Non-contact micropore form and roughness measurement method and system

Abstract

The invention relates to the field of visual measurement of parts, in particular to a non-contact micropore form and roughness measurement method and system. The method comprises the following steps: carrying out initial contour scanning on a precision part, carrying out part detection base point analysis, and extracting an optimal coordinate base point; performing full-coverage scanning planning and panoramic image splicing based on the optimal coordinate base point to obtain a panoramic fusion image; geometric structure point depth identification is carried out on the panoramic fusion image, three-dimensional point cloud modeling is carried out, and a part CAD digital model is constructed; carrying out micropore shape key size calculation based on a part CAD (Computer Aided Design) digital model, carrying out geometric comparison calculation, and extracting an error value of each measurement point; and performing deviation visual mapping and comprehensive evaluation according to the error value of each measurement point, and outputting a comprehensive evaluation report. According to the invention, the precision part detection efficiency and precision are improved, and the production quality control and later problem traceability of the precision part are enhanced.

Description

Technical Field

[0001] This invention relates to the field of visual measurement of parts, and in particular to a non-contact method and system for measuring the morphology and roughness of micropores. Background Technology

[0002] Traditional methods for detecting micropore morphology and surface roughness mainly rely on equipment such as contact profilometers or scanning electron microscopes (SEM). While these methods can obtain relatively accurate measurement results to a certain extent, they have many limitations: contact methods are prone to damaging the surface of small parts and have low measurement efficiency, making them unsuitable for batch testing; while high-end equipment such as SEM is expensive and complex to operate, making it unsuitable for real-time testing on the production site. Furthermore, when dealing with complex part structures, traditional methods often only provide limited two-dimensional cross-sectional information, making it difficult to comprehensively characterize the three-dimensional morphology and roughness features of micropores.

[0003] With the rapid development of high-resolution imaging technology and computer vision, image recognition-based non-destructive testing methods have gradually become an important direction for quality control of precision parts. By combining high-definition cameras, optical microscopes, and motion platforms, global images and local details of micropores can be acquired without contacting the workpiece. Image processing and pattern recognition algorithms can then be used to automate the analysis of aperture, shape deviations, and surface roughness. However, traditional image recognition algorithms often rely on manually designed features, making it difficult to accurately identify and measure high-precision features such as complex micropore edges and fine textures. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a non-contact method and system for measuring the morphology and roughness of micropores, thereby resolving at least one of the aforementioned technical problems.

[0005] To achieve the above objectives, the present invention provides a non-contact method for measuring the morphology and roughness of micropores, comprising the following steps: Step S1: Perform initial contour scanning on the precision part and analyze the part inspection base points to extract the optimal coordinate base points; Step S2: Based on the optimal coordinate base point, perform full-coverage scanning planning and panoramic image stitching to obtain a panoramic fused image; Step S3: Perform depth recognition of geometric structure points on the panoramic fusion image, and perform 3D point cloud modeling to construct a CAD digital model of the part; Step S4: Calculate the key dimensions of the micropore morphology based on the CAD digital model of the part, perform geometric comparison calculations, and extract the error values ​​of each measurement point; Step S5: Perform deviation visualization mapping and comprehensive evaluation based on the error values ​​of each measurement point, and output a comprehensive evaluation report.

[0006] This specification provides a non-contact micropore morphology and roughness measurement system for performing the non-contact micropore morphology and roughness measurement method described above, comprising: The detection base point analysis module is used to perform initial contour scanning of precision parts, analyze the detection base points of the parts, and extract the optimal coordinate base points; The panoramic stitching module is used to perform full-coverage scanning planning and panoramic image stitching based on the optimal coordinate base points to obtain a panoramic fused image. The point cloud modeling module is used to perform geometric structure point depth recognition on panoramic fusion images and to perform 3D point cloud modeling to build CAD digital models of parts. The shape and size calculation module is used to calculate the key dimensions of microhole shape based on the CAD digital model of the part, perform geometric comparison calculation, and extract the error value of each measurement point. The error assessment module is used to perform deviation visualization mapping and comprehensive assessment based on the error values ​​of each measurement point, and output a comprehensive assessment report.

[0007] The beneficial effects of this invention are as follows: Through high-precision initial contour scanning and base point analysis, the most suitable "optimal coordinate base points" for subsequent scanning and positioning can be extracted, ensuring spatial consistency during subsequent image stitching and 3D modeling. For complex or irregularly shaped parts, automatic base point extraction can adapt to different shapes, reducing human intervention and improving the system's versatility and robustness. A unified coordinate system reduces stitching errors between multiple scans, facilitating accurate extraction of micro-hole size and shape. Intelligent scanning path planning ensures full coverage scanning of complex surfaces and micro-hole areas, avoiding blind spots. Multi-view image stitching generates high-quality panoramic fusion images, making detailed structures (such as micro-hole edges) clearly discernible, facilitating subsequent processing. Automatic planning and stitching reduce manual operation time and improve processing efficiency. Deep learning or algorithms identify geometric structure points in images, generating high-density, high-precision 3D point clouds to realistically restore the part's shape and micro-hole structure. Accurate modeling can also be achieved for complex or irregular micro-hole shapes, suitable for various application scenarios such as irregularly shaped parts and complex workpieces. CAD models serve as the benchmark for subsequent measurements and comparisons, improving the accuracy of error analysis and visualization mapping. Based on high-precision CAD models, micron-level measurements of key dimensions such as micropore diameter, depth, and roundness can be achieved. By comparing actual modeling data with the design model, the error value of each measurement point is accurately extracted, supporting the analysis of complex micropore geometric features. Beyond dimensional errors, surface information such as roughness and morphological contours can also be extracted, providing data support for comprehensive quality assessment. Through visualized deviation mapping, the dimensional errors and morphological anomalies in the micropore region are presented, making problem points readily apparent. The comprehensive assessment report provides quantitative data for quality control during the manufacturing process, facilitating the traceability of non-conforming parts and the adjustment of processing parameters. Automatic assessment report generation supports batch inspection and quality trend analysis, improving inspection efficiency and subsequent problem traceability capabilities. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the steps of a non-contact micropore morphology and roughness measurement method according to the present invention.

[0009] Figure 2 This is a flowchart illustrating the detailed implementation steps of step S1.

[0010] Figure 3 This is a flowchart illustrating the detailed implementation steps of step S2.

