technical solution: An online inspection method for the machining quality of precision mold parts includes the following steps: Simultaneously acquire geometric contour optical data and surface imaging optical data of the same inspection area for precision mold parts, and associate unified acquisition time sequence information for the two types of optical data; perform contour continuity analysis on the geometric contour optical data to generate contour change data characterizing the contour change trend, and mark candidate positions with significant contour changes in the contour change data; perform edge stability analysis on the surface imaging optical data to generate edge change data characterizing the continuous state of the surface edges, and mark candidate regions with abnormal edge concentrations in the edge change data; based on the candidate positions in the contour change data and the candidate regions in the edge change data, perform spatial position alignment processing to generate data for associating geometric contour changes with surface edges. The system generates characteristic alignment position information data of edge changes; under the constraint of characteristic alignment position information data, it performs local reconstruction processing on geometric contour optical data to obtain local contour enhancement data of the corresponding feature region; under the constraint of characteristic alignment position information data, it performs local contrast enhancement processing on surface imaging optical data to obtain local imaging enhancement data of the corresponding feature region; based on the consistency relationship between local contour enhancement data and local imaging enhancement data, it generates consistency evaluation data reflecting the degree of correlation between geometric changes and surface changes; based on the consistency evaluation data, it performs joint judgment processing on geometric contour optical data and surface imaging optical data to generate quality judgment results for the corresponding detection area of ​​precision mold parts.

[0014] The present invention has the following beneficial effects: 1. In this invention, geometric contour optical data and surface imaging optical data are acquired simultaneously during the processing, and time-series information is collected and associated with the two types of data in a unified manner. This ensures that different optical detection results are consistent in both time and spatial dimensions, providing a reliable data foundation for subsequent multi-source data collaborative analysis. This avoids the information bias caused by a single detection method. By performing contour continuity analysis and edge stability analysis on the geometric contour optical data and surface imaging optical data respectively, contour change data and edge change data are generated, and candidate positions and candidate regions are marked in them. This allows abnormal areas of geometric and surface changes to be focused in advance, improving the targeting and efficiency of subsequent processing. By introducing a feature alignment position information generation mechanism, a stable spatial correspondence is established between contour change data and edge change data, enabling geometric contour changes and surface imaging changes to be accurately aligned within the same detection area. This effectively overcomes the differences in spatial scale and coordinate system between multi-source optical data.

[0015] 2. In this invention, by generating local contour enhancement data and local imaging enhancement data under the constraint of feature alignment position information, the influence of sampling noise, local discrepancies, and background interference on the detection results is weakened, and the real geometric changes and surface detail changes are enhanced, thereby improving the stability and reliability of local feature expression. By performing consistency evaluation processing based on local enhancement data, dynamic window adjustment parameters and local weighting parameters are introduced to adaptively analyze the degree of change density in different detection areas, which can more accurately reflect the real correlation between geometric changes and surface changes, and avoid evaluation bias caused by fixed window or uniform weight strategy. By performing stability verification and joint correlation verification on consistency evaluation data, and outputting quality judgment results accordingly, the detection conclusion not only depends on the single detection result, but also can comprehensively consider the change trend in the continuous processing process, thereby improving the reliability, stability, and judgment accuracy of online detection of precision mold parts processing quality. Attached Figure Description

[0016] Fig. 1 This is a system architecture diagram of the present invention; Fig. 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Reference Figs. 1-2This invention provides an embodiment of an online inspection system for the processing quality of precision mold parts, comprising: an optical data acquisition module, which, based on a line laser profile measurement device, simultaneously acquires geometric profile optical data and surface imaging optical data of the same inspection area of ​​the precision mold part to be inspected during the processing of the precision mold part, and associates the two types of optical data with unified acquisition timing information; a profile edge change data processing module, which performs profile continuity analysis on the geometric profile optical data to form profile change data characterizing the profile change trend, and marks candidate positions with significant profile changes in the profile change data; performs edge stability analysis on the surface imaging optical data to form edge change data characterizing the continuous state of the surface edges, and marks candidate areas with abnormal edge concentrations in the edge change data; and a feature alignment position information generation module. The system comprises four modules: a block that performs spatial alignment processing on candidate positions in contour change data and candidate regions in edge change data to generate feature alignment position information data for associating geometric contour changes and surface edge changes; a local enhancement data generation module that, under the constraint of feature alignment position information data, performs local reconstruction processing on geometric contour optical data and surface imaging optical data respectively to generate local contour enhancement data and local imaging enhancement data for corresponding feature regions; a consistency evaluation data generation module that, based on the consistency relationship between local contour enhancement data and local imaging enhancement data, generates consistency evaluation data reflecting the degree of correlation between geometric changes and surface changes; and a quality judgment module that, based on the consistency evaluation data, performs joint judgment processing on geometric contour optical data and surface imaging optical data, and outputs the quality judgment results for the corresponding detection areas of the precision mold parts.

[0019] The data chain between modules includes the following data flow relationship: the geometric contour optical data, surface imaging optical data, and corresponding acquisition timing information generated by the optical data acquisition module are transmitted in parallel to the contour edge change data processing module; the contour edge change data processing module generates contour change data based on the geometric contour optical data and edge change data based on the surface imaging optical data. The contour change data contains candidate locations with significant contour changes, and the edge change data contains candidate regions with concentrated edge anomalies. The contour change data and edge change data are synchronously output to the feature alignment position information generation module; the feature alignment position information generation module uses the contour change data and edge change data as input to generate feature alignment position information data, and sends the feature alignment position information data to the local enhancement data generation module and a... The consistency evaluation data generation module and the local enhancement data generation module generate local contour enhancement data and local imaging enhancement data respectively from geometric contour optical data and surface imaging optical data under the constraint of feature alignment position information data. The local contour enhancement data and local imaging enhancement data are output as a set of data alignment enhancement data to the consistency evaluation data generation module. The consistency evaluation data generation module generates consistency evaluation data based on the data alignment enhancement data set and feature alignment position information data, and transmits the consistency evaluation data to the quality judgment module. The quality judgment module generates the quality judgment result of the corresponding inspection area of ​​the precision mold parts based on the consistency evaluation data and combined with the spatial index relationship recorded in the feature alignment position information data, realizing a complete data closed-loop transmission from the original optical data to the quality judgment result.

