A three-dimensional scanning efficiency and precision balancing method and system based on an adaptive sampling strategy

Abstract

The application relates to the technical field of three-dimensional scanning, and particularly discloses a three-dimensional scanning efficiency and precision balancing method and system based on an adaptive sampling strategy. The application obtains prior data and scanning state data of a target object, fuses initial three-dimensional point clouds, multispectral reflection images and reflection intensity information, accurately extracts and corrects surface curvature distribution characteristics, generates a local geometric feature vector sequence in combination with geometric and optical characteristics, quantizes scanning priorities, marks a scanning path based on the sequence, constructs a scanning strategy matrix in combination with device geometric positioning accuracy, maximum scanning speed and time constraints, realizes differentiated scanning parameter matching, dynamically generates a local compensation sub-path through real-time quality monitoring during scanning, splices a main path and a compensation scanning path to output a result, balances scanning efficiency and precision in a targeted manner, avoids local data loss or quality decline, guarantees the precision of key areas, and improves overall scanning efficiency.

Description

Technical Field

[0001] This invention relates to the field of 3D scanning technology, and in particular to a method and system for balancing 3D scanning efficiency and accuracy based on an adaptive sampling strategy. Background Technology

[0002] 3D scanning technology has been widely used in reverse engineering, precision inspection, cultural relic digitization, robot navigation and other fields. Its core goal is to acquire complete 3D point cloud data of the surface of a target object efficiently and with high accuracy within a limited time. However, there is an inherent contradiction between scanning efficiency and accuracy: increasing scanning speed (efficiency) usually means reducing sampling density, which may miss fine features or lead to sparse point clouds, affecting reconstruction accuracy; while pursuing high accuracy requires reducing speed and increasing sampling density, which leads to a significant increase in scanning time.

[0003] However, existing 3D scanning path planning methods are usually based on preset uniform sampling strategies or simple geometric features (such as normal vectors) for path planning. They lack depth perception and dynamic response capabilities for the complex geometric and optical properties of the object surface. As a result, when the device encounters reflections, occlusions, or slight device drift, it is easy to cause local data loss or quality degradation. Therefore, a 3D scanning efficiency and accuracy balancing method based on adaptive sampling strategies is needed to solve the above problems. Summary of the Invention

[0004] The purpose of this invention is to provide a method for balancing the efficiency and accuracy of three-dimensional scanning based on an adaptive sampling strategy, comprising:

[0005] A method for balancing efficiency and accuracy in 3D scanning based on an adaptive sampling strategy includes:

[0006] Acquire 3D scan data of the target object, which includes prior data of the target object and scan state data;

[0007] Based on the prior data of the target object, obtain the initial three-dimensional point cloud data, multispectral reflectance image data and reflectance intensity information of the target object, and obtain the surface curvature distribution characteristics of the target object based on the initial three-dimensional point cloud data and the multispectral reflectance image data;

[0008] A sequence of local geometric feature vectors is obtained based on the surface curvature distribution characteristics and the reflection intensity information;

[0009] An initial scanning path is obtained based on the initial 3D point cloud data, and the initial scanning path is marked based on the local geometric feature vector sequence to obtain a marked scanning path;

[0010] The geometric positioning accuracy parameters and maximum scanning speed are obtained based on the scanning state data, and a scanning strategy matrix is ​​constructed based on the local geometric feature vector sequence, the geometric positioning accuracy parameters, and the maximum scanning speed.

[0011] The marked scanning path is scanned in a balance between efficiency and accuracy according to the scanning strategy matrix, and the scanning results are output.

[0012] Preferably, the step of obtaining the surface curvature distribution characteristics of the target object based on the initial three-dimensional point cloud data and the multispectral reflectance image data includes:

[0013] The spatial mesh corresponding to the target object is obtained based on the initial three-dimensional point cloud data, and the spatial mesh is divided into non-overlapping parts based on a preset local region to obtain multiple local three-dimensional mesh units.

[0014] The geometric feature values ​​within each local three-dimensional mesh cell are obtained based on the plurality of local three-dimensional mesh cells, wherein the geometric feature values ​​include the local principal curvature and the rate of change of the normal vector;

[0015] Multiple local curvature feature values ​​are obtained based on multiple local principal curvatures and multiple normal vector change rates;

[0016] The initial 3D point cloud data and the multispectral reflectance image data are registered to obtain 3D point cloud-multispectral reflectance image reference data, and each local 3D mesh cell is assigned a corresponding local spectral feature based on the 3D point cloud-multispectral reflectance image reference data.

[0017] The local curvature feature values ​​are corrected based on the local spectral features to obtain the corrected local curvature feature values, and the corrected local curvature feature values ​​are used as the surface curvature distribution features.

[0018] Preferably, the step of obtaining a local geometric feature vector sequence based on the surface curvature distribution characteristics and the reflection intensity information includes:

[0019] Based on the surface curvature distribution characteristics, the curvature extreme points of each region on the surface of the target object are extracted, and multiple curvature extreme points are subjected to region growing to generate multiple initial geometric feature regions, wherein each initial geometric feature region corresponds to an average curvature value.

[0020] Based on the reflection intensity information, multiple reflection intensities of the target object within a preset time period are obtained, and an average reflection intensity is obtained based on the multiple reflection intensities. A standard reflection intensity difference is obtained based on the average reflection intensity and the multiple reflection intensities.

[0021] The reflection intensity feature value is obtained based on the difference between the average reflection intensity and the standard reflection intensity, and the reflection intensity feature value is sequentially weighted and fused with multiple average curvature values ​​to obtain multiple local feature fusion coefficients.

[0022] The local feature fusion coefficients are concatenated into an incremental vector to obtain a sequence of local geometric feature vectors.

[0023] Preferably, the step of obtaining an initial scan path based on the initial 3D point cloud data and marking the initial scan path based on the local geometric feature vector sequence to obtain a marked scan path includes:

[0024] Based on the initial 3D point cloud data, the corresponding spatial bounding box and normal vector field are obtained, and an initial scanning path covering all visible surfaces of the target object is generated by the spatial bounding box and the normal vector field based on the spatial filling curve algorithm. The initial scanning path includes multiple path points.

[0025] The local geometric feature vector sequence is mapped to each path point of the initial scan path, and a feature label is assigned to each path point, wherein the feature label contains a sequence priority value;

[0026] The relationship between multiple sequence priority values ​​and a preset sequence priority value range is determined sequentially, and corresponding scan action identifiers are assigned to the multiple sequence priority values. The scan action identifiers include fast scan identifiers, standard scan identifiers, and fine scan identifiers.

[0027] When the sequence priority value is less than the minimum value of the preset sequence priority value range, the path point corresponding to the sequence priority value is mapped to a fast scan identifier.

[0028] When the sequence priority value is within the preset sequence priority value range, the path point corresponding to the sequence priority value is mapped to a standard scan identifier.

[0029] When the sequence priority value is greater than the maximum value of the preset sequence priority value range, the path point corresponding to the sequence priority value is mapped to a fine scan identifier.

[0030] The path points assigned to the scan action identifier in the multiple sequence priority values ​​are optimized by interval merging, and adjacent path points with the same scan identifier are merged into the same scan path segment.

[0031] A marked scan path is generated based on multiple scan path segments and their corresponding scan action identifiers.