[0011] Figure 4 This is a flowchart illustrating the detailed implementation steps of step S3. Detailed Implementation

[0012] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0013] This application provides a non-contact method and system for measuring the morphology and roughness of micropores. The execution entity of the non-contact method and system includes, but is not limited to, mechanical equipment, data processing platforms, cloud server nodes, network upload devices, etc., which can be considered as general computing nodes in this application. The data processing platform includes, but is not limited to, at least one of an audio-visual management system, an information management system, and a cloud-based data management system.

[0014] Please see Figures 1 to 4 This invention provides a non-contact method for measuring the morphology and roughness of micropores, comprising the following steps: Step S1: Perform initial contour scanning on the precision part and analyze the part inspection base points to extract the optimal coordinate base points; Step S2: Based on the optimal coordinate base point, perform full-coverage scanning planning and panoramic image stitching to obtain a panoramic fused image; Step S3: Perform depth recognition of geometric structure points on the panoramic fusion image, and perform 3D point cloud modeling to construct a CAD digital model of the part; Step S4: Calculate the key dimensions of the micropore morphology based on the CAD digital model of the part, perform geometric comparison calculations, and extract the error values ​​of each measurement point; Step S5: Perform deviation visualization mapping and comprehensive evaluation based on the error values ​​of each measurement point, and output a comprehensive evaluation report.

[0015] In the embodiments of the present invention, see Figure 1 This is a schematic flowchart illustrating the steps of a non-contact micropore morphology and roughness measurement method according to the present invention. In this example, the steps of the non-contact micropore morphology and roughness measurement method include: Step S1: Perform initial contour scanning on the precision part and analyze the part inspection base points to extract the optimal coordinate base points; In this embodiment, a high-resolution CCD camera mounted on a high-precision motion platform is used to perform a 360-degree omnidirectional initial contour scan of the precision part. The scanning resolution is set to 0.5μm pixel accuracy to ensure accurate capture of minute features. The Canny edge detection algorithm combined with the Harris corner detection algorithm is used to automatically identify key geometric feature points of the part, including outer contour boundary points, internal micro-hole edge points, and geometric center points. The spatial coordinates of the detected feature points are optimized using the least squares fitting method to establish a set of candidate coordinate base points. The positional accuracy of each candidate base point is controlled within ±0.1μm. A multi-objective optimization algorithm is used to comprehensively evaluate the geometric stability, measurement coverage, and repeatability of each candidate base point, with geometric stability weighted at 40%, measurement coverage weighted at 35%, and repeatability weighted at 25%. The measurement uncertainty distribution under different base point selection schemes is simulated using Monte Carlo simulation, and the scheme with the minimum overall uncertainty is selected as the optimal coordinate base point. A transformation matrix between the base point coordinate system and the machine coordinate system is established, with transformation accuracy controlled within ±0.05μm. The accuracy and stability of the coordinate transformation are verified through repeated measurements using a standard sphere.

[0016] Step S2: Based on the optimal coordinate base point, perform full-coverage scanning planning and panoramic image stitching to obtain a panoramic fused image; In this embodiment, a spiral scanning trajectory is planned based on the determined optimal coordinate base point, with the scanning interval set to 20μm to ensure a 30% overlap between adjacent scanning strips to guarantee stitching accuracy. A six-axis linkage control system of the motion platform is used to execute precise scanning actions, achieving a positioning accuracy of ±0.1μm for each axis and a repeatability accuracy better than ±0.05μm. High-resolution image sequences are simultaneously acquired during the scanning process, with the camera exposure time set to 5ms and the gain value adjusted to 64 to ensure a signal-to-noise ratio greater than 40dB. The SIFT feature point matching algorithm is used to perform feature point matching on adjacent images. Feature point extraction and registration are performed with a matching success rate of over 95% and registration error controlled within 0.5 pixels. A multi-band fusion algorithm is used to seamlessly stitch the registered images, employing a combination of Laplacian and Gaussian pyramids to eliminate stitching gaps, achieving sub-pixel accuracy. Histogram equalization and contrast enhancement algorithms are used to optimize the quality of the stitched panoramic image, ensuring overall brightness uniformity and detail clarity. An image quality evaluation system is established, including sharpness, contrast, and noise level assessments, to ensure the fused panoramic image meets the requirements for subsequent analysis.

[0017] Step S3: Perform depth recognition of geometric structure points on the panoramic fusion image, and perform 3D point cloud modeling to construct a CAD digital model of the part; In this embodiment, multi-scale geometric feature analysis is performed on the panoramic fusion image. The difference of Gaussian operator is used to detect feature points at different scales, with a detection threshold set to 0.03 to ensure accurate identification of micropore boundaries. A deep learning convolutional neural network model is used for semantic segmentation of the geometric structures in the image. The network training uses a ResNet-50 backbone architecture, and the training dataset contains 10,000 labeled images, achieving a model accuracy of 98.5%. The surface morphology is reconstructed based on stereo vision principles combined with structured light projection technology. The structured light stripe spacing is set to 50 μm, the projection angle to 30 degrees, and the depth resolution to [missing information]. The depth accuracy was achieved to 0.1 μm. Discrete depth point cloud data was converted into a continuous 3D mesh model using the Poisson surface reconstruction algorithm, with a mesh density controlled at 10,000 triangular faces per square millimeter. The mesh topology was optimized using the Delaunay triangulation algorithm to eliminate duplicate vertices and degenerate triangles, ensuring the geometric integrity of the model. NURBS surface fitting technology was employed for high-precision modeling of complex freeform surfaces, with the fitting error controlled within ±0.05 μm. CAD model quality evaluation indicators were established, including mesh quality, fitting accuracy, and topological integrity, to ensure the reliability and accuracy of the digital model.

[0018] Step S4: Calculate the key dimensions of the micropore morphology based on the CAD digital model of the part, perform geometric comparison calculations, and extract the error values ​​of each measurement point; In this embodiment, key geometric parameters of the microhole are extracted based on the constructed CAD digital model of the part, including hole diameter, roundness, cylindricity, and positional accuracy, with a measurement accuracy of ±0.02μm. The equivalent diameter and center coordinates of the microhole are calculated using a least-squares circle fitting algorithm, with the fitting residual controlled within 0.01μm. Roundness error and diameter deviation are also calculated. The cylindricity error of the microhole is evaluated using a cylindrical surface fitting method, and the cylindricity value is calculated using the minimum region method. The straightness and perpendicularity of the microhole axis are also evaluated. A geometric comparison and analysis system between the theoretical CAD model and the measured model is established, using point-to-surface distance... The deviation value of each measurement point was obtained by calculation method, with a deviation calculation accuracy of ±0.01μm. Statistical analysis methods were used to perform normality tests and outlier identification on the deviation data, eliminating outlier data points other than those defined by the 3σ principle, and establishing a reliable error distribution model. The passability of each dimensional parameter was evaluated based on tolerance zone analysis theory, and different judgment thresholds were set. The first-level accuracy requirement was a deviation of less than ±0.05μm, and the second-level accuracy requirement was a deviation of less than ±0.1μm. An error source tracing system was constructed to analyze the contribution of geometric error, systematic error, and random error, providing a theoretical basis for subsequent error correction.