[0020] In this embodiment, during the machining of precision mold parts, the optical data acquisition module remains continuously operational and synchronized with the machining cycle of the machining equipment. A line laser profile measuring device is positioned on one side of the machining path of the precision mold part, with its laser emitter facing the detection area of ​​the precision mold part to be inspected. A laser beam is projected along the cross-sectional direction of the detection area, causing the laser beam to form continuous profile projection lines on the surface of the precision mold part, thereby obtaining preliminary signals of geometric profile optical data used to characterize the geometric shape changes on the surface of the precision mold part.

[0021] After the laser beam illuminates the detection area, the optical sensor receives the reflected laser signal and simultaneously acquires the contour image data formed by the reflection. This contour image data is stored as the original geometric contour signal data. Due to factors such as vibration, changes in ambient light, or slight equipment offsets in the processing environment, the original geometric contour signal data may contain interference noise. Therefore, noise filtering is performed on the original geometric contour signal data to reduce random interference components caused by environmental factors or equipment errors, resulting in processed geometric contour signal data.

[0022] After obtaining the processed geometric contour signal data, the signal data is converted into two-dimensional spatial coordinate data according to a predetermined spatial mapping relationship, so that each sampling position corresponds to a clear spatial coordinate expression, thereby forming the final geometric contour optical data. This geometric contour optical data is used to represent the geometric feature state of the surface shape of the precision mold parts as the processing process changes, and serves as the basic input data for subsequent contour continuity analysis.

[0023] Meanwhile, the area array industrial vision imaging device is positioned corresponding to the line laser contour measurement device, and its imaging field of view covers the same detection area as the geometric contour optical data. During the machining of precision mold parts, the area array industrial vision imaging device continuously acquires images of the detection area, obtaining preliminary signals of surface imaging optical data. The imaging sensor receives the reflected light from the detection area, forming raw surface imaging signal data, which includes texture distribution information and edge structure information of the surface of the precision mold parts.

[0024] After acquiring the raw surface imaging signal data, distortion correction processing is performed on the raw surface imaging signal data to eliminate image geometric distortion caused by imaging lens distortion or installation deviation, so that the position of each pixel in the image is consistent with the actual spatial position. Subsequently, brightness normalization processing is performed on the distortion-corrected image data to unify the brightness distribution at different acquisition times and under different lighting conditions, so that the surface imaging signal data has a consistent contrast basis in terms of brightness.

[0025] After distortion correction and brightness normalization, the processed image data is converted into a two-dimensional image data format to form the final surface imaging optical data. This surface imaging optical data is used to represent the surface condition and detailed features of precision mold parts, and serves as the basic input data for subsequent edge stability analysis and edge change data generation.

[0026] While geometric contour optical data and surface imaging optical data are being generated, the optical data acquisition module collects time-series information for the two types of optical data in a unified manner, so that the data acquired by different optical methods maintain a corresponding relationship in the time dimension, providing a time-series consistency basis for the generation of subsequent contour change data, edge change data and feature alignment position information data.

[0027] In this embodiment, the contour edge change data processing module performs a timing consistency check on the geometric contour optical data. By comparing the timing information corresponding to adjacent acquisition times in the geometric contour optical data, it confirms that each contour data is arranged continuously according to the acquisition order, and removes data segments with timing abnormalities or missing data. Subsequently, contour coordinate continuity correction processing is performed on the geometric contour optical data that has passed the timing check, and local coordinate jumps caused by sampling jitter or processing vibration are smoothed and corrected to form corrected contour data.

[0028] After obtaining the corrected contour data, adjacent contour difference processing is performed on the corrected contour data according to the acquisition time sequence. The changes in contour coordinates of corresponding spatial positions at consecutive acquisition times are calculated to obtain the difference results reflecting the magnitude of contour changes. Subsequently, change trend aggregation processing is performed on the adjacent contour difference results, and the difference results from multiple consecutive acquisition times are time-series aggregated to form contour change data used to characterize the contour change trend.

[0029] After generating the contour change data, local change intensity marking processing is performed on the contour change data. By analyzing the degree of concentration of changes in the spatial distribution of the contour change data, spatial locations with relatively prominent changes are identified, and these spatial locations are marked in the contour change data to form candidate locations with significant contour changes for subsequent spatial location alignment processing.

[0030] During the generation of edge variation data, the surface imaging optical data undergoes a temporal consistency check to ensure that the image data is arranged continuously in the acquisition order. Subsequently, edge region extraction processing is performed on the temporal-checked surface imaging optical data to extract regions that may contain edge structures from the image data, forming edge candidate data.

[0031] After obtaining the candidate edge data, edge continuity verification is performed on the candidate edge data to analyze the spatial coherence of the edge structure and remove isolated noisy edges. Subsequently, adjacent edge difference processing is performed on the edge structures that pass the continuity verification to calculate the changes in the edge positions at adjacent acquisition times, forming edge difference data.

[0032] After generating edge difference data, anomaly clustering calculation is performed on the edge difference data to analyze the spatial concentration of edge changes, forming edge change data that characterizes the continuous state of surface edges. Subsequently, anomaly clustering labeling processing is performed on the edge change data to mark candidate regions with abnormally concentrated edges, which are used to characterize areas with significant edge changes in surface imaging optical data.

[0033] Through the above processing, contour change data and its corresponding candidate positions, as well as edge change data and its corresponding candidate regions, are generated, providing a spatial constraint basis for the subsequent generation of feature alignment position information data.

[0034] In this embodiment, the feature alignment position information generation module reads candidate positions with significant contour changes from the contour change data. For each candidate position, based on the spatial coordinate information of the candidate position in the geometric contour optical data, it extracts the corresponding contour candidate feature points and organizes multiple contour candidate feature points into a contour candidate feature point set. While forming the contour candidate feature point set, the correspondence between each contour candidate feature point and the acquisition time sequence information is retained in the set, so that each feature point is associated with clear point sequence index information.

[0035] Simultaneously, candidate regions with concentrated edge anomalies are read from the edge variation data. For each candidate region, edge candidate feature points located within the pixel coordinate range of the candidate region in the surface imaging optical data are extracted, and these extracted edge candidate feature points are organized into an edge candidate feature point set. Since the edge candidate feature point set is initially represented in pixel coordinate form, a spatial coordinate transformation process is further performed on the edge candidate feature point set to convert the pixel coordinate expression into a spatial coordinate expression consistent with the geometric contour optical data, thereby forming an imaging candidate spatial point set.

[0036] After obtaining the candidate feature point set for the contour and the candidate spatial point set for imaging, coordinate scale consistency correction is performed on the two types of point sets. By uniformly correcting the differences between the two types of point sets in spatial scale, coordinate origin, and direction, the two types of point sets are placed in the same spatial coordinate system. After completing the coordinate scale consistency correction, the two types of point sets are determined as the contour point set to be aligned and the imaging point set to be aligned, respectively.