[0032] Preferably, the step of constructing the scanning strategy matrix based on the local geometric feature vector sequence, the geometric positioning accuracy parameter, and the maximum scanning speed includes:

[0033] The minimum feasible point distance is obtained based on the geometric positioning accuracy parameters.

[0034] Multiple sampling densities are obtained based on the local geometric feature vector sequence and the minimum feasible point distance;

[0035] Obtain the total task pre-constraint time, and construct a scanning strategy matrix based on the total task pre-constraint time, multiple sampling densities, and the maximum scanning speed.

[0036] Preferably, the step of performing an efficiency- and accuracy-balanced scan of the marked scan path according to the scan strategy matrix and outputting the scan result includes:

[0037] Multiple spatial coordinates are obtained according to the marked scanning path, and the multiple spatial coordinates are paired according to the scanning strategy matrix to obtain multiple pairing strategies, wherein the pairing strategy includes target scanning speed and target point density;

[0038] The preset 3D scanning device is started and moves along the marked scanning path. The marked scanning path is scanned in a balance between efficiency and accuracy according to the multiple target scanning speeds and multiple target point densities to obtain the initial scanning result. The initial scanning result includes real-time acquisition of multiple point cloud data subsets acquired in the current scanning path segment, and acquisition of the point cloud density uniformity and point cloud fitting residual of each point cloud data subset.

[0039] Multiple real-time quality indicators are obtained by weighting the density uniformity of multiple points cloud and the fitting residuals of multiple points cloud;

[0040] The real-time quality indicators are compared sequentially with preset quality thresholds;

[0041] If the real-time quality index is greater than or equal to the preset quality threshold, then the preset 3D scanning device is controlled to continue scanning the next path segment until the main body path scanning is completed, and the main body scanning path is obtained.

[0042] If the real-time quality index is less than the preset quality threshold, the position of the current scanning path segment and the scanning strategy matrix corresponding to the real-time quality index are obtained, and a local compensation scanning sub-path is dynamically generated according to the position of the current scanning path segment and the scanning strategy matrix.

[0043] The main scanning path and multiple local compensation scanning sub-paths are concatenated to obtain the scanning path result.

[0044] This application also provides a 3D scanning efficiency and accuracy balancing system based on an adaptive sampling strategy, including:

[0045] The first acquisition module is used to acquire the three-dimensional scan data of the target object, wherein the three-dimensional scan data includes the prior data of the target object and the scan state data;

[0046] The second acquisition module is used to acquire the initial three-dimensional point cloud data, multispectral reflectance image data and reflectance intensity information of the target object based on the prior data of the target object, and to acquire the surface curvature distribution characteristics of the target object based on the initial three-dimensional point cloud data and the multispectral reflectance image data.

[0047] The third acquisition module is used to acquire a sequence of local geometric feature vectors based on the surface curvature distribution characteristics and the reflection intensity information;

[0048] The fourth acquisition module is used to acquire an initial scanning path based on the initial 3D point cloud data, and to mark the initial scanning path based on the local geometric feature vector sequence to obtain a marked scanning path;

[0049] The fifth acquisition module is used to acquire geometric positioning accuracy parameters and maximum scanning speed based on the scanning state data, and to construct a scanning strategy matrix based on the local geometric feature vector sequence, the geometric positioning accuracy parameters, and the maximum scanning speed;

[0050] The scanning module is used to perform a balanced scanning of the marked scanning path based on the scanning strategy matrix, and output the scanning results.

[0051] Preferably, the second acquisition module includes:

[0052] The first acquisition unit is used to acquire the spatial grid corresponding to the target object based on the initial three-dimensional point cloud data, and to divide the spatial grid into non-overlapping parts based on a preset local region to obtain multiple local three-dimensional grid units.

[0053] The second acquisition unit is used to acquire geometric feature values ​​within each local three-dimensional mesh cell based on the plurality of local three-dimensional mesh cells, wherein the geometric feature values ​​include local principal curvature and normal vector change rate;

[0054] The third acquisition unit is used to acquire multiple local curvature feature values ​​based on multiple local principal curvatures and multiple normal vector change rates;

[0055] The registration unit is used to register the initial three-dimensional point cloud data and the multispectral reflectance image data to obtain three-dimensional point cloud-multispectral reflectance image reference data, and to assign corresponding local spectral features to each local three-dimensional mesh unit according to the three-dimensional point cloud-multispectral reflectance image reference data.

[0056] The correction unit is used to correct the multiple local curvature feature values ​​according to the multiple local spectral features, to obtain multiple corrected local curvature feature values, and to use the multiple corrected local curvature feature values ​​as surface curvature distribution features.

[0057] This application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0058] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0059] The beneficial effects of this application are as follows: This invention obtains prior data and scanning state data of the target object, integrates initial three-dimensional point cloud, multispectral reflectance image and reflectance intensity information, accurately extracts and corrects surface curvature distribution features, and then generates a local geometric feature vector sequence by combining geometric and optical properties, quantifies scanning priority, marks the scanning path based on the sequence, and constructs a scanning strategy matrix by combining the device's geometric positioning accuracy, maximum scanning speed and time constraints to achieve differentiated scanning parameter matching. During scanning, local compensation sub-paths are dynamically generated through real-time quality monitoring, and the main path and supplementary scanning path are spliced ​​together to output the results. In this way, scanning efficiency and accuracy are balanced in a targeted manner, avoiding local data loss or quality degradation, ensuring the accuracy of key areas and improving the overall scanning efficiency. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of a method flow according to an embodiment of this application.

[0061] Figure 2 This is a schematic diagram of the system structure according to an embodiment of this application.

[0062] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0063] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0064] like Figures 1-2 As shown, this application provides a method for balancing the efficiency and accuracy of 3D scanning based on an adaptive sampling strategy, including:

[0065] A method for balancing efficiency and accuracy in 3D scanning based on an adaptive sampling strategy includes:

[0066] S1. Obtain the three-dimensional scan data of the target object, wherein the three-dimensional scan data includes the prior data of the target object and the scan state data;

[0067] S2. Obtain the initial three-dimensional point cloud data, multispectral reflectance image data and reflectance intensity information of the target object based on the prior data of the target object, and obtain the surface curvature distribution characteristics of the target object based on the initial three-dimensional point cloud data and the multispectral reflectance image data;

[0068] S3. Obtain a sequence of local geometric feature vectors based on the surface curvature distribution characteristics and the reflection intensity information;

[0069] S4. Obtain the initial scanning path based on the initial 3D point cloud data, and mark the initial scanning path based on the local geometric feature vector sequence to obtain the marked scanning path;

[0070] S5. Obtain the geometric positioning accuracy parameters and maximum scanning speed based on the scanning status data, and construct a scanning strategy matrix based on the local geometric feature vector sequence, the geometric positioning accuracy parameters, and the maximum scanning speed;

[0071] S6. Perform a balanced scan of the marked scanning path based on the scanning strategy matrix, and output the scan results.