[0019] Step S5: Perform deviation visualization mapping and comprehensive evaluation based on the error values ​​of each measurement point, and output a comprehensive evaluation report.

[0020] In this embodiment, pseudo-color mapping technology is used to convert three-dimensional deviation data into an intuitive color distribution map. The color mapping range is set to -0.2μm to +0.2μm, and a blue-green-yellow-red gradient color spectrum is used to represent the deviation magnitude. A thermal mapping algorithm is used to construct a spatial density analysis of the deviation distribution, identifying areas of high deviation concentration and deviation distribution patterns. The thermal mapping resolution is set to 10μm × 10μm. An interactive deviation analysis interface is constructed based on three-dimensional visualization technology, supporting 360-degree rotation observation and local magnification functions, with a magnification of up to 100x. A deviation statistical analysis model is established to calculate the average deviation, standard deviation, maximum positive deviation, and maximum... The system analyzes statistical parameters such as negative deviation and assesses the normality and symmetry of the deviation distribution. A comprehensive evaluation algorithm is used to evaluate the measurement results across multiple dimensions, including dimensional accuracy, shape accuracy, surface quality, and overall pass rate, with weights of 30%, 25%, 25%, and 20%, respectively. A standardized measurement report template is constructed, containing complete content such as measurement parameter settings, raw data records, deviation analysis results, statistical evaluation conclusions, and improvement suggestions. An automatic report generation system is established, integrating data processing, chart generation, and document editing functions to generate measurement and testing reports that conform to industry standards, providing a scientific basis for quality control and process improvement.

[0021] In this embodiment, see Figure 2 The diagram below illustrates the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: An initial contour scan of a precision part is performed using a high-precision motion measurement platform to identify the boundary feature points of the precision part. The boundary feature points are located in three-dimensional space, and their three-dimensional coordinates are extracted. The overall shape is evaluated based on three-dimensional positioning coordinates to obtain the main geometric features; Based on the above, topological correlation analysis is performed on the boundary feature points, and the symmetry axis distribution is identified to extract the symmetry axis distribution information; Calculate the geometric center of gravity and principal direction of the part based on the main geometric features and the distribution information of the axis of symmetry; Based on the geometric center and main direction of the part, the part detection base point is analyzed, and the optimal coordinate base point is extracted.

[0022] In this embodiment, a precision part is placed on a high-precision motion measurement platform, which can achieve nanometer to micrometer-level displacement control in the X, Y, and Z directions. The measurement system is equipped with a high-definition imaging unit, such as a high-resolution industrial camera or a confocal microscope lens, and a stable light source to ensure imaging contrast and clarity. The platform acquires contours point by point or line by line according to a set scanning path and step size (e.g., 2 μm or 5 μm), forming a continuous sequence of boundary images. Subsequently, edge recognition and feature extraction algorithms (e.g., Canny, Sobel, or sub-pixel boundary detection methods) are used to process the images and extract the feature point set of the part's outer boundary. To ensure data stability, noise can be suppressed during the scanning process through multiple sampling and signal averaging. After the boundary feature points are identified, these points need to be mapped from the image coordinate system to the real space coordinate system. Through the platform's own high-precision displacement feedback system, the physical coordinates in the X and Y directions of each acquisition position can be directly obtained. At the same time, optical ranging units (e.g., laser confocal displacement sensors, white light interferometry devices, etc.) are used to obtain height information in the Z direction. To achieve consistency between pixel coordinates and physical coordinates, a calibration model needs to be established in the early stages of the system to correct imaging distortion and construct the pixel-space transformation relationship. During the acquisition process, triaxial position data is recorded simultaneously for each boundary point, forming a complete three-dimensional coordinate point cloud set. To improve coordinate accuracy, filtering, fitting, and other methods can be used to optimize discrete points, eliminate outliers, and form a stable three-dimensional coordinate set.

[0023] Surface reconstruction and geometric fitting based on point clouds, such as using least-squares surface fitting, B-spline reconstruction, or the Alpha-Shape algorithm to model the surface of boundary points. Surface fitting removes isolated points and local noise, resulting in a smoother overall shape that conforms to the actual part's shape. Subsequently, based on the modeled point cloud or surface, the main geometric features of the part, such as aperture, thickness, radius of curvature, and flatness, are extracted. These geometric features not only reflect the overall shape of the part but can also be compared with the theoretical design model to determine shape deviations and geometric error distributions. After shape evaluation, a geometric feature library of the part can be created, and a point cloud adjacency graph or a Delaunay triangulation structure can be constructed to transform the discrete point set into an analyzable topological mesh. This mesh allows for local curvature, eigenvector, and boundary connectivity analysis, thereby identifying the regularity of the part's boundaries. If the part has rotational or axisymmetric properties, the axis of symmetry can be calculated by fitting the boundary point set to an ideal geometric model (such as a circle, ellipse, or cylinder), and the spatial distribution of the axis of symmetry can be further extracted. For parts with multiple symmetrical features, a global optimization algorithm can be used to extract the optimal symmetry axis direction, ensuring the stability and uniqueness of the identification. Obtaining the symmetry axis distribution information of the part provides a reference benchmark for calculating the geometric centroid and principal directions.

[0024] The calculation of the geometric centroid is based on the overall volume or surface area distribution of a 3D point cloud. By dividing the point cloud into small units and combining the position and weight of each unit, the overall spatial average position is obtained, which is the geometric centroid. The calculation of the principal direction typically relies on Principal Component Analysis (PCA). By calculating the covariance matrix of the point cloud and extracting eigenvectors, the main extension direction of the part in 3D space is obtained. This direction is often highly correlated with the part's axis of symmetry or main geometric features. To avoid interference from noise points in the calculation results, the point cloud can be smoothed first, and outliers can be removed. Determining the geometric centroid and principal direction is a crucial step in establishing a spatial reference system for the part, ensuring that the part's positioning and alignment no longer depend on the clamping position but are based on the part's inherent geometric features. The geometric centroid is determined as the origin of the coordinate system, ensuring the consistency of the part's reference in space. Secondly, the principal direction is used as the Z-axis of the coordinate system, aligning the part's detection with its geometric features. To further improve the stability of the coordinate system, several base points can be selected at the part's boundary or specific feature points as reference points in the X and Y directions. By minimizing the deviation between the base point and the overall point cloud, the coordinate system is optimized to accurately reflect the actual geometric features of the part. The resulting optimal detection base point not only serves as a unified reference for morphology and roughness calculations but also effectively reduces deviations caused by measurement platform errors or optical distortions, thereby improving the repeatability and reliability of the entire detection system.