[0037] Before performing point set alignment processing, the centroid positions of the two types of point sets are calculated based on the spatial distribution of points in the contour point set to be aligned and the imaging point set to be aligned. The spatial relationship between the centroid positions is used as the initial alignment basis to perform initial pose alignment processing on the two types of point sets, forming initial alignment pose data.

[0038] Under the constraints of the initial alignment pose data, iterative nearest-neighbor alignment is performed on the contour point set to be aligned and the imaging point set to be aligned. In each iteration, based on the nearest neighbor matching rule, the spatially nearest imaging point is matched for each contour point in both point sets, generating point-pair matching data. Subsequently, pose update data is calculated based on the point-pair matching data, and the current alignment pose data is updated using this data. After each iteration, residual calculation is performed on the current point-pair matching result to obtain matching residual data characterizing the current alignment state. The formula for calculating the point-pair matching residual is: ; For the first The position vectors of candidate contour feature points in the spatial coordinate system. In order to be with the first The position vector of the imaging candidate spatial point matched by each contour candidate feature point. The matching index relationship determined by the nearest neighbor matching rule. For the first The matching residuals for the matched points characterize the spatial deviation in the current alignment state. The formula for aggregating the matching residual data is as follows: , For the first In the first iteration The matching residuals for the matching points, This represents the number of point pairs participating in the matching in the current iteration round. For the first The overall matching residual data corresponding to each iteration. Formula for calculating residual change: , This refers to the matching residual data corresponding to the previous iteration. This refers to the matching residual data corresponding to the current iteration. This represents the change in residual between two adjacent iterations, used to reflect the convergence speed of the residual.

[0039] In the iterative nearest point alignment process, an iterative update adjustment coefficient is introduced. Based on the change relationship of the matching residual data in two adjacent iterations, a corresponding iterative update adjustment coefficient is generated. This coefficient is then used to adjust the update amplitude of the pose update data, thereby avoiding over-adjustment during the pose update process and forming adjusted pose update data.

[0040] At the end of the current iteration, the matching residual data corresponding to this iteration is read, and the matching residual data corresponding to the previous iteration is retained to form residual comparison data between two adjacent iterations. The change amount calculation is performed on the matching residual data between two adjacent iterations to obtain residual change amount information reflecting the residual convergence speed. Based on the residual change amount information, the downward trend state of the matching residual during the current iteration is determined, distinguishing between rapid decline, steady decline, or fluctuation. Based on the residual change amount information and the alignment pose change corresponding to the current iteration, an iterative update adjustment coefficient is generated to reflect the current alignment stability. The iterative update adjustment coefficient is applied to the pose update data calculated in this iteration to adjust the update amplitude of the pose update data, obtaining the adjusted pose update data. The adjusted pose update data is used to update the alignment pose data, and the process proceeds to the next iteration. Pose update adjustment formula: , This is the alignment pose data from the previous iteration. This is pose update data calculated based on point-pair matching data. To iteratively update the adjustment coefficient, To adjust and update the current wheel alignment pose data, This refers to the pose combination operator, used to represent pose update operations. The formula for calculating the iterative update adjustment coefficient is as follows: , For the first The change in residuals during each iteration. The matching residual data from the previous iteration, To avoid extremely small positive numbers with a denominator of zero, it is used only as a numerically stable term. For the first The iteration update adjustment coefficients corresponding to each iteration are used to adjust the pose update amplitude.

[0041] Meanwhile, a noise suppression weight coefficient is introduced during the iterative nearest point alignment process. Based on the dispersion of the point pair matching data within the local spatial range, a corresponding noise suppression weight coefficient is generated. This coefficient is then used to perform weighted residual calculation on the point pair matching data, forming weighted matching residual data. The matching residual data is then updated based on this weighted matching residual data to reduce the impact of outliers on the overall alignment result.

[0042] Read point-pair matching data, extract spatial distance information of each matching point pair, and form a point-pair distance information set. Using each matching point pair as the center, locate other spatially adjacent matching point pairs in the point-pair matching data, forming a local neighborhood point pair set corresponding to that matching point pair. Perform discreteness calculation on the point-pair distance information in the local neighborhood point pair set to obtain the local discreteness information corresponding to that matching point pair, ensuring that the local discreteness information reflects the consistency of the distance distribution between the matching point pair and its neighboring matching point pairs. Within the same iteration, perform normalization processing on the local discreteness information of all matching point pairs to form comparable discreteness normalization results, ensuring that the discreteness of different matching point pairs is on a uniform scale. Based on discreteness... The normalization result generates a noise suppression weight coefficient, ensuring that the larger the normalized dispersion result, the smaller the noise suppression weight coefficient, and vice versa. This suppresses highly discrete matching point pairs at the weight level. A range constraint is applied to the noise suppression weight coefficient to keep it non-negative and bounded, preventing instability in the weighted residual calculation process due to abnormal weights. The noise suppression weight coefficient is written into the point pair matching data, ensuring that each matching point pair carries the corresponding noise suppression weight coefficient. Then, this noise suppression weight coefficient is used to perform weighted residual calculation on the point pair matching data, forming weighted matching residual data, which is then used to update the matching residual data. Local dispersion calculation formula: , For the first The matching residuals for the matching points, For the first For the matching point neighborhood of the th The matching residuals for the matching points, For the first The number of local neighborhood point pairs corresponding to the matching point. For the first The local discreteness information of a matching point is used to characterize its consistency with neighboring matching points. The local discreteness normalization formula is as follows: , For the first For the local dispersion information of the matching points, This represents the minimum local dispersion of all matching point pairs in the current iteration round. The maximum local dispersion of all matching point pairs in the current iteration round. This is the normalized local dispersion result. The formula for calculating the noise suppression weighting coefficient is: , For the first Normalized local discreteness of the matching points For the first For the noise suppression weight coefficients corresponding to the matching points, the larger the dispersion, the smaller the weight. The formula for calculating the weighted matching residual is: , For the original matching residual, For noise suppression weighting coefficients, This is the weighted matching residual, used to update the matching residual data. The formula for calculating the alignment pose change is: , For the current wheel alignment pose data, For the previous round of alignment pose data, The pose change between two adjacent rounds is used to generate the iteration termination flag.