[0072] As described in steps S1-S6 above, the core requirement of 3D scanning is to acquire complete and accurate 3D information of the target object's surface within a limited time. However, the complexity of the object's surface geometry (such as differences in curvature in different areas) and the diversity of its optical properties (such as changes in reflection intensity) lead to an inherent contradiction between efficiency and accuracy during the scanning process. Rapid scanning is prone to missing fine features due to insufficient sampling, while high-precision scanning will prolong the time due to excessively high sampling density. At the same time, the limitations of the scanning equipment's geometric positioning accuracy, the maximum scanning speed constraint, and the possible occurrence of reflections, occlusions, or slight equipment drift during the scanning process further exacerbate the problem of unstable local data quality. Therefore, this invention first acquires the 3D scanning data of the target object, which includes prior data of the target object and scanning status data. The prior data of the target object can be obtained through pre-scanning, calling 3D model libraries, or acquiring data from visual imaging equipment. It contains initial information related to the object's basic geometric contour and surface optical reflection characteristics. The scanning status data is acquired in real time by the 3D scanning equipment and covers the equipment's current geometric positioning accuracy parameters (such as positioning error range), maximum scanning speed (the limit of movement speed determined by the equipment's hardware performance), and other operating status parameters. This provides the foundational input data for formulating subsequent scanning strategies. Only by fully understanding the characteristics of the object itself and the operating status of the equipment can the targetedness and feasibility of the subsequent strategy design be ensured. For example, in the scenario of scanning cultural relics, prior data can be obtained through low-precision rapid pre-scanning to acquire initial information on the approximate outline and surface reflectivity of the cultural relic. The scanning status data is then fed back in real time by the laser scanning equipment, providing data support for subsequent scanning strategies that distinguish between areas of fine decoration and smooth areas of the cultural relic.

[0073] Initial 3D point cloud data, multispectral reflectance image data, and reflectance intensity information are obtained based on prior data of the target object. The surface curvature distribution characteristics of the target object are then extracted by combining the initial 3D point cloud data and the multispectral reflectance image data. The initial 3D point cloud data is obtained through a preliminary scan of the target object using a 3D scanning device. It is a dataset composed of the 3D coordinates of a large number of discrete points, reflecting the basic spatial morphology of the object. The multispectral reflectance image data is acquired using a multispectral imaging device, capturing the reflectance characteristics of the object in different spectral bands. The reflectance intensity information is the intensity data of the received signal after the detection signal (such as laser) emitted by the scanning device is reflected from the object's surface, reflecting the object's surface reflectivity to the detection signal. After acquiring this basic data, a spatial mesh corresponding to the target object is first constructed based on the initial 3D point cloud data. Then, the spatial mesh is divided into multiple non-overlapping local 3D mesh units according to a preset local region to ensure partitioned coverage of the object's surface. Subsequently, the local principal curvature (the core parameter reflecting the degree of curvature of the mesh unit's surface) and the rate of change of the normal vector (the parameter reflecting the change of the surface direction of the mesh unit) within each local 3D mesh unit are calculated, and the local curvature feature value is obtained accordingly. Next, the initial 3D point cloud data is registered with the multispectral reflectance image data to make the spatial positions of the point cloud data and the image data correspond one-to-one, forming 3D point cloud-multispectral reflectance image reference data. Then, each local 3D mesh unit is assigned a corresponding local spectral feature. Finally, the local curvature feature value is corrected using the local spectral feature to obtain the corrected local curvature feature value, which is the surface curvature distribution feature. This precise characterization of the geometric differences across different regions of an object's surface, combined with correction using multispectral reflectance image data, is crucial because the spectral reflectance characteristics of an object's surface are often related to its geometric shape (e.g., the spectral reflectance of a rough surface differs from that of a smooth surface, and rough surfaces typically exhibit more complex curvature variations). The corrected surface curvature distribution features more accurately reflect the object's surface geometry. For instance, in scanning automotive parts, the local principal curvature and rate of change of the normal vector in areas such as grooves and sharp edges are significantly higher than in planar areas. Correction using multispectral reflectance image data allows for a more accurate distinction between these high-curvature feature areas and smooth areas, providing a basis for subsequent differentiated scanning strategies.

[0074] Local geometric feature vector sequences are obtained based on surface curvature distribution characteristics and reflection intensity information. First, curvature extrema (points with significantly higher curvature values ​​than surrounding areas, typically corresponding to key surface features such as edges and textures) are extracted from the surface curvature distribution characteristics. These curvature extrema are then expanded into multiple initial geometric feature regions using a region growing algorithm, each corresponding to an average curvature value, thus aggregating the surface feature regions. Next, based on reflection intensity information, multiple reflection intensity data points within a preset time period are statistically analyzed to calculate the difference between the average and standard reflection intensity (reflecting the fluctuation of reflection intensity). Reflection intensity feature values ​​are then extracted, comprehensively reflecting the surface's ability to reflect and the stability of the detection signal. Subsequently, the reflection intensity feature values ​​are weighted and fused with the average curvature value of each initial geometric feature region to obtain multiple local feature fusion coefficients. The weights during fusion are determined based on the influence of geometric features and reflection characteristics on scanning quality in the application scenario. Finally, these local feature fusion coefficients are concatenated in ascending order to form a local geometric feature vector sequence. This method fuses the geometric and optical reflection properties of an object's surface to form a feature sequence that quantifies the scanning priority of each region, providing a core basis for marking subsequent scanning paths. For example, in precision mold scanning, the cavity texture region of the mold has a high density of curvature extrema and a high average curvature value. Furthermore, if the surface material's reflection intensity fluctuates significantly in this region (large difference in standard reflection intensity), the corresponding local feature fusion coefficient will be larger, placing it with higher priority in the vector sequence. Subsequent scans will then provide greater accuracy assurance for this region.

[0075] An initial scan path is obtained based on the initial 3D point cloud data, and then marked using a sequence of local geometric feature vectors to obtain a marked scan path. First, the spatial bounding box (the smallest cube that completely encloses the object, defining the scan area) and normal vector field (the field composed of the normal vectors of each point cloud point, reflecting the orientation of each point on the object's surface) of the target object are determined based on the initial 3D point cloud data. Then, a space-filling curve algorithm (a curve algorithm that can efficiently cover two-dimensional or three-dimensional space, such as the Hilbert curve) is used to generate an initial scan path covering the entire visible surface of the object. This path contains multiple consecutive path points to ensure no omissions in the scan. Next, the sequence of local geometric feature vectors is mapped to each path point of the initial scan path, assigning a feature label containing a sequence priority value to each path point. The sequence priority value directly corresponds to the magnitude of the local feature fusion coefficient, quantifying the scan priority of the region where each path point is located. A preset sequence priority value range is then set. By judging the relationship between the sequence priority value of each path point and this range, a corresponding scanning action identifier is assigned to the path point. When the sequence priority value is less than the minimum value of the range, a fast scanning identifier is assigned (suitable for areas with low scanning priority and simple geometric features). When the sequence priority value is within the range, a standard scanning identifier is assigned (suitable for areas with medium geometric features and scanning priority). When the sequence priority value is greater than the maximum value of the range, a fine scanning identifier is assigned (suitable for areas with high scanning priority and complex geometric features). Then, consecutive path points assigned the same scanning action identifier are optimized by range merging to form multiple scanning path segments. Finally, based on these scanning path segments and their corresponding scanning action identifiers, a marked scanning path is generated. This quantifies the scanning priority, which integrates geometric and optical characteristics, into specific scanning path markers, providing a clear path basis for the execution of subsequent differentiated scanning strategies. Range merging optimization reduces action switching during the scanning process and improves scanning efficiency. For example, in scanning building components, the path point sequence corresponding to the decorative carving area of ​​the component has a high priority value and is marked as a fine scan identifier, while the path point sequence corresponding to the flat side of the component has a low priority value and is marked as a fast scan identifier. After merging, they form a fine scan path segment for the carving area and a fast scan path segment for the side, allowing the scanning process to be adjusted accordingly.