[0025] In this embodiment, see Figure 3 The diagram below illustrates the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Regional complexity analysis is performed based on key geometric features to identify the geometric complexity of different parts of precision components; Based on the aforementioned geometric complexity, an adaptive scanning method is defined, resulting in the adaptive scanning method. Full coverage scanning planning is performed based on optimal coordinate base points and adaptive scanning method to generate scanning planning path; Based on the scanning planning path, a full-range measurement and imaging of precision parts is performed, and the full-range measurement image is extracted. A panoramic fused image is obtained by stitching together panoramic images from all-around measurement images.

[0026] In this embodiment, the basic principle of complexity analysis is to quantify the complexity of a part's surface or boundary by using curvature distribution, feature density, and local geometric change rate. For example, the complexity is significantly higher in areas with drastic changes in boundary curvature or where micropore clusters exist than in flat or smooth areas. Specifically, the overall point cloud or reconstructed surface of the part is divided into several region units. The area of ​​each unit can be set according to the scanning accuracy, such as 50 μm × 50 μm or smaller. The geometric curvature, surface normal change rate, and boundary feature point density are calculated for each region unit, and a region complexity index is obtained through a weighted average. Regions with higher complexity values ​​represent more subtle geometric features and faster surface changes, thus requiring higher scanning accuracy. Through complexity analysis, the part can be divided into "high-complexity regions," "medium-complexity regions," and "low-complexity regions," forming a complexity distribution map, which provides a basis for adaptive adjustments to the subsequent scanning method. Adaptive scanning methods are defined for different complexity regions. For high-complexity regions, a high-resolution, small-step scanning strategy is required. For example, the platform's step size can be set between 1 μm and 2 μm to ensure the capture of minute geometric changes. Regarding imaging parameters, the magnification should be increased and the field of view reduced to achieve higher clarity in local details. For moderately complex areas, a moderate step size (e.g., around 5 μm) can be used to improve scanning efficiency while maintaining feature integrity. For low-complexity areas, a larger step size (e.g., 10 μm or more) can be used to quickly cover large areas without affecting overall accuracy. The core of the adaptive scanning method is to define the scanning strategy hierarchically based on complexity information, ensuring high accuracy in complex areas and high efficiency in simple areas, thus achieving a balance between scanning accuracy and efficiency. This method not only avoids the low efficiency problem caused by globally uniform high-precision scanning but also avoids the loss of key features caused by low-precision scanning, achieving comprehensive optimization of accuracy and efficiency.

[0027] Using the geometric centroid as the scanning starting point and the main direction as the Z-axis, the scanning path is aligned with the features of the part. Within this coordinate system, complexity distribution is combined with an adaptive scanning strategy to generate a fully covered scanning path. Specifically, a partitioned scanning plan is used: the entire part surface or boundary is first divided into regions, with dense scanning grids for high-complexity regions and sparse scanning grids for low-complexity regions. Path planning can employ a serpentine scanning method or a block scanning method, optimized in conjunction with the travel range and speed limits of the motion platform. To avoid omissions or overlaps during scanning, a certain overlap rate (e.g., 10% to 15%) needs to be reserved during path generation to ensure continuity between adjacent scanning areas. When executing the planned path, the motion platform drives the imaging system to acquire images point-by-point or region-by-region, forming comprehensive measurement data. The imaging process needs to be performed under constant illumination conditions, using a stable cold light source or collimated light source to reduce the impact of illumination changes on image edge sharpness. In highly complex regions, the platform operates with small step sizes and acquires high-resolution images to obtain details of micropore boundaries and surface roughness. In low-complexity regions, the platform moves rapidly with larger step sizes to form a large field-of-view image, improving imaging efficiency. The imaging system can output grayscale or color images, with resolutions set from sub-micrometer to micrometer levels to meet the accuracy requirements of different regions. Throughout the process, the system automatically records the coordinate position and orientation information of each image, enabling subsequent image stitching and fusion to be precisely registered based on spatial location. Multiple discrete images are stitched together to form a continuous panoramic image, achieving unified observation of the overall morphology of the part. The key to stitching lies in image registration and fusion. Initial geometric registration is performed based on the position information recorded during acquisition, locating each image to its corresponding coordinate position. Subsequently, image feature matching algorithms (such as SIFT, SURF, or ORB features) are used to extract key points in the overlapping areas of adjacent images, and precise stitching parameters are calculated through matching and optimization. To eliminate brightness differences and seam marks at the splicing points, multi-resolution fusion or weighted averaging methods can be used for image fusion, resulting in a natural transition between spliced ​​areas. The final panoramic image covers the entire surface or boundary of the part, presenting both the overall shape and local details within a single image.

[0028] In this embodiment, the adaptive scanning method is specifically defined as follows: Based on the aforementioned geometric complexity, flat regions, curved regions, edge regions, and hole regions are divided. The flat area is divided into a regular grid for grid scanning; Calculate the principal curvature directions of the surface region; Perform a spiral scan based on the main curvature direction; identify the edge lines of the edge and hole regions; Contour tracking scan is performed based on the edge lines.

[0029] In this embodiment, after the complexity analysis is completed, the surface regions of the part need to be classified according to the complexity index. The basic principle of classification is to use local curvature, rate of change of normal vector, and boundary feature distribution to determine different geometric regions. Regions with curvature close to zero and minimal change in normal vector direction can be classified as flat regions; surfaces with continuous curvature changes and large normal vector inflection points are classified as curved regions; if a region has obvious boundary breakpoints or abrupt changes in shape, and the boundary lines are continuously distributed, it is classified as an edge region; and when a closed contour or a large negative curvature distribution is detected, it is identified as a hole region. During the classification process, the point cloud or curved surface is usually divided into fixed-size analysis units (e.g., 20 μm × 20 μm), and the average curvature and boundary feature index are calculated within each unit. Then, a threshold is used to determine which region category it belongs to. Since the geometric features of flat regions are relatively simple, the curvature is almost zero, and the surface changes are small, it is suitable to use a regular mesh division method for scanning. The specific method involves using the boundaries of a flat area as constraints, projecting it onto the XY plane, and dividing it into uniform rectangular grid cells. The grid size can be flexibly set according to the required resolution, such as 10 μm × 10 μm or finer. The scanning path traverses these grids sequentially in a row-determinant order, i.e., raster scanning. This method ensures uniform coverage of the entire planar area and facilitates subsequent image stitching. To avoid missing minute surface features due to excessively large scanning steps, a certain overlap (e.g., 5% to 10%) can be added to the scanning path. During imaging, the motion platform moves along a regular grid trajectory, and the high-definition imaging unit acquires the area image point by point. Through raster scanning, not only can data acquisition of flat areas be completed quickly, but the integrity and accuracy of image coverage can also be ensured, providing high-quality input data for overall panoramic stitching.