[0043] After each pose update, the aligned pose data obtained from two adjacent iterations are compared, and an iteration termination flag is generated based on the changes in the aligned pose data. When the iteration termination flag indicates that the aligned pose data has entered a stable state, the nearest point alignment process is terminated.

[0044] After the iteration terminates, based on the finally obtained aligned pose data, spatial position alignment mapping processing is performed on the candidate positions in the contour change data and the candidate regions in the edge change data to establish a spatial correspondence between the two types of change data. Based on this, feature alignment position information data for associating geometric contour changes and surface edge changes is generated. Iteration termination determination relationship: When the alignment pose change remains stable during continuous iterations, the nearest point alignment process is determined to have entered a stable phase and the iteration is terminated.

[0045] In this embodiment, the local enhancement data generation module performs local enhancement processing on the geometric contour optical data and the surface imaging optical data respectively, under the constraint of the feature alignment position information data, to form a data basis for subsequent consistency evaluation.

[0046] In the process of generating local contour enhancement data, the spatial alignment mapping relationship recorded in the feature alignment position information data is first read. Based on this spatial alignment mapping relationship, the spatial region corresponding to the candidate position with significant contour changes is located in the geometric contour optical data, and this spatial region is determined as the local processing region. At the same time, contour local processing region index data is formed to identify the location range of this region.

[0047] After determining the local processing area of ​​the contour, the original contour sub-data located within that area is extracted from the geometric contour optical data based on the contour local processing area index data. Since the original contour sub-data may have local discontinuities or slight fluctuations during the acquisition process, contour continuity preservation processing is performed on the original contour sub-data according to the acquisition time sequence to maintain the continuity of contour changes at adjacent acquisition times, forming continuous contour sub-data.

[0048] After obtaining the continuous contour sub-data, local morphological compensation processing is performed on it. This processing reduces abnormal fluctuations caused by sparse sampling, local discretization, or noise interference, making the contour changes smoother and more stable, thus forming local contour enhancement data to characterize the true geometric changes.

[0049] In the process of generating local imaging enhancement data, the spatial alignment mapping relationship recorded in the feature alignment position information data is also used to locate the spatial range corresponding to the candidate region with concentrated edge anomalies in the surface imaging optical data, and this range is determined as the imaging local processing region, while forming the imaging local processing region index data.

[0050] After determining the local processing area of ​​the imaging, the original imaging sub-data within the corresponding area is extracted from the surface imaging optical data based on the local processing area index data. To address potential brightness imbalances in the original imaging sub-data, brightness distribution consistency correction processing is performed to stabilize the brightness distribution within the area, resulting in brightness-corrected imaging sub-data.

[0051] After obtaining the brightness-corrected imaging sub-data, local contrast-preserving processing is performed on it. This processing enhances grayscale differences in edge variation areas while suppressing background interference in non-edge areas, thereby forming locally enhanced imaging data that highlights surface detail variations.

[0052] After generating local contour enhancement data and local imaging enhancement data, the local contour enhancement data and local imaging enhancement data are associated and organized according to the spatial correspondence in the feature alignment position information data to form a data alignment enhancement data set for subsequent consistency evaluation processing.

[0053] In this embodiment, the consistency evaluation data generation module reads local contour enhancement data and local imaging enhancement data, and simultaneously reads the spatial correspondence recorded in the feature alignment position information data. Based on this spatial correspondence, spatial pairing processing is performed on the local contour enhancement data and local imaging enhancement data, so that the two types of enhancement data located at the same feature alignment position form a one-to-one corresponding enhanced feature data pair. The formula for representing the enhanced feature data pair is: , For the first Local contour enhancement data corresponding to each feature alignment position For the first Local imaging enhancement data corresponding to each feature alignment position To increase the number of feature data pairs, the feature alignment position information data is used.

[0054] After forming enhanced feature data pairs, contour change feature sequences characterizing geometric change trends are extracted from the local contour enhancement data, and imaging change feature sequences characterizing surface edge change states are extracted from the local imaging enhancement data. Subsequently, temporal consistency correction processing is performed on both types of feature sequences to ensure that the contour change feature sequences and imaging change feature sequences are consistent in time order. Contour change feature sequences: Imaging change feature sequence: , For the first The enhanced feature data pair in the first... Contour change feature values ​​under a specific acquisition time series For the first The enhanced feature data pair in the first... Imaging change feature values ​​under each acquisition time sequence The acquisition time sequence length is used to ensure consistency between the two sequences after acquisition time sequence consistency correction.

[0055] Before performing consistency analysis, a dynamic window adjustment parameter is introduced under the constraint of the enhanced feature data pairs. Based on the density of spatial distribution changes between the local contour enhancement data and the local imaging enhancement data, a dynamic window adjustment parameter corresponding to each enhanced feature data is generated, and this parameter is used to determine the local analysis window range for performing cross-correlation analysis. Local contour change density: The density of local imaging changes: , For the spatial index length, For the first Each enhanced feature data represents the density of corresponding contour changes. For the first Each enhanced feature data point represents the density of corresponding imaging changes.

[0056] The process involves reading pairs of enhanced feature data, performing spatial correspondence verification on the corresponding local contour enhancement data and local imaging enhancement data within each pair, ensuring consistent spatial indices for both types of data at the same feature alignment location. Under spatial index constraints, the variation magnitude between adjacent indices is calculated for the local contour enhancement data in spatial index order, and the variation magnitudes are aggregated along the spatial index dimension to obtain a representation of the variation density of the local contour enhancement data. Similarly, under the same spatial index constraints, the variation magnitude between adjacent indices is calculated for the local imaging enhancement data in spatial index order, and the variation magnitudes are aggregated along the spatial index dimension to obtain a representation of the variation density of the local imaging enhancement data. Finally, the representations of variation density of the local contour enhancement data and the representations of variation density of the local imaging enhancement data are subjected to consistent scaling. To make the two comparable at the same scale, a fusion calculation is performed based on the two types of change density representations after consistent scaling, resulting in a comprehensive change density representation corresponding to the enhanced feature data. This comprehensive change density representation is then normalized within the current set of enhanced feature data pairs, ensuring that the comprehensive change density corresponding to different enhanced feature data is at a uniform scale. Based on the normalized comprehensive change density, a monotonic mapping process is performed to generate dynamic window adjustment parameters to constrain the adjustment of the local analysis window range. This ensures that the window adjustment direction corresponding to the dynamic window adjustment parameters remains consistent with the analysis requirements of the change-dense region when the comprehensive change density increases. The dynamic window adjustment parameters are then written into the cross-correlation analysis input configuration corresponding to the enhanced feature data, enabling subsequent cross-correlation analyses to determine and adjust the local analysis window range accordingly. Change density after consistent scaling: , , These represent the minimum and maximum values ​​of the density of contour changes in the current enhanced feature data set. These represent the minimum and maximum values ​​of the imaging change density in the current augmented feature data set. Overall change density: , For the first Each enhanced feature data pair represents the corresponding comprehensive change density.