[0076] The geometric positioning accuracy parameters and maximum scanning speed are obtained based on the scanning status data. A scanning strategy matrix is ​​then constructed by combining the local geometric feature vector sequence, the geometric positioning accuracy parameters, and the maximum scanning speed. The geometric positioning accuracy parameter represents the maximum positioning accuracy achievable by the scanning device under its current operating state, determining the minimum feasible distance between adjacent sampling points during the scanning process (i.e., the minimum feasible point distance; the higher the positioning accuracy, the smaller the minimum feasible point distance). The maximum scanning speed is the limit of movement speed determined by both the device's hardware performance and its operating state, thus limiting the upper limit of scanning efficiency. First, the minimum feasible point distance is calculated based on the geometric positioning accuracy parameters to ensure that the distance between sampling points is not less than this value. This avoids data redundancy or exceeding the device's positioning accuracy range due to excessively dense sampling points. Then, the scanning priority of each scanning path segment is determined based on the local geometric feature vector sequence. Combined with the minimum feasible point distance, corresponding sampling densities are assigned to path segments with different priorities (higher scanning priority means higher sampling density, i.e., more sampling points per unit length). Finally, the total task pre-constraint time (the upper limit of scanning time set by the user according to actual needs) is obtained. Combining the total task pre-constraint time, the sampling density of each path segment, and the device's maximum scanning speed, a scanning strategy matrix is ​​constructed. This matrix contains the matching relationship between the target scanning speed and target point density for each scanning path segment, ensuring that within the total time constraint, a balance between efficiency and accuracy is achieved through different combinations of speed and point density for different path segments. This combines device performance constraints, scanning time constraints, and regional scanning priorities to form an executable quantitative scanning strategy, avoiding inefficiency or insufficient accuracy caused by unconstrained speed or density settings. For example, in batch scanning scenarios of industrial parts, where the total pre-constraint time is relatively short, for critical mating surfaces of the parts (high-priority path segments), a smaller minimum feasible point distance and a higher sampling density are set based on the equipment's geometric positioning accuracy parameters. Simultaneously, the target scanning speed is appropriately reduced within the equipment's maximum scanning speed range. For non-matting surfaces of the parts (low-priority path segments), a larger minimum feasible point distance and a lower sampling density are set to increase the target scanning speed. This combination ensures that the scanning is completed within the specified time, and the accuracy of critical areas meets the standards.

[0077] The scanning strategy matrix is ​​used to perform a balanced scanning of the marked scanning path, balancing efficiency and accuracy, and outputs the scanning results. First, multiple spatial coordinates of each path segment are extracted from the marked scanning path. Based on the scanning strategy matrix, target scanning speed and target point density are paired for each path point corresponding to the spatial coordinates, forming specific scanning execution parameters. Then, a preset 3D scanning device is started and moves along the marked scanning path, scanning according to the paired target scanning speed and target point density. Multiple point cloud data subsets of the current scanning path segment are collected in real time, and the point cloud density uniformity (reflecting the uniformity of the spatial distribution of point cloud data) and point cloud fitting residual (reflecting the deviation between the point cloud data and the ideal geometric model) of each point cloud data subset are calculated. These two parameters are weighted to obtain a real-time quality index. The weights are determined according to the requirements of density uniformity and fitting accuracy in the application scenario. Finally, the real-time quality index is compared with a preset quality index. A threshold (set according to the accuracy requirements of the scanning task) is compared. If the real-time quality index is greater than or equal to the preset quality threshold, it means that the scanning quality of the current path segment meets the standard, and the control device continues to scan the next path segment until the main path scanning is completed, obtaining the main scanning path. If the real-time quality index is less than the preset quality threshold, it means that the scanning quality of the current path segment does not meet the standard, possibly due to missing data or insufficient accuracy. At this time, based on the position of the current scanning path segment and the scanning strategy matrix, a local compensation scanning sub-path is dynamically generated. This sub-path uses a higher sampling density and a slower scanning speed to supplement the scanning of the area with substandard quality. Finally, the main scanning path and multiple local compensation scanning sub-paths are stitched together to form a complete scanning path result, completing the acquisition and output of scanning data. In this way, closed-loop control of the scanning process is achieved through real-time quality feedback, dynamically compensating for areas with substandard scanning quality, ensuring that the overall scanning result meets the accuracy requirements while maximizing the utilization of equipment performance to improve efficiency. For example, in the digital scanning of cultural relics, when scanning the fine decorative areas of a cultural relic, if the real-time calculated point cloud fitting residual is too large, causing the real-time quality index to be lower than the preset threshold, it indicates that the current scan has failed to accurately capture the decorative details. At this time, the system will generate a local compensation scanning sub-path based on the location of the area and the original scanning strategy matrix, and rescan the area with a higher sampling density and a slower speed to supplement the detailed data, ultimately ensuring that the decorative accuracy of the digital model of the cultural relic meets the standard. At the same time, other smooth areas are completed according to the original fast scanning strategy, balancing efficiency and accuracy. This can effectively solve the problems of difficulty in balancing efficiency and accuracy, and easy loss or quality degradation of local data in existing technologies.

[0078] In one embodiment, step S2, which involves obtaining the surface curvature distribution features of the target object based on the initial three-dimensional point cloud data and the multispectral reflectance image data, includes:

[0079] S21. Obtain the spatial grid corresponding to the target object based on the initial three-dimensional point cloud data, and divide the spatial grid into non-overlapping parts based on a preset local region to obtain multiple local three-dimensional grid units.

[0080] S22. Obtain geometric feature values ​​within each local three-dimensional mesh cell based on the plurality of local three-dimensional mesh cells, wherein the geometric feature values ​​include local principal curvature and normal vector change rate;

[0081] S23. Obtain multiple local curvature feature values ​​based on multiple local principal curvatures and multiple normal vector change rates;

[0082] S24. Register the initial three-dimensional point cloud data and the multispectral reflectance image data to obtain three-dimensional point cloud-multispectral reflectance image reference data, and assign corresponding local spectral features to each local three-dimensional mesh unit according to the three-dimensional point cloud-multispectral reflectance image reference data;

[0083] S25. Correct the multiple local curvature feature values ​​according to the multiple local spectral features to obtain multiple corrected local curvature feature values, and use the multiple corrected local curvature feature values ​​as surface curvature distribution features.

[0084] As described in steps S21-S25 above, this invention obtains the spatial mesh corresponding to the target object based on the initial 3D point cloud data, and divides the spatial mesh into non-overlapping units based on a preset local region to obtain multiple local 3D mesh units. The initial 3D point cloud data is obtained by a 3D scanning device through a preliminary scan of the target object. It is a dataset composed of a large number of discrete 3D coordinate points, which can reflect the basic spatial contour of the object. Through spatial mesh construction, the discrete point cloud data can be transformed into a structured mesh model, making the geometric shape of the object surface easier to analyze. The non-overlapping division according to the preset local region is to achieve uniform coverage and fine segmentation of the object surface, ensuring that each local 3D mesh unit can correspond to a small area of ​​the object surface, avoiding the local geometric features from being averaged due to the large area, which would fail to accurately reflect the local details. For example, when scanning a ceramic artifact with fine patterns, its initial 3D point cloud data contains discrete points of the overall contour of the artifact and the details of the patterns. By constructing a spatial mesh and dividing it into multiple local 3D mesh units, each small area where the pattern details are located can correspond to an independent mesh unit, laying the foundation for the subsequent extraction of accurate geometric features of the area.