[0030] Curved surfaces, due to their significant curvature variations, cannot be directly scanned using planar regular meshes. Instead, the scanning direction must be designed based on their geometric characteristics. In this case, the curvature information of the curved surface region needs to be extracted first. This is done by calculating second-order geometric features, such as principal and secondary curvatures, from the surface point cloud data, and extracting the maximum curvature value and its corresponding direction vector within each region cell. This method determines the dominant curvature direction of each surface cell, i.e., the direction of the fastest surface change. To ensure the stability of direction recognition, the curvature directions are typically smoothed by weighted averaging of the curvature vectors of adjacent cells, avoiding direction jumps caused by noise points. The calculation of the dominant curvature direction not only helps guide the arrangement of the scanning path but also reveals the geometric patterns of the curved surface region. For example, for regions resembling cylinders or spheres, their dominant curvature directions exhibit a clear radial or tangential distribution. Analysis can deduce the global curvature dominance pattern, providing geometric constraints for subsequent helical scanning. After obtaining the dominant curvature direction of the curved surface region, a helical path is selected as the scanning strategy. The advantage of helical scanning lies in its ability to closely conform to the geometry of curved surfaces, expanding coverage along the main curvature direction to ensure scanning accuracy and continuity. This is achieved by using the geometric center of the curved surface region as the initial point and generating a helical trajectory along the main curvature direction. The trajectory spacing can be set according to the required resolution, for example, 2 μm to 5 μm. The helical path automatically expands to the entire curved surface region during coverage, forming a continuous and seamless acquisition line. To avoid excessive overlap or omissions, a curvature adaptive control mechanism is introduced during path generation: the path spacing is reduced at locations with higher curvature and appropriately increased at locations with lower curvature. The platform moves along the helical path, and the imaging unit acquires data point-by-point at a fixed frequency, ensuring data density and uniform distribution across the curved surface region. Helical scanning not only effectively captures detailed morphology on curved surfaces but also reduces stitching errors caused by linear scanning, resulting in more coherent image data of the curved surface region that more closely resembles the real structure.

[0031] Edges and holes are typically the most critical feature areas in precision parts, as their morphology directly affects the assembly accuracy and functional performance of the parts. Therefore, these areas require special processing. First, edge detection algorithms (such as Canny edge detection and sub-pixel boundary extraction) are used to identify the precise boundary lines of edges and holes. These boundary lines exist as continuous curves, and discrete points can be transformed into smooth trajectories through curve fitting (such as B-spline or polynomial curve fitting). Subsequently, a scanning strategy performs contour tracking along these boundary lines; that is, the platform moves according to the edge line trajectory, completing image acquisition point by point. To improve scanning stability, a parallel auxiliary scan line can be added on each side of the edge line to supplement data in the transition area near the edge. For hole areas, contour tracking scanning completely encircles the inner edge of the hole, ensuring that hole diameter, roundness, and edge roughness information are fully captured. In this way, the minute morphological features of edge and hole areas can be recorded with high precision, providing crucial data support for subsequent calculations of micro-hole morphology and roughness errors.

[0032] In this embodiment, reference Figure 4 The diagram below illustrates the detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include: Perform depth recognition of geometric structure points in panoramic fusion images and mark the geometric structure points; Spatial positioning of geometric structural points generates structural point location information; Perform type parsing on the geometric structure points to obtain the structure point type; Discrete point cloud conversion is performed on the panoramic fusion image to generate discrete point cloud data of the parts; The discrete point cloud data of the part is labeled according to the type of structural points, and a 3D point cloud model is constructed based on the location information of the structural points to build a CAD digital model of the part.

[0033] In this embodiment, after obtaining the panoramic fused image, representative geometric structural points need to be identified in the overall image. Geometric structural points refer to key points that reflect the characteristic shape of the part, such as the center of an opening, edge intersections, and extreme points of surface curvature. These points not only serve as the skeleton in subsequent 3D modeling but also provide crucial references for calculating shape errors. This is achieved by using image analysis methods to identify the grayscale gradient, texture changes, and geometric boundary features of the panoramic image. For example, gradient operators can be used to detect edge direction, and structural points can be extracted at edge intersections or abrupt curvature changes. For micro-hole features, circle fitting or ellipse detection methods can be used to obtain the center of the opening as the structural point. To improve recognition accuracy, a deep learning target detection network can be added during the recognition process to classify and filter feature points, avoiding misidentification. The identified geometric structural points are marked on the panoramic image, forming a feature point distribution map with geometric reference significance. After the geometric structural points are marked on the panoramic image, they need to be mapped to 3D space to obtain their actual physical coordinates. Since panoramic images are stitched together from multiple images captured by a scanning platform at different locations, the spatial position of each pixel can be calculated using imaging calibration parameters and platform displacement data. Specifically, this is achieved by establishing a one-to-one correspondence between image coordinates and actual spatial coordinates based on the calibration model of the optical imaging system and the platform's 3D coordinate feedback. For complex curved surfaces, depth information can be obtained using binocular stereo matching or optical phase reconstruction methods, further improving spatial positioning accuracy. Through this process, each geometric point can be assigned a unique 3D coordinate value, forming a set of spatial position information for each point.

[0034] After obtaining the spatial location information of geometric structural points, it is necessary to analyze and classify the types of these points. The goal of type analysis is to determine which type of structural feature the point belongs to, such as an opening point, edge point, planar feature point, curved surface feature point, or corner point. This is achieved through multi-dimensional analysis based on the geometric attributes of the structural points: first, the magnitude of local curvature is used to determine whether the point is in a flat or curved region; second, boundary detection is used to determine whether the point is on an edge; for points with closed contour features, they can be further analyzed as opening points or inner contour points. During type analysis, pattern recognition algorithms can be combined to perform cluster analysis on the neighborhood geometric distribution of the points, thereby improving classification accuracy. For example, edge points are usually located in areas with a large rate of change of normal vector, while planar feature points are distributed in positions with curvature close to zero. Through type analysis, a set of structural point data with clear semantics can be formed, providing data support for subsequent point cloud annotation and 3D modeling. To transition from 2D panoramic images to 3D digital modeling, the image data needs to be converted into discrete point clouds. Specifically, the pixels in the panoramic image are mapped to point cloud data according to their corresponding 3D coordinates, with each pixel corresponding to a 3D point. At this point, the point cloud density is determined by both the image resolution and the scan step size. For example, when the scan step size is set to 2 μm, the generated point cloud density can reach hundreds of thousands of points per square millimeter. During the conversion process, geometric correction of the point cloud is required to eliminate the effects of optical distortion or cumulative errors in the scan path. The Iterative Closest Point (ICP) method can be used for global alignment of the point cloud, ensuring spatial consistency between point clouds in different regions.