[0057] After obtaining the local analysis window, the local analysis window range involved in the cross-correlation analysis is adaptively adjusted according to the dynamic window adjustment parameters. This ensures that regions with concentrated geometric changes and regions with concentrated edge changes use different window scales during the analysis, thereby improving the relevance of the local consistency assessment. Dynamic window adjustment parameters: , To comprehensively measure the minimum and maximum values ​​of the intensity of change in the current set of data pairs, This is a monotonic mapping relationship used to map the intensity of overall changes to the direction of window adjustment. For the first The dynamic window adjustment parameters corresponding to each enhanced feature data.

[0058] In the cross-correlation analysis, local weighting parameters are introduced. Based on the significance of changes at different spatial locations within the enhanced feature data pair, local weighting parameters are generated for the corresponding enhanced feature data pair. These local weighting parameters are then used to weight the contour change feature sequences and imaging change feature sequences participating in the cross-correlation calculation, forming a weighted feature sequence. Significance of spatial index location changes: , , For the first A pair of enhanced feature data in spatial index The degree of significance of geometric changes at the location For the first A pair of enhanced feature data in spatial index The degree of significance of surface changes at a given location. The degree of significance of combined changes: ,in, , , For normalization processing performed within the spatial index range, For spatial indexing The significance of joint changes at a given location is characterized. Local weighting parameters: , The minimum and maximum values ​​of the joint significance of change within the spatial index range. It is a monotonic mapping relationship. For the first A pair of enhanced feature data in spatial index Local weighting parameters at a given location. Weighted feature sequence: , , , These are the weighted eigenvalues ​​used in the cross-correlation analysis. Local cross-correlation results: , This is a local analysis window obtained by adjusting the parameters of the dynamic window. For cross-correlation displacement, For the first The results of local correlation of a pair of enhanced feature data.

[0059] Read the enhanced feature data pairs and, under the spatial index constraint of the feature alignment position information data, establish a per-spatial index correspondence between the local contour enhancement data and the local imaging enhancement data. For each spatial index position of the local contour enhancement data, calculate the magnitude of change relative to adjacent spatial index positions to form a geometric change significance representation corresponding to that spatial index position. For each spatial index position of the local imaging enhancement data, calculate the magnitude of change relative to adjacent spatial index positions to form a surface change significance representation corresponding to that spatial index position. Perform consistent scaling on the geometric change significance representation and the surface change significance representation at the same spatial index position, allowing them to be combined at the same scale. Perform joint scaling on the consistent scaling results at the same spatial index position. The significance calculation process yields a representation of the joint change significance corresponding to the spatial index position. Within the spatial index range of the current enhanced feature data pair, the joint change significance representation is normalized to ensure that the joint change significance of each spatial index position is on a uniform scale. Based on the normalized joint change significance, a monotonic mapping process is performed to generate local weighting parameters corresponding to each spatial index position, giving higher weights to spatial index positions with higher joint change significance and lower weights to spatial index positions with lower joint change significance. The local weighting parameters are applied to the corresponding spatial index positions of the contour change feature sequence and the imaging change feature sequence, and the feature values ​​participating in the cross-correlation calculation are weighted to form a weighted feature sequence for subsequent cross-correlation analysis calculations.

[0060] After completing the above parameter constraints and feature weighting, based on the weighted feature sequence and the local analysis window after the parameter constraints are adjusted by the dynamic window, cross-correlation analysis is performed on the local contour enhancement data and the local imaging enhancement data to obtain local correlation results used to characterize the correspondence between geometric changes and surface changes.

[0061] After obtaining the local correlation results corresponding to each enhanced feature data, the local correlation results are summarized and consistency constraint processing is performed. Multiple local correlation results in the same detection area are comprehensively analyzed to form consistency evaluation data that reflects the overall correlation between geometric changes and surface changes.

[0062] After the consistency assessment data is generated, it is organized and output according to the spatial index relationship in the feature alignment location information data. This ensures that the consistency assessment data corresponds to the detection area, and the consistency assessment data is used as input data for the quality judgment process. Consistency assessment data: , For the first The optimal local correlation result for each enhanced feature data pair Consistency assessment data to reflect the overall correlation between geometric changes and surface changes.

[0063] In this embodiment, the quality assessment module reads the consistency evaluation data and simultaneously reads the spatial index relationship recorded in the feature alignment position information data. Based on this spatial index relationship, the consistency evaluation data is reorganized according to the detection area, so that each detection area corresponds to a set of consistency evaluation results, forming a set of consistency evaluation results corresponding to the detection area.

[0064] After forming the consistency evaluation result set, stability verification is performed on the consistency evaluation result set. By comparing and analyzing the consistency evaluation results under adjacent acquisition time sequences, consistency evaluation results that maintain a consistent trend of change across multiple consecutive acquisition times are identified, and consistency evaluation results that conform to this trend are selected to form a stable consistency evaluation subset.

[0065] After obtaining the stable consistency evaluation subset, a joint correlation verification process is performed on the stable consistency evaluation subset by combining the local contour enhancement data and local imaging enhancement data of the corresponding detection area. In the joint correlation verification process, the spatial correspondence and consistency of geometric contour changes and surface imaging changes within the same detection area are comprehensively confirmed to determine whether the change correlation within the detection area is valid.

[0066] After completing the joint correlation verification process, based on the joint correlation verification results, a region-level quality judgment label is generated for the detection areas that pass the joint correlation verification process; at the same time, an anomaly judgment label is generated for the detection areas that fail the joint correlation verification process, which is used to distinguish the detection areas with different processing quality states.

[0067] After generating regional quality judgment labels and anomaly judgment labels, the judgment labels corresponding to each detection area are summarized and organized according to the spatial index relationship of the detection areas to form the quality judgment results of the detection areas corresponding to the precision parts of the mold.

[0068] After the quality judgment result is formed, the quality judgment result is output as the final judgment data for the online inspection of the machining quality of precision mold parts, which is used to characterize the machining quality status of precision mold parts in the corresponding inspection area.