[0085] Geometric feature values ​​are obtained for each local 3D mesh element, including local principal curvature and rate of change of normal vector. Local principal curvature is the core parameter describing the degree of surface curvature of the mesh element; a larger principal curvature value indicates more severe surface curvature in that area (e.g., raised patterns on artifacts or sharp edges of parts). The rate of change of normal vector reflects the rate of change of the surface normal direction; a high rate of change of normal vector indicates a rapid change in surface orientation (e.g., areas of curved transition). These two parameters jointly characterize the geometric features of the local mesh element from different dimensions. By calculating these two parameters for each local 3D mesh element, the geometric complexity of different areas on an object's surface can be quantitatively distinguished. For example, in a car bumper scan, the rate of change of normal vector is higher in the curved transition area of ​​the bumper, while it is lower in the flat mounting surface. By extracting these two parameters, areas with different geometric characteristics can be clearly defined.

[0086] Multiple local curvature feature values ​​are obtained based on multiple local principal curvatures and multiple normal vector change rates. This step integrates the local principal curvatures and normal vector change rates through comprehensive calculation, fusing the geometric parameters of the two dimensions into a unified local curvature feature value, thereby achieving an overall quantification of the geometric complexity of local mesh units. The fusion calculation method can be set according to the application scenario requirements. For example, a weighted summation method can be used, assigning corresponding weights to the local principal curvatures and normal vector change rates respectively, and then adding them together to obtain the local curvature feature value. The weight setting is determined based on the importance of the two parameters to the geometric complexity. For example, in precision mold scanning, the local principal curvatures have a greater impact on the accuracy of the mold cavity, and can be assigned a higher weight, so that the local curvature feature value focuses more on reflecting the degree of curvature, ensuring that the subsequent scanning strategy can specifically guarantee the accuracy of the bending details of the cavity.

[0087] The initial 3D point cloud data and multispectral reflectance image data are registered to obtain 3D point cloud-multispectral reflectance image reference data, and each local 3D mesh cell is assigned a corresponding local spectral feature. The multispectral reflectance image data is acquired by a multispectral imaging device, which can capture the reflectance intensity information of an object in different spectral bands. The reflectance characteristics of object surfaces with different materials and roughnesses differ under multispectral conditions (e.g., the multispectral reflectance curves of metal surfaces and plastic surfaces are different). Data registration uses algorithms to precisely align the spatial positions of the 3D point cloud data and the multispectral reflectance image data, ensuring a one-to-one correspondence between each point cloud point and an image pixel. This allows each local 3D mesh cell to be associated with corresponding multispectral reflectance information, i.e., local spectral features. The core significance of this step is to establish the correlation between geometric data and optical data, providing data support for subsequent correction of geometric features using optical properties. For example, in scanning metal parts, the multispectral reflectance characteristics of the oxide layer surface of the part are different from those of the unoxidized metal surface. The oxide layer surface is usually rougher and has more complex geometric features. By registering each grid cell with corresponding local spectral features, it is possible to determine whether the region is an oxide layer through the spectral features and then correct its geometric feature values.

[0088] Multiple local curvature feature values ​​are corrected based on multiple local spectral features to obtain multiple corrected local curvature feature values, which are then used as the surface curvature distribution features. Since there is an inherent correlation between the spectral and geometric features of an object's surface—for example, smooth surfaces have more uniform spectral reflection and relatively stable local spectral features, while rough surfaces (with complex geometric features) have more chaotic spectral reflection and larger fluctuations in local spectral features—local spectral features can be used to correct previously calculated local curvature feature values, eliminating misjudgments of geometric features caused by point cloud noise, surface reflection, and other factors. For instance, a local mesh cell on an object's surface might have an overestimated local curvature feature value due to noise in the point cloud data caused by surface reflection. However, its corresponding local spectral features reveal uniform reflection in this area, consistent with the spectral characteristics of a smooth surface. Therefore, this local curvature feature value can be corrected downwards to more closely approximate the true geometric features. Through this correction process, the final corrected local curvature feature values ​​can more accurately and comprehensively reflect the curvature distribution of the target object's surface, forming a precise surface curvature distribution feature.

[0089] In one embodiment, step S3, which involves obtaining a sequence of local geometric feature vectors based on the surface curvature distribution characteristics and the reflection intensity information, includes:

[0090] S31. Extract the curvature extreme points of each region on the surface of the target object according to the surface curvature distribution characteristics, and perform region growth on multiple curvature extreme points to generate multiple initial geometric feature regions, wherein each initial geometric feature region corresponds to an average curvature value.

[0091] S32. Obtain multiple reflection intensities of the target object within a preset time period based on the reflection intensity information, obtain an average reflection intensity based on the multiple reflection intensities, and obtain a standard reflection intensity difference based on the average reflection intensity and the multiple reflection intensities.

[0092] S33. Obtain a reflection intensity feature value based on the difference between the average reflection intensity and the standard reflection intensity, and then perform weighted fusion with multiple average curvature values ​​according to the reflection intensity feature value to obtain multiple local feature fusion coefficients.

[0093] S34. The multiple local feature fusion coefficients are concatenated into an incremental vector to obtain a local geometric feature vector sequence.

[0094] As described in steps S31-S34 above, this invention extracts the curvature extrema points of each region on the surface of a target object based on the surface curvature distribution characteristics, and performs region growing on multiple curvature extrema points to generate multiple initial geometric feature regions, each corresponding to an average curvature value. The surface curvature distribution characteristics are a set of corrected local curvature feature values ​​obtained by fusing initial 3D point cloud data and multispectral reflectance image data, which can accurately reflect the curvature degree of each region on the object's surface. Curvature extrema points are points in the surface curvature distribution characteristics whose values ​​are significantly higher or lower than those of the surrounding areas. These points usually correspond to key geometric feature positions on the object's surface, such as the corners of parts, the inflection points of patterns on cultural relics, etc. Extracting these extrema points can quickly locate the core areas with complex geometric features. The region growing algorithm uses the curvature extrema points as seed points and gradually incorporates points with similar curvature features and spatial adjacency to form continuous initial geometric feature regions. This process can transform discrete extrema points into continuous regions with actual physical meaning, avoiding interference from isolated points in feature judgment. The average curvature value corresponding to each initial geometric feature region is the arithmetic mean of the modified local curvature feature values ​​of all points within that region. It is used to quantify the overall geometric complexity of the region; the higher the average curvature value, the more complex the geometric features of the region. For example, in the scanning of aero-engine blades, there are a large number of curvature extrema points around the leading edge, trailing edge, and film vents on the blade surface. After generating initial geometric feature regions through region growing, the average curvature values ​​of these regions are significantly higher than those of the smooth blade surface, clearly distinguishing between geometrically complex and simple regions.