[0035] The structural point type information obtained from the previous analysis is fused with the point cloud data to complete point cloud annotation. This is done by using different types of structural points as annotation benchmarks and labeling the adjacent regions in the point cloud with corresponding tags, such as edge region points, hole boundary points, and surface points. This gives the point cloud geometric semantic features, rather than just a meaningless collection of spatial points. After annotation, 3D modeling is performed based on the spatial location of the structural points. Modeling methods can employ surface reconstruction algorithms (such as Poisson reconstruction, B-spline fitting, or Delaunay triangulation) to transform the point cloud into a continuous surface or solid model. For hole regions, a complete micropore structure can be generated by fitting cylindrical or conical surfaces; for planar regions, regular geometric surfaces are generated through planar fitting. The final CAD digital model not only closely matches the real part in geometry but also possesses a complete structural classification at the semantic level, allowing it to be directly used for subsequent calculations of micropore morphology and roughness errors.

[0036] In this embodiment, step S4 includes the following steps: Calculate the key dimensions of micropore morphology based on the CAD digital model of the part, and extract the morphological and dimensional parameters of multiple micropore morphologies. Geometric relationship analysis is performed on the CAD digital model of the parts to obtain the geometric relationships between structures; Based on the geometric relationship between the structures, the morphological and dimensional parameters are measured and the deviation is corrected to obtain the micropore corrected morphological parameters; Based on the preset theoretical micropore morphology parameters, the micropore correction morphology parameters are geometrically compared and calculated, and the error values ​​of each measurement point are extracted.

[0037] In this embodiment, feature segmentation is performed on all micropores identified in the model to distinguish the pore portion from its surrounding substrate region. Then, a geometric measurement algorithm is used to calculate the key morphological dimensions of each micropore, including pore diameter, roundness, ellipticity, pore depth, tilt angle, and pore edge thickness. The calculation method can be based on least squares fitting and geometric constraint extraction: for example, the boundary point set of the micropore is fitted as a circle or ellipse to obtain the average pore diameter and shape deviation; for the depth parameter, it is directly calculated based on the point cloud range along the pore axis. To ensure accuracy, multi-point averaging and curve / surface smoothing algorithms can be introduced during the calculation to eliminate local measurement noise. The analysis of geometric relationships includes not only the spatial distribution relationship between micropores but also the positional and directional constraints between micropores and the main features of the part. For example, micropores may be distributed in an array on the surface of the part, requiring the calculation of the consistency of the spacing between their centers and the arrangement accuracy; for inclined or intersecting holes, the angle relationship between the hole axis and the reference plane of the part needs to be analyzed. The implementation method is based on the three-dimensional coordinate data of the CAD digital model. Through vector operations and geometric constraint detection algorithms, the relative positions, parallelism, perpendicularity, and symmetry between structures are calculated and verified. To improve reliability, the geometric relationship analysis can be divided into two levels: global analysis and local analysis. Global analysis is used to extract the overall arrangement pattern, while local analysis is used to capture subtle deviations in individual hole positions.

[0038] After calculating the morphological parameters and overall geometric relationships of the microholes, the measurement results need to be corrected for deviations. Sources of deviation include imaging distortion, point cloud fitting errors, and cumulative platform motion errors, all of which affect the final dimensional accuracy. The correction process is based on geometric relationships: for example, when multiple microholes should theoretically be arranged in a straight line, but slight offsets occur in actual measurements, the offset can be corrected using global straight-line fitting; when there are known geometric constraints between the hole axis direction and the reference plane, the hole depth or tilt angle parameters can be compensated based on the included angle deviation. Specifically, least-squares optimization is used to jointly solve for the dimensional parameters of each microhole with global geometric constraints, thereby obtaining correction results that better conform to the actual geometry. The corrected microhole morphological parameters are then geometrically compared with preset theoretical parameters to extract the error values ​​at each measurement point. Theoretical parameters are derived from part design models or standard part specifications, such as design hole diameter, ideal roundness, hole depth, and distribution spacing. The comparison method typically involves calculating the difference between the actual and theoretical parameters and quantifying the error according to different indicators. For example, aperture error is the difference between the actual average aperture and the theoretical aperture; roundness error is the maximum deviation of the aperture boundary point from the ideal fitted circle; and position error is the difference between the coordinates of the aperture center point and the theoretical arrangement coordinates. To ensure the accuracy of error calculation, distributed error mapping can be introduced during the comparison, comparing the point cloud of the entire aperture with the theoretical circle point by point to generate a two-dimensional or three-dimensional error field. Through this process, not only can the overall deviation of each micro-aperture be obtained, but also the minute irregularities caused by processing can be revealed at the local scale.

[0039] In this embodiment, the specific steps of step S5 are as follows: Multi-scale decomposition of the panoramic fused image and surface morphology recognition are performed to obtain the surface morphology data of the part. Roughness data and ripple data are extracted based on the surface morphology data of the part. The coarse data is then subjected to coarseness quantization to obtain a coarseness quantization coefficient. Identify the principal direction and anisotropic characteristics of the ripple data; The surface texture pattern of the part is evaluated based on the main direction, anisotropic characteristics, and roughness quantization coefficient, and surface texture pattern features are generated. The error values ​​at each measurement point are visualized using color deviation and comprehensively evaluated in conjunction with surface texture pattern features.