[0069] Example 1: This example is applied to the online quality inspection of the mold cavity sidewall during the finishing stage. The inspection object is the cavity sidewall area formed during the continuous cutting process of the mold precision parts.

[0070] During the machining process, the optical data acquisition module maintains synchronization with the CNC machining equipment. It simultaneously acquires geometric contour optical data and surface imaging optical data for the detection area corresponding to the cavity sidewall, and associates the acquisition timing information for both types of optical data. The geometric contour optical data characterizes the contour morphology changes of the cavity sidewall along the height direction during machining, while the surface imaging optical data characterizes the changes in the surface texture and edge structure of the cavity sidewall.

[0071] The contour edge change data processing module performs contour continuity analysis on the geometric contour optical data to form contour change data, and marks candidate locations where contour changes are concentrated due to machining vibration or tool wear in the contour change data; at the same time, it performs edge stability analysis on the surface imaging optical data to form edge change data, and marks candidate regions where surface edges are abnormally concentrated in the edge change data.

[0072] The feature alignment position information generation module performs spatial alignment processing based on candidate positions in the contour change data and candidate regions in the edge change data to generate feature alignment position information data for associating the geometric contour changes of the cavity sidewall with the surface edge changes. This feature alignment position information data is used to determine the one-to-one spatial correspondence between geometric contour changes and surface changes.

[0073] Under the constraint of feature alignment position information data, the local enhancement data generation module performs local enhancement processing on the geometric contour optical data and surface imaging optical data respectively, generating local contour enhancement data and local imaging enhancement data of the corresponding cavity sidewall detection area, and organizes the two types of enhancement data into a data alignment enhancement data set according to the spatial correspondence.

[0074] The consistency assessment data generation module, based on the consistency relationship between local contour enhancement data and local imaging enhancement data, and combined with dynamic window adjustment parameters and local weighting parameters, performs cross-correlation analysis on geometric changes and surface changes to form consistency assessment data that reflects the degree of correlation between geometric changes and surface changes in the cavity sidewall region.

[0075] The quality assessment module performs joint assessment processing on the corresponding detection areas of the cavity sidewall based on the consistency assessment data. When the consistency assessment data remains stable and consistent during continuous acquisition, the quality assessment result for the corresponding detection area is output; when the consistency assessment data shows abnormal fluctuations, the abnormal assessment identification result for the corresponding detection area is output to indicate that there is a deviation in the processing quality of the cavity sidewall.

[0076] Example 2: This example is applied to the online quality inspection scenario after the parting surface of the mold is finished. The inspection object is the geometric flatness and surface edge condition of the parting surface area of ​​the precision parts of the mold.

[0077] During the parting surface finishing process, the optical data acquisition module simultaneously acquires geometric contour optical data and surface imaging optical data of the corresponding detection area of ​​the parting surface, and uniformly associates the acquisition timing information of the two types of optical data. The geometric contour optical data is used to characterize the height variation trend of the parting surface along the detection direction, while the surface imaging optical data is used to characterize the edge sharpness and texture continuity of the parting surface.

[0078] The contour edge change data processing module performs contour continuity analysis on the geometric contour optical data, generates contour change data, and marks candidate positions where the local height change of the parting surface is concentrated in the contour change data; it performs edge stability analysis on the surface imaging optical data, generates edge change data, and marks candidate regions where the edge of the parting surface is abnormally concentrated in the edge change data.

[0079] The feature alignment position information generation module performs spatial position alignment processing based on contour change data and edge change data to generate feature alignment position information data for associating parting surface geometric changes and surface edge changes, so that the parting surface geometric changes and surface changes form a clear correspondence in spatial position.

[0080] Under the constraint of feature alignment position information data, the local enhancement data generation module generates local contour enhancement data and local imaging enhancement data of the detection area corresponding to the fractal surface, and forms a data alignment enhancement data set for consistency evaluation based on spatial correspondence.

[0081] The consistency assessment data generation module performs consistency assessment processing on the parting surface detection area based on the consistency relationship between local contour enhancement data and local imaging enhancement data, forming consistency assessment data that reflects the overall correlation between geometric changes and surface changes of the parting surface.

[0082] The quality assessment module performs stability and joint correlation verification on the consistency assessment data, and generates quality assessment results for the corresponding detection areas of the parting surface based on the verification results. When geometric changes and surface changes are consistent within the parting surface area, a qualified quality assessment result is output; when geometric changes and surface changes show inconsistent correlation, an abnormal assessment result is output to indicate the machining quality deviation of the parting surface.

[0083] An online inspection method for the machining quality of precision mold parts includes the following steps: synchronously acquiring geometric contour optical data and surface imaging optical data of the same inspection area of ​​the precision mold parts, and associating the two types of optical data with unified acquisition time sequence information; performing contour continuity analysis on the geometric contour optical data to form contour change data characterizing the contour change trend, and marking candidate positions with significant contour changes in the contour change data; performing edge stability analysis on the surface imaging optical data to form edge change data characterizing the continuous state of the surface edges, and marking candidate regions with abnormal edge concentrations in the edge change data; and performing spatial alignment processing based on the candidate positions in the contour change data and the candidate regions in the edge change data to generate a method for... The process involves: 1) Using feature alignment information data relating geometric contour changes to surface edge changes; 2) Under the constraint of this feature alignment information data, performing local reconstruction processing on the geometric contour optical data to obtain local contour enhancement data for the corresponding feature region; 3) Under the constraint of this feature alignment information data, performing local contrast enhancement processing on the surface imaging optical data to obtain local imaging enhancement data for the corresponding feature region; 4) Based on the consistency relationship between the local contour enhancement data and the local imaging enhancement data, generating consistency evaluation data reflecting the degree of correlation between geometric changes and surface changes; and 5) Performing joint judgment processing on the geometric contour optical data and the surface imaging optical data based on the consistency evaluation data to generate quality judgment results for the corresponding inspection area of ​​the precision mold parts.