[0095] Based on the reflection intensity information, multiple reflection intensities of the target object within a preset time period are obtained, and the average reflection intensity is obtained from these multiple reflection intensities. Then, the standard reflection intensity difference is obtained from the average reflection intensity and the multiple reflection intensities. The calculation formula is as follows:

[0096] ;

[0097] in, This represents the difference in standard reflection intensity. Indicates the intensity of the i-th reflection. This indicates the number of reflection intensities, where i represents the index of the reflection intensity, i = 1, 2, 3...n. Indicates the average reflected intensity;

[0098] Reflectance intensity information is acquired in real time by the signal receiving module of the 3D scanning equipment. It is the signal intensity data detected by the receiver after the detection signal (such as laser or structured light) emitted by the equipment is reflected from the object's surface. Its value is related to factors such as the object's surface material, roughness, and angle of incidence. The preset time is a reasonable time window set according to the scanning equipment's sampling frequency and the target object's scanning range, ensuring that a sufficient amount of reflection intensity data can be collected within this time to reflect the area's reflection characteristics. The average reflection intensity is the arithmetic mean of all reflection intensity data within the preset time, used to characterize the overall level of reflection intensity in that area. The standard reflection intensity difference is used to quantify the degree of fluctuation in reflection intensity; the larger the standard reflection intensity difference, the more unstable the reflection intensity in that area, and the more prone it is to data noise during the scanning process. For example, in scanning glass products, reflective areas on the glass surface cause drastic fluctuations in reflection intensity data, with a standard reflection intensity difference much larger than that of frosted areas. These two parameters can clearly distinguish between areas with stable and unstable reflection.

[0099] The characteristic value of reflection intensity is obtained based on the difference between the average reflection intensity and the standard reflection intensity, where the calculation formula is:

[0100] ;

[0101] in, Represents the characteristic value of reflection intensity. Indicates the average reflection intensity. This represents the difference in standard reflection intensity. This represents a smaller constant set according to actual conditions, where, A small positive number ϵ (smoothing factor) is introduced into the denominator to prevent the denominator from being zero or too small, and the reflection intensity characteristic value reflects the "average reflection intensity under unit reflection fluctuation". This ensures that even under extremely stable reflection conditions, the indicator has a reasonable upper limit, avoiding extreme situations. Multiple local feature fusion coefficients are obtained by sequentially weighting and fusing the reflection intensity feature value with multiple average curvature values. Next, these local feature fusion coefficients are concatenated as an incremental vector to obtain a sequence of local geometric feature vectors. Weighted fusion involves linearly weighting the average curvature value of each initial geometric feature region with its corresponding reflection intensity feature value: Local Feature Fusion Coefficient = α × Average Curvature Value + (1-α) × Reflection Intensity Feature Value, where α represents the weight of the average curvature value. This allows for the quantitative integration of multiple features through the local feature fusion coefficients. The subsequent incremental vector concatenation arranges the local feature fusion coefficients of all initial geometric feature regions in ascending order, forming an ordered vector sequence. The order of this sequence directly corresponds to the scanning priority of each region; the larger the coefficient, the higher the scanning priority of the corresponding region, requiring a more refined scanning strategy. For example, in scanning a mobile phone casing, the button recess area has a high average curvature (geometric complexity), and the material splicing in this area may cause large fluctuations in reflection intensity (large difference in standard reflection intensity). The corresponding local feature fusion coefficient is large, so it ranks higher in the vector sequence. Subsequent scans will allocate a higher sampling density and a slower scanning speed to this area to ensure scanning accuracy. On the other hand, the flat back area of ​​the casing has a low average curvature and stable reflection intensity, and a small local feature fusion coefficient. It ranks lower and will be scanned using a fast scanning strategy to improve efficiency.

[0102] Multiple local feature fusion coefficients are obtained by weighting and fusing the reflection intensity feature value with the average curvature value of each initial geometric feature region. Each coefficient corresponds to an initial geometric feature region, quantifying the comprehensive scanning priority of that region. The larger the coefficient, the more complex the region geometry or the more unstable the reflection, and the higher the scanning priority. The specific process of incremental vector concatenation is as follows: First, all local feature fusion coefficients are sorted by numerical value in ascending order. Then, the sorted coefficients are expanded sequentially by vector dimension to form a one-dimensional ordered vector, i.e., a sequence of local geometric feature vectors. This ensures the accuracy of the coefficient order. Vector concatenation is achieved by expanding the array or matrix. Each coefficient is an element of the vector, and its index position directly reflects the priority. The larger the index, the larger the corresponding coefficient value, and the higher the scanning priority.

[0103] In one embodiment, step S4, which involves obtaining an initial scan path based on the initial 3D point cloud data and marking the initial scan path based on the local geometric feature vector sequence to obtain a marked scan path, includes:

[0104] S41. Obtain the corresponding spatial bounding box and normal vector field based on the initial three-dimensional point cloud data, and generate an initial scanning path covering all visible surfaces of the target object based on the spatial bounding box and the normal vector field using the spatial filling curve algorithm. The initial scanning path includes multiple path points.

[0105] S42. Map the local geometric feature vector sequence to each path point of the initial scan path, and assign a feature label to each path point, wherein the feature label contains a sequence priority value;

[0106] S43. Sequentially determine the relationship between multiple sequence priority values ​​and preset sequence priority value intervals, and assign corresponding scan action identifiers to multiple sequence priority values, wherein the scan action identifiers include fast scan identifiers, standard scan identifiers and fine scan identifiers;

[0107] When the sequence priority value is less than the minimum value of the preset sequence priority value range, the path point corresponding to the sequence priority value is mapped to a fast scan identifier.

[0108] When the sequence priority value is within the preset sequence priority value range, the path point corresponding to the sequence priority value is mapped to a standard scan identifier.

[0109] When the sequence priority value is greater than the maximum value of the preset sequence priority value range, the path point corresponding to the sequence priority value is mapped to a fine scan identifier.

[0110] S44. Perform interval merging optimization on the path points assigned to the scan action identifier in the multiple sequence priority values, and merge adjacent path points continuously assigned the same scan identifier into the same scan path segment.

[0111] S45. Generate a marked scan path based on the multiple scan path segments and their corresponding scan action identifiers.

[0112] As described in steps S41-S45 above, this invention obtains the corresponding spatial bounding box and normal vector field based on the initial 3D point cloud data, and generates an initial scanning path covering the entire visible surface of the target object based on the spatial filling curve algorithm. The initial scanning path includes multiple path points. The initial 3D point cloud data is pre-scanned and obtained by a 3D scanning device, containing discrete 3D coordinate information of the object surface. The spatial bounding box is the smallest cubic structure that can completely enclose the target object, and its size is determined by the maximum and minimum values ​​of the coordinates in the initial 3D point cloud data. It is used to define the spatial range of the scan and avoid scanning beyond the object boundary, thus preventing invalid work. The normal vector field is a set composed of the normal vectors of each point in the initial 3D point cloud data. The normal vectors reflect the orientation of each point on the object surface, providing a basis for the direction planning of the scanning path, ensuring that the detection direction of the scanning device is reasonably matched with the orientation of the object surface, and improving the data acquisition quality. The space-filling curve algorithm is a type of algorithm that can generate continuous, non-overlapping curves (such as Hilbert curves and Runge-Kutta curves) that cover the entire area within a limited space. Applying this algorithm to the spatial bounding box and normal vector field generates a continuous initial scan path that ensures the path covers all visible surfaces of the object. Furthermore, the multiple path points on the path are distributed sequentially, providing a basis for subsequent point-by-point marking. For example, in industrial part scanning, the spatial bounding box of the part is determined using initial 3D point cloud data. Combined with the normal vector field of the part's surface, the initial scan path generated using the Hilbert curve algorithm can systematically cover all surfaces of the part, including visible areas of complex structures such as grooves and holes, avoiding omissions in the scan range.