[0040] In this embodiment, the core idea of ​​scale decomposition is to decompose the image into high-frequency, mid-frequency, and low-frequency components, corresponding to roughness details, ripple features, and macroscopic morphological contours, respectively. This can be achieved using wavelet decomposition or multi-scale Fourier transform to process the grayscale information of the image surface in layers, allowing for the separation of geometric features at different scales. Subsequently, feature extraction methods are used to identify morphological information at different scales, such as using gradient fields and local height differences to determine roughness texture and using low-frequency components to fit the surface ripple direction. Through this process, complex surface texture information can be expressed in a structured data form, yielding complete part surface morphological data. After obtaining the layered surface morphological data, it needs to be further analyzed into roughness data and ripple data. Roughness data mainly comes from high-frequency components, exhibiting micrometer-level surface detail undulations, and its extraction can be achieved through high-pass filtering or local height difference calculation; while ripple data comes from mid-frequency components, reflecting the periodic fluctuation characteristics generated by surface processing, and its extraction can be achieved using band-pass filtering combined with periodic signal fitting. The specific approach involves separating the topographic data into different frequency bands and performing statistical analysis on each band. For example, the distribution density, amplitude, and distribution pattern of wave peaks and troughs are calculated, thereby obtaining independent sets of coarseness and ripple data. This process ensures that coarseness and ripple do not interfere with each other, allowing subsequent quantization and directional analysis to be carried out separately, avoiding misjudgments or calculation errors caused by scale mixing.

[0041] A combined statistical and geometric analysis is performed on the roughness data. First, conventional roughness indices are calculated, such as the average arithmetic height Ra, root mean square height Rq, and maximum height Rz. These parameters can be obtained by averaging the absolute values, squares, or peak-to-valley differences of surface height differences. Second, power spectral density analysis is used to further characterize the spatial distribution of roughness, distinguishing between random and directional roughness. Finally, multiple roughness indices are converted into a single roughness quantification coefficient through weighted combination or normalization methods. This coefficient reflects the degree of surface roughness within a certain range. Through this quantification process, the surface micromorphology is transformed from a qualitative description into a quantitative indicator, providing a unified reference scale for texture pattern evaluation and error calculation. After extracting the ripple data, its principal direction and anisotropic characteristics need to be identified. The principal direction refers to the main extension direction of the surface ripples, which is usually related to the machining process. For example, turning lines are often distributed in concentric circles, while milling lines may be distributed in parallel. The implementation involves performing spectral analysis on the ripple data using a two-dimensional Fast Fourier Transform (FFT) or directional filter to extract the direction corresponding to the region of maximum energy concentration, which is the main ripple direction. Anisotropy identification involves analyzing the distribution differences of the ripples in different directions. The energy ratio in different directions is calculated through directional power spectral density or structure tensor analysis to determine whether the ripples exhibit isotropic or strongly directional distribution characteristics. The identified main direction and anisotropy can not only serve as evaluation indicators of processing stability but also provide directional constraints for texture pattern classification and comparison. A texture pattern analysis framework is constructed, using roughness level as a quantitative description of texture details, the main ripple direction as a control parameter for macroscopic orientation, and anisotropy as an indicator of surface consistency and uniformity. Through multi-dimensional feature fusion, complete surface texture pattern features can be formed. For example, when the surface roughness quantification coefficient is low but the ripple directionality is strong, the surface pattern may exhibit regular processing textures; while when the roughness coefficient is high and the ripple anisotropy is weak, it may exhibit a random roughness pattern. The resulting surface texture pattern features can not only numerically characterize the surface quality of parts, but also provide important basis for roughness error calculation and surface functionality evaluation (such as friction performance and fatigue strength).

[0042] In this embodiment, the specific steps for visualizing the error values ​​at each measurement point using color and comprehensively evaluating them in conjunction with surface texture pattern features are as follows: The error values ​​of each measurement point are located in three dimensions, and the deviation is visualized by color to obtain the deviation color visualization data; Based on the deviation color visualization data, a deviation visualization mapping is performed on the CAD digital model of the part to construct a deviation distribution density map; Based on the deviation distribution density map and surface texture pattern characteristics, a comprehensive evaluation of morphological deviation and roughness is performed, and a comprehensive evaluation report is output.

[0043] In this embodiment, the measurement point error value is matched one-to-one with its coordinates in point cloud data or CAD model to form a spatial coordinate point set with error attributes. Subsequently, these error values ​​undergo color mapping, converting errors of different magnitudes into different color gradients. For example, positions with deviations close to zero are displayed as green, positive deviations gradually transition to red, and negative deviations gradually transition to blue. This method visually reflects the distribution of errors on the spatial model, making the relationship between local deviations and the overall structure immediately apparent. To ensure consistency between visual effects and quantitative indicators, a hierarchical interval method can be used during color mapping, dividing the error range into several level intervals and corresponding them to color ranges, thereby avoiding local oversaturation or distortion. A correspondence is established between the error color data and the mesh or surface units of the CAD model, so that each surface unit or mesh node has an error value and corresponding color information. Through this mapping, the deviation distribution characteristics of each region can be displayed on the complete part CAD model. Furthermore, to analyze the concentration of deviations, spatial density calculations can be performed on the mapping results to construct a deviation distribution density map. This density map not only shows the numerical magnitude of the deviations but also reveals the clustering trends of deviations in different regions. For example, a curved surface area may have a large range of continuous deviations, while some orifice edges may only have localized, isolated deviations. In this way, the deviation distribution density map can comprehensively reveal the systematic and random errors present in the machining and forming process of parts.

[0044] After obtaining the deviation distribution density map and surface texture pattern characteristics, a comprehensive evaluation of the overall geometry and surface quality of the part is required. The evaluation process first involves joint analysis of the deviation distribution and surface texture data: the deviation distribution reveals macroscopic geometric errors, such as aperture deviation, positional deviation, and irregular surface deformation; the texture pattern characteristics reflect the microscopic surface state, including roughness level, ripple directionality, and anisotropy. By constructing a comprehensive evaluation model, macroscopic and microscopic indicators are unified under the same evaluation framework. For example, a weighted multi-index method or fuzzy comprehensive evaluation method is used to combine the deviation density value, roughness quantification coefficient, and texture directionality index into a comprehensive evaluation coefficient. Subsequently, a complete evaluation result is generated based on this coefficient, including a morphological deviation distribution map, roughness level classification, and overall machining consistency evaluation. The final comprehensive evaluation report is presented in a combined text and graphic format, including both visual results (deviation mapping and texture pattern map) and quantitative indicators (error value statistics and roughness level), providing a systematic and traceable basis for quality inspection, process optimization, and error tracing of precision parts.

[0045] In this embodiment, a non-contact micropore morphology and roughness measurement system is provided for performing the non-contact micropore morphology and roughness measurement method described above, including: The detection base point analysis module is used to perform initial contour scanning of precision parts, analyze the detection base points of the parts, and extract the optimal coordinate base points; The panoramic stitching module is used to perform full-coverage scanning planning and panoramic image stitching based on the optimal coordinate base points to obtain a panoramic fused image. The point cloud modeling module is used to perform geometric structure point depth recognition on panoramic fusion images and to perform 3D point cloud modeling to build CAD digital models of parts. The shape and size calculation module is used to calculate the key dimensions of microhole shape based on the CAD digital model of the part, perform geometric comparison calculation, and extract the error value of each measurement point. The error assessment module is used to perform deviation visualization mapping and comprehensive assessment based on the error values ​​of each measurement point, and output a comprehensive assessment report.