[0084] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An online inspection system for the machining quality of precision mold parts, characterized in that: include: The optical data acquisition module, based on the line laser profile measurement device, simultaneously acquires the geometric profile optical data and surface imaging optical data of the same detection area of ​​the precision mold parts to be inspected during the processing of precision mold parts, and associates the two types of optical data with unified acquisition timing information; The contour edge change data processing module performs contour continuity analysis on geometric contour optical data to form contour change data that characterizes the trend of contour change, and marks candidate positions with significant contour changes in the contour change data. It also performs edge stability analysis on surface imaging optical data to form edge change data that characterizes the continuous state of surface edges, and marks candidate regions with abnormally concentrated edges in the edge change data. The feature alignment position information generation module performs spatial position alignment processing on the candidate positions in the contour change data and the candidate regions in the edge change data to generate feature alignment position information data for associating geometric contour changes and surface edge changes. The local enhancement data generation module performs local reconstruction processing on the geometric contour optical data and surface imaging optical data respectively under the constraint of feature alignment position information data, to generate local contour enhancement data and local imaging enhancement data for the corresponding feature regions. The consistency assessment data generation module generates consistency assessment data that reflects the degree of correlation between geometric changes and surface changes, based on the consistency relationship between local contour enhancement data and local imaging enhancement data. The quality assessment module performs joint assessment processing on geometric contour optical data and surface imaging optical data based on consistency evaluation data, and outputs the quality assessment results for the corresponding inspection areas of the precision mold parts.

2. The online inspection system for the machining quality of precision mold parts according to claim 1, characterized in that: In the process of machining precision mold parts, the optical data acquisition module uses a line laser profile measuring device to project a laser beam onto the detection area of ​​the precision mold part to be inspected, forming a preliminary signal of geometric profile optical data; The optical sensor receives the reflected laser beam and acquires the reflected image to form the original geometric contour signal data. The original geometric contour signal data is subjected to noise filtering to remove noise caused by environmental factors or equipment errors, resulting in processed geometric contour signal data. The processed geometric contour signal data is converted into two-dimensional or three-dimensional spatial coordinate data to form the final geometric contour optical data, which represents the changes in the surface shape and geometric features of the precision mold parts. A line laser profilometry device is used to acquire images of the area to be inspected, thereby obtaining preliminary signals of surface imaging optical data. Image data of the area to be detected is acquired by an imaging sensor, and raw surface imaging signal data is generated, in which the image data reflects the surface texture and edge features of the precision mold parts. The original surface imaging signal data is subjected to distortion correction processing to eliminate image distortion caused by lens distortion or other factors. By performing brightness normalization processing, the brightness distribution of the image is adjusted to adapt to the imaging needs under different ambient lighting conditions. The image data, after distortion correction and brightness normalization, is converted into two-dimensional image data to form the final surface imaging optical data, representing the surface condition and detailed features of the precision mold parts.

3. The online inspection system for the machining quality of precision mold parts according to claim 1, characterized in that: The contour edge change data processing module performs acquisition timing consistency verification and contour coordinate continuity correction on the geometric contour optical data to form corrected contour data; The corrected contour data is processed by performing adjacent contour difference processing and change trend aggregation processing according to the acquisition time sequence to form contour change data that characterizes the contour change trend; Perform local change intensity marking processing on the contour change data to mark candidate locations with significant contour changes in the contour change data; Perform acquisition timing consistency verification and edge region extraction processing on the surface imaging optical data to form edge candidate data; Edge continuity verification and adjacent edge difference processing are performed on the edge candidate data to form edge difference data. Anomaly clustering degree calculation is performed on the edge difference data to form edge change data characterizing the continuous state of the surface edge. Perform anomaly clustering labeling on edge change data to mark candidate regions with edge anomaly clustering in the edge change data.

4. The online inspection system for the machining quality of precision mold parts according to claim 1, characterized in that: The feature alignment position information generation module reads candidate positions with significant contour changes in the contour change data, extracts a set of contour candidate feature points based on the spatial coordinates of the candidate positions in the geometric contour optical data, and retains the point sequence index information corresponding to the acquisition time sequence information in the set of contour candidate feature points. Read the candidate regions with concentrated edge anomalies in the edge change data, extract the set of edge candidate feature points based on the pixel coordinates of the candidate regions in the surface imaging optical data, and convert the set of edge candidate feature points into a spatial coordinate expression consistent with the geometric contour optical data to form an imaging candidate spatial point set. Coordinate scale consistency correction is performed on the contour candidate feature point set and the imaging candidate spatial point set to form the contour point set to be aligned and the imaging point set to be aligned. Initial pose alignment is then performed based on the centroid positions of the two types of point sets to form initial alignment pose data. Under the constraints of the initial alignment pose data, iterative nearest point alignment processing is performed on the contour point set to be aligned and the imaging point set to be aligned. During the iteration, point pair matching data is generated based on the nearest neighbor matching rule, pose update data is calculated based on the point pair matching data, alignment pose data is updated based on the pose update data, and matching residual data is calculated after each iteration. In the process of iterative nearest point alignment, an iterative update adjustment coefficient is introduced. The iterative update adjustment coefficient is generated based on the relationship between the changes in the matching residual data of two adjacent iterations. The update amplitude of the pose update data is adjusted using the iterative update adjustment coefficient to form the adjusted pose update data. In the iterative nearest point alignment process, a noise suppression weight coefficient is introduced. The noise suppression weight coefficient is generated based on the local dispersion of the point pair matching data. The point pair matching data is then processed by weighted residual calculation using the noise suppression weight coefficient to form weighted matching residual data. The matching residual data is then updated based on the weighted matching residual data. An iteration termination flag is generated based on the relationship between the alignment pose data changes between two adjacent iterations. The iteration nearest point alignment process is terminated when the iteration termination flag indicates that the alignment pose data has entered a stable state. Based on the alignment pose data corresponding to the termination iteration, the candidate positions in the contour change data and the candidate regions in the edge change data are subjected to spatial position alignment mapping processing to generate feature alignment position information data for associating geometric contour changes and surface edge changes.

5. The online inspection system for the machining quality of precision mold parts according to claim 1, characterized in that: The local enhancement data generation module reads the feature alignment position information data, determines the local processing region corresponding to the candidate position with significant contour changes in the geometric contour optical data based on the spatial alignment mapping relationship recorded in the feature alignment position information data, and forms the contour local processing region index data. Based on the local contour processing region index data, the original contour sub-data of the corresponding region is extracted from the geometric contour optical data, and contour continuity preservation processing is performed on the original contour sub-data according to the acquisition time sequence to form contour continuous sub-data. Local morphological compensation processing is performed on the continuous contour sub-data to weaken abnormal fluctuations caused by sparse sampling, local discretization, or noise interference, forming local contour enhancement data to characterize the real geometric changes. Based on the spatial alignment mapping relationship recorded in the feature alignment position information data, the local processing region corresponding to the candidate region with concentrated edge anomalies in the surface imaging optical data is determined, and the imaging local processing region index data is formed. Based on the imaging local processing region index data, the original imaging sub-data of the corresponding region is extracted from the surface imaging optical data, and the brightness distribution consistency correction processing is performed on the original imaging sub-data to form brightness-corrected imaging sub-data. Local contrast-preserving processing is performed on the brightness-corrected imaging sub-data to enhance gray-level differences in edge variation areas and suppress background interference in non-edge areas, forming local imaging enhancement data for characterizing surface detail changes; The local contour enhancement data and local imaging enhancement data are associated and organized according to the spatial correspondence in the feature alignment position information data to form a data alignment enhancement data set for subsequent consistency assessment.