[0113] The local geometric feature vector sequence is mapped to each path point of the initial scan path, and a feature label containing a sequence priority value is assigned to each path point. The local geometric feature vector sequence is an ordered vector generated by fusing surface curvature distribution features and reflection intensity information. Each vector element corresponds to a local feature fusion coefficient, which comprehensively reflects the geometric complexity and reflection stability of the region. The mapping process is achieved through spatial coordinate matching. The fusion coefficient corresponding to a certain region of the object in the local geometric feature vector sequence is assigned to path points in the initial scan path that pass through that region, so that each path point can be associated with the corresponding comprehensive feature quantization value. The sequence priority value in the feature label is directly equivalent to the local feature fusion coefficient corresponding to that path point. The larger the fusion coefficient, the higher the sequence priority value, indicating a higher scanning priority for the region where the path point is located, requiring more refined scanning actions. For example, in the scanning of cultural relics, the local feature fusion coefficient corresponding to densely decorated areas is large, and the sequence priority value mapped to the path points in that area is also high, providing a direct basis for subsequent allocation of refined scan identifiers.

[0114] The system sequentially determines the relationship between multiple sequence priority values ​​and a preset sequence priority value range, and assigns corresponding scan action identifiers to these sequence priority values. These scan action identifiers include fast scan identifiers, standard scan identifiers, and fine scan identifiers. The preset sequence priority value range is a numerical range pre-defined based on the accuracy requirements of the scanning task, equipment performance, and application scenario. For example, in a typical industrial inspection scenario, the preset sequence priority value range can be set to [0.3, 0.7]. The minimum and maximum values ​​of the range can be dynamically adjusted according to actual needs. The determination logic is as follows: when the sequence priority value is less than the minimum value of the preset sequence priority value range, it indicates that the region where the path point is located has simple geometry and stable reflection, resulting in a low scan priority, mapped to a fast scan identifier, corresponding to a higher scan speed and lower sampling density; when the sequence priority value is within the preset sequence priority value range, the region has moderate geometric complexity and reflection stability, mapped to a standard scan identifier, corresponding to a balanced scan speed and sampling density; when the sequence priority value is greater than the maximum value of the preset sequence priority value range, the region has complex geometry or unstable reflection, resulting in a high scan priority, mapped to a fine scan identifier, corresponding to a lower scan speed and higher sampling density. This classification method can transform abstract sequence priority values ​​into specific scanning action instructions that the device can recognize, ensuring that different priority areas receive matching scanning processing. For example, in the scanning of building components, the path point sequence priority value of the flat outer surface of the component is low and is marked as a fast scan identifier, while the path point sequence priority value of the carved and decorative area of ​​the component is high and is marked as a fine scan identifier, enabling the scanning process to be adjusted accordingly.

[0115] Interval merging optimization is performed on path points assigned scan action identifiers from multiple sequence priority values, merging adjacent path points with the same scan identifier into a single scan path segment. Since path points on the initial scan path are continuously distributed, adjacent path points often belong to the same region, and their scan action identifiers may be identical. Performing scan actions point by point would cause the device to frequently switch scan modes, reducing scanning efficiency. Interval merging optimization iterates through the scan action identifiers of all path points, merging consecutive path points with the same identifier into a single scan path segment. Each scan path segment corresponds to a single scan action identifier, reducing the number of action switching during the scan process. For example, if 100 consecutive path points on the initial scan path are all marked with a fast scan identifier, after interval merging optimization, these path points are merged into a fast scan path segment. The device only needs to switch to fast scan mode when entering this path segment and continue switching modes until the end of the path segment, significantly improving scanning efficiency.

[0116] A marked scanning path is generated based on multiple scanning path segments and their corresponding scanning action identifiers. After interval merging optimization, the initial scanning path, originally composed of a large number of discrete path points, is transformed into a structured path composed of multiple continuous scanning path segments. Each path segment carries a clear scanning action identifier, forming a complete marked scanning path. This path ensures coverage of all visible surfaces of the target object and clarifies the scanning mode of each path segment through the scanning action identifiers, providing a clear and efficient path basis for subsequent differentiated scanning in conjunction with the scanning strategy matrix. For example, in scanning the casing of an electronic device, the marked scanning path includes a fast scanning path segment (corresponding to the flat back of the casing), a standard scanning path segment (corresponding to the sides of the casing), and a fine scanning path segment (corresponding to the button hole area of ​​the casing). When the device scans along this path, it can automatically adjust the scanning speed and sampling density according to the identifiers of each path segment, achieving a balance between efficiency and accuracy.

[0117] In one embodiment, step S5, which involves constructing a scanning strategy matrix based on the local geometric feature vector sequence, the geometric positioning accuracy parameter, and the maximum scanning speed, includes:

[0118] S51. Obtain the minimum feasible point distance based on the geometric positioning accuracy parameters;

[0119] S52. Obtain multiple sampling densities based on the local geometric feature vector sequence and the minimum feasible point distance;

[0120] S53. Obtain the total task pre-constraint time, and construct a scanning strategy matrix based on the total task pre-constraint time, multiple sampling densities, and the maximum scanning speed.

[0121] As described in steps S51-S53 above, this invention obtains the minimum feasible point distance based on the geometric positioning accuracy parameter. The geometric positioning accuracy parameter is a core component of the scanning state data, fed back in real-time by the 3D scanning equipment during operation, reflecting the spatial positioning error range of the scanning points under the current working state (e.g., the positioning accuracy parameter of a laser scanning equipment may be ±0.01mm). The minimum feasible point distance refers to the minimum distance between two adjacent sampling points that ensures data validity under the current geometric positioning accuracy. Its calculation must be based on the geometric positioning accuracy parameter—if the sampling point distance is less than the error range corresponding to the geometric positioning accuracy, the spatial positions of adjacent sampling points will overlap or become confused due to positioning errors, resulting in data redundancy and inability to effectively distinguish them; if the sampling point distance is too large, key geometric features will be missed. Typically, 1.5-2 times the geometric positioning accuracy parameter is used as the minimum feasible point distance (the specific multiple can be adjusted according to the accuracy requirements of the application scenario). For example, when the geometric positioning accuracy parameter is ±0.01mm, the minimum feasible point distance can be set to 0.015mm, which ensures both the spatial distinguishability of the sampling points and fully utilizes the positioning accuracy potential of the equipment. This sets feasible boundaries for sampling density at the device performance level, ensuring that subsequent sampling density settings do not exceed the device's positioning capabilities, thus providing a basic guarantee for data validity.