[0046] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0047] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein are implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A non-contact method for measuring the morphology and roughness of micropores, characterized in that, Includes the following steps: Step S1: Perform initial contour scanning on the precision part and analyze the part inspection base points to extract the optimal coordinate base points; Step S2: Based on the optimal coordinate base point, perform full-coverage scanning planning and panoramic image stitching to obtain a panoramic fused image; Step S3: Perform depth recognition of geometric structure points on the panoramic fusion image, and perform 3D point cloud modeling to construct a CAD digital model of the part; Step S4: Calculate the key dimensions of the micropore morphology based on the CAD digital model of the part, perform geometric comparison calculations, and extract the error values ​​of each measurement point; Step S5: Perform deviation visualization mapping and comprehensive evaluation based on the error values ​​of each measurement point, and output a comprehensive evaluation report.

2. The non-contact micropore morphology and roughness measurement method according to claim 1, characterized in that, The specific steps of step S1 are as follows: An initial contour scan of a precision part is performed using a high-precision motion measurement platform to identify the boundary feature points of the precision part. The boundary feature points are located in three-dimensional space, and their three-dimensional coordinates are extracted. The overall shape is evaluated based on three-dimensional positioning coordinates to obtain the main geometric features; Based on the above, topological correlation analysis is performed on the boundary feature points, and the symmetry axis distribution is identified to extract the symmetry axis distribution information; Calculate the geometric center of gravity and principal direction of the part based on the main geometric features and the distribution information of the axis of symmetry; Based on the geometric center and main direction of the part, the part detection base point is analyzed, and the optimal coordinate base point is extracted.

3. The non-contact micropore morphology and roughness measurement method according to claim 1, characterized in that, The specific steps of step S2 are as follows: Regional complexity analysis is performed based on key geometric features to identify the geometric complexity of different parts of precision components; Based on the aforementioned geometric complexity, an adaptive scanning method is defined, resulting in the adaptive scanning method. Full coverage scanning planning is performed based on optimal coordinate base points and adaptive scanning method to generate scanning planning path; Based on the scanning planning path, a full-range measurement and imaging of precision parts is performed, and the full-range measurement image is extracted. A panoramic fused image is obtained by stitching together panoramic images from all-around measurement images.

4. The non-contact micropore morphology and roughness measurement method according to claim 3, characterized in that, The adaptive scanning method is specifically defined as follows: Based on the aforementioned geometric complexity, flat regions, curved regions, edge regions, and hole regions are divided. The flat area is divided into a regular grid for grid scanning; Calculate the principal curvature directions of the surface region; Perform a spiral scan based on the main curvature direction; identify the edge lines of the edge and hole regions; Contour tracking scan is performed based on the edge lines.

5. The non-contact micropore morphology and roughness measurement method according to claim 1, characterized in that, Step S3 is as follows: Perform depth recognition of geometric structure points in panoramic fusion images and mark the geometric structure points; Spatial positioning of geometric structural points generates structural point location information; Perform type parsing on the geometric structure points to obtain the structure point type; Discrete point cloud conversion is performed on the panoramic fusion image to generate discrete point cloud data of the parts; The discrete point cloud data of the part is labeled according to the type of structural points, and a 3D point cloud model is constructed based on the location information of the structural points to build a CAD digital model of the part.

6. The non-contact micropore morphology and roughness measurement method according to claim 1, characterized in that, The specific steps of step S4 are as follows: Calculate the key dimensions of micropore morphology based on the CAD digital model of the part, and extract the morphological and dimensional parameters of multiple micropore morphologies. Geometric relationship analysis is performed on the CAD digital model of the parts to obtain the geometric relationships between structures; Based on the geometric relationship between the structures, the morphological and dimensional parameters are measured and the deviation is corrected to obtain the micropore corrected morphological parameters; Based on the preset theoretical micropore morphology parameters, the micropore correction morphology parameters are geometrically compared and calculated, and the error values ​​of each measurement point are extracted.

7. The non-contact micropore morphology and roughness measurement method according to claim 1, characterized in that, The specific steps of step S5 are as follows: Multi-scale decomposition of the panoramic fused image and surface morphology recognition are performed to obtain the surface morphology data of the part. Roughness data and ripple data are extracted based on the surface morphology data of the part. The coarse data is then subjected to coarseness quantization to obtain a coarseness quantization coefficient. Identify the principal direction and anisotropic characteristics of the ripple data; The surface texture pattern of the part is evaluated based on the main direction, anisotropic characteristics, and roughness quantization coefficient, and surface texture pattern features are generated. The error values ​​at each measurement point are visualized using color deviation and comprehensively evaluated in conjunction with surface texture pattern features.

8. The non-contact micropore morphology and roughness measurement method according to claim 7, characterized in that, The specific steps for visualizing the error values ​​at each measurement point using color and combining them with surface texture pattern features for comprehensive evaluation are as follows: The error values ​​of each measurement point are located in three dimensions, and the deviation is visualized by color to obtain the deviation color visualization data; Based on the deviation color visualization data, a deviation visualization mapping is performed on the CAD digital model of the part to construct a deviation distribution density map; Based on the deviation distribution density map and surface texture pattern characteristics, a comprehensive evaluation of morphological deviation and roughness is performed, and a comprehensive evaluation report is output.

9. A non-contact micropore morphology and roughness measurement system, characterized in that, A method for performing the non-contact micropore morphology and roughness measurement as described in claim 1, comprising: The detection base point analysis module is used to perform initial contour scanning of precision parts, analyze the detection base points of the parts, and extract the optimal coordinate base points; The panoramic stitching module is used to perform full-coverage scanning planning and panoramic image stitching based on the optimal coordinate base points to obtain a panoramic fused image. The point cloud modeling module is used to perform geometric structure point depth recognition on panoramic fusion images and to perform 3D point cloud modeling to build CAD digital models of parts. The shape and size calculation module is used to calculate the key dimensions of microhole shape based on the CAD digital model of the part, perform geometric comparison calculation, and extract the error value of each measurement point. The error assessment module is used to perform deviation visualization mapping and comprehensive assessment based on the error values ​​of each measurement point, and output a comprehensive assessment report.

Images