6. The online inspection system for the machining quality of precision mold parts according to claim 1, characterized in that: The consistency assessment data generation module reads local contour enhancement data and local imaging enhancement data, and performs spatial pairing processing on the two types of enhancement data according to the spatial correspondence recorded in the feature alignment position information data, forming a one-to-one pair of enhancement feature data. Based on the enhanced feature data pairs, contour change feature sequences representing geometric change trends in local contour enhancement data and imaging change feature sequences representing surface edge change states in local imaging enhancement data are extracted respectively, and acquisition timing consistency correction processing is performed on the two types of feature sequences. Under the constraint of enhanced feature data pairs, a dynamic window adjustment parameter is introduced to determine the local analysis window used to perform cross-correlation analysis, and the dynamic window adjustment parameter of the corresponding enhanced feature data pair is generated according to the density of changes in the spatial distribution of local contour enhancement data and local imaging enhancement data. Based on the dynamic window adjustment parameters, the local analysis window range involved in the calculation during the cross-correlation analysis is adaptively adjusted so that the regions with concentrated geometric changes and the regions with concentrated edge changes use different analysis window scales. In the cross-correlation analysis process, local weighting parameters are introduced. Based on the significance of changes in different spatial locations in the enhanced feature data pair, local weighting parameters are generated for the corresponding enhanced feature data pair. The local weighting parameters are then used to perform weighting processing on the feature values ​​participating in the cross-correlation calculation to form a weighted feature sequence. Based on the weighted feature sequence and the local analysis window constrained by the dynamic window adjustment parameters, cross-correlation analysis is performed on the local contour enhancement data and the local imaging enhancement data to obtain local correlation results that characterize the correspondence between geometric changes and surface changes. The local correlation results corresponding to each enhanced feature data are summarized and consistency constraint processed to form consistency assessment data that reflects the overall correlation between geometric changes and surface changes; The consistency assessment data is organized and output according to the spatial index relationship in the feature alignment location information data, and used as input data for the quality judgment process.

7. The online inspection system for the machining quality of precision mold parts according to claim 1, characterized in that: The quality assessment module reads the consistency assessment data and, based on the spatial index relationship recorded in the feature alignment position information data, reorganizes the consistency assessment data according to the detection area to form a set of consistency assessment results corresponding to the detection area. Stability verification is performed on the consistency assessment result set to identify consistency assessment results that maintain a consistent trend of change in adjacent acquisition time series, and a stable consistency assessment subset is formed. Based on the stable consistency evaluation subset, combined with the local contour enhancement data and local imaging enhancement data of the corresponding detection area, joint correlation verification processing is performed to confirm the validity of the correspondence between geometric changes and surface changes within the same detection area. A region-level quality judgment label is generated for the detection area that passes the joint correlation verification process, and an anomaly judgment label is generated for the detection area that fails the joint correlation verification process. The regional quality judgment marks and anomaly judgment marks are summarized according to the detection area index relationship to form the quality judgment results of the corresponding detection area for precision mold parts; The output quality judgment results serve as the final judgment data for the online inspection of the machining quality of precision mold parts.

8. The online inspection system for the machining quality of precision mold parts according to claim 1, characterized in that: The data links between modules include the following data flow relationships: The geometric contour optical data, surface imaging optical data, and corresponding acquisition timing information generated by the optical data acquisition module are transmitted in parallel to the contour edge change data processing module. The contour and edge change data processing module generates contour change data based on geometric contour optical data and edge change data based on surface imaging optical data. The contour change data contains candidate locations with significant contour changes, and the edge change data contains candidate regions with concentrated edge anomalies. The contour change data and edge change data are synchronously output to the feature alignment position information generation module. The feature alignment location information generation module takes contour change data and edge change data as input, generates feature alignment location information data, and sends the feature alignment location information data to the local enhancement data generation module and the consistency evaluation data generation module respectively. Under the constraint of feature alignment position information data, the local enhancement data generation module generates local contour enhancement data and local imaging enhancement data from geometric contour optical data and surface imaging optical data, respectively, and outputs the local contour enhancement data and local imaging enhancement data as a set of data alignment enhancement data to the consistency evaluation data generation module. The consistency assessment data generation module generates consistency assessment data based on the data alignment enhancement dataset and feature alignment position information data, and then transmits the consistency assessment data to the quality judgment module. The quality assessment module generates quality assessment results for the corresponding inspection areas of precision mold parts based on consistency assessment data and combined with the spatial index relationship recorded in the feature alignment position information data, thus realizing a complete data closed-loop transfer from the original optical data to the quality assessment results.

9. A method for applying to an online inspection system for the machining quality of precision mold parts as described in any one of claims 1-8, characterized in that: Includes the following steps: Simultaneously collect geometric contour optical data and surface imaging optical data of the same inspection area of ​​precision mold parts, and associate unified acquisition timing information for the two types of optical data; Perform contour continuity analysis on the geometric contour optical data to form contour change data that characterizes the trend of contour change, and mark candidate locations with significant contour changes in the contour change data; Edge stability analysis is performed on surface imaging optical data to generate edge variation data characterizing the continuous state of surface edges, and candidate regions with concentrated edge anomalies are marked in the edge variation data; Based on the candidate locations in the contour change data and the candidate regions in the edge change data, spatial position alignment processing is performed to generate feature alignment position information data for associating geometric contour changes and surface edge changes. Under the constraint of feature alignment position information data, the geometric contour optical data is locally reconstructed to obtain the local contour enhancement data of the corresponding feature region; Under the constraint of feature alignment position information data, local contrast enhancement processing is performed on the surface imaging optical data to obtain local imaging enhancement data of the corresponding feature region; Based on the consistency relationship between local contour enhancement data and local imaging enhancement data, consistency assessment data reflecting the degree of correlation between geometric changes and surface changes is generated. Based on the consistency assessment data, joint judgment processing is performed on the geometric contour optical data and surface imaging optical data to generate the quality judgment results of the corresponding inspection area of ​​the precision parts of the mold.

Images