[0122] Multiple sampling densities are obtained based on the local geometric feature vector sequence and the minimum feasible point distance. The local geometric feature vector sequence is a vector sequence concatenated from multiple local feature fusion coefficients in ascending order. Each fusion coefficient corresponds to the scanning priority of a region; the larger the coefficient, the higher the scanning priority. Sampling density refers to the number of sampling points per unit length or unit area, and is inversely proportional to the point distance (the smaller the point distance, the higher the sampling density). In this step, the sampling density matching follows the principle of "priority is positively correlated with density, and density is not less than the maximum density corresponding to the minimum feasible point distance": First, each local feature fusion coefficient in the local geometric feature vector sequence is divided into different intervals (such as high, medium, and low priority intervals). Then, a corresponding sampling density is assigned to each interval. The sampling density of the high priority interval is the highest, and the corresponding point distance is equal to the minimum feasible point distance (fully utilizing the positioning accuracy of the device to ensure detail capture); the sampling density of the medium priority interval is the second highest, and the point distance is 1.5-2 times the minimum feasible point distance; the sampling density of the low priority interval is the lowest, and the point distance is 2-3 times the minimum feasible point distance (the specific multiple can be flexibly adjusted according to the total task pre-constraint time). In this way, the region scanning priority is transformed into specific sampling density parameters, realizing the accurate adaptation of sampling density to region geometric complexity and reflection stability, laying the foundation for subsequent speed matching.

[0123] The total task pre-constraint time is obtained, and a scanning strategy matrix is ​​constructed based on the total task pre-constraint time, multiple sampling densities, and the maximum scanning speed. The total task pre-constraint time is set by the user according to the actual application scenario (e.g., in industrial batch inspection, the scanning time for a single item may be set to no more than 5 minutes, while for the digital scanning of cultural relics, a longer time may be allowed to ensure accuracy). The maximum scanning speed is a core performance parameter of the scanning equipment, determined by the equipment's hardware performance (e.g., the maximum scanning speed of a laser scanning equipment may be 1000 points / second), reflecting the upper limit of the number of sampling points that the equipment can complete per unit time. The construction process of the scanning strategy matrix is ​​a multi-parameter optimization and matching process: First, based on the sampling density of each region and the scanning path length of that region (obtained from the length of the scanning path segment in the marked scanning path), the theoretical scanning time required for that region is calculated (theoretical scanning time = scanning path length ÷ point spacing × scanning time of a single sampling point, scanning time of a single sampling point = 1 ÷ maximum scanning speed); then, the theoretical scanning times of all regions are summed and compared with the total task pre-constraint time. If the total theoretical scanning time is less than or equal to the total task pre-constraint time, the sampling density of each region is directly summed with the maximum scanning speed (or based on region priority). The scanning strategy involves matching the sampling density of high-priority areas with appropriately reduced speeds to form a pairing strategy. If the total theoretical scanning time exceeds the total task pre-constraint time, the point spacing of medium and low-priority areas is appropriately increased (sampling density is reduced) while maintaining the sampling density of high-priority areas. Alternatively, the scanning speed of medium and low-priority areas is increased within the limits of equipment performance (not exceeding the maximum scanning speed), and the theoretical scanning time is recalculated until the total time meets the pre-constraint requirements. Finally, the target scanning speed (determined based on the above optimization matching results) and target point density (i.e., the point spacing parameter corresponding to the sampling density) for each scanning path segment are organized into a matrix to form a scanning strategy matrix. The rows of the matrix correspond to each scanning path segment in the scanning path, and the columns correspond to the target scanning speed and target point density. The value in each cell is the specific execution parameter for that path segment. This quantitatively matches the regionally differentiated sampling density with the scanning speed and performs global optimization through the total task pre-constraint time. This ensures that the scanning strategy meets the time requirements and equipment performance while maximizing the adaptation to the accuracy requirements of different regions, providing clear and feasible parameter basis for subsequent scanning execution.

[0124] In one embodiment, step S6, which involves performing an efficiency- and accuracy-balanced scan of the marked scanning path based on the scanning strategy matrix and outputting the scan results, includes:

[0125] S61. Obtain multiple spatial coordinates corresponding to the marked scanning path, and pair the multiple spatial coordinates based on the scanning strategy matrix to obtain multiple pairing strategies, wherein the pairing strategy includes target scanning speed and target point density.

[0126] S62. Start the preset three-dimensional scanning device to move along the marked scanning path, and perform an efficiency and accuracy balance scan on the marked scanning path according to the multiple target scanning speeds and multiple target point densities to obtain the initial scanning result. The initial scanning result includes real-time acquisition of multiple point cloud data subsets acquired in the current scanning path segment, and acquisition of the point cloud density uniformity and point cloud fitting residual of each point cloud data subset.

[0127] S63. Weighted calculation is performed on the density uniformity of multiple point clouds and the fitting residuals of multiple point clouds to obtain multiple real-time quality indicators.

[0128] S64. The multiple real-time quality indicators are compared with preset quality thresholds in sequence;

[0129] If the real-time quality index is greater than or equal to the preset quality threshold, then the preset 3D scanning device is controlled to continue scanning the next path segment until the main body path scanning is completed, and the main body scanning path is obtained.

[0130] If the real-time quality index is less than the preset quality threshold, the position of the current scanning path segment and the scanning strategy matrix corresponding to the real-time quality index are obtained, and a local compensation scanning sub-path is dynamically generated according to the position of the current scanning path segment and the scanning strategy matrix.

[0131] S65. The main scanning path and the multiple local compensation scanning sub-paths are spliced ​​together to obtain the scanning path result.

[0132] As described in steps S61-S65 above, this invention obtains multiple spatial coordinates corresponding to the marked scanning path, and pairs these spatial coordinates based on the scanning strategy matrix to obtain multiple pairing strategies. Each pairing strategy includes a target scanning speed and a target point density. The marked scanning path contains multiple scanning path segments with scanning action identifiers. Its spatial coordinates are the three-dimensional position information of each path point, used to guide the movement trajectory of the scanning device. The scanning strategy matrix contains the optimal target scanning speed and target point density parameters corresponding to each scanning path segment. The pairing process is achieved through the correspondence between spatial coordinates and scanning path segments: first, the multiple spatial coordinates of the marked scanning path are grouped according to the scanning path segments, with each group corresponding to a continuous scanning path segment; then, the target scanning speed and target point density corresponding to that scanning path segment are extracted from the scanning strategy matrix and associated with that group of spatial coordinates to form a pairing strategy. This transforms the abstract strategy matrix parameters into executable instructions bound to specific path coordinates, providing the scanning device with a clear combination of movement trajectory and scanning parameters, ensuring the accurate implementation of differentiated strategies.

[0133] The preset 3D scanning device is activated and moves along the marked scanning path. It performs a balanced scan of the marked scanning path, considering both efficiency and accuracy, based on multiple target scanning speeds and point densities, to obtain initial scan results. These initial scan results include real-time acquisition of multiple subsets of point cloud data for the current scanning path segment, along with the point cloud density uniformity and point cloud fitting residuals for each subset. The preset 3D scanning device can be a laser scanning device, structured light scanning device, etc. As it moves along the marked scanning path, it adjusts the scanning speed and sampling point density in real time according to a pairing strategy: in fine scanning path segments, the device reduces its moving speed and increases the sampling frequency to ensure target point density; in fast scanning path segments, the device increases its moving speed and reduces the sampling frequency to improve efficiency. A point cloud data subset is a local point cloud data collected in real time by the device during the scanning process at fixed time intervals (e.g., 0.1 seconds) or fixed path lengths (e.g., 1 mm). It is used to dynamically evaluate the quality of the current scanning area. Point cloud density uniformity refers to the degree of uniformity of the spatial distribution of sampling points within the point cloud data subset. It is obtained by calculating the ratio of the standard deviation of the distance between adjacent sampling points within the subset to the average value. The smaller the ratio, the better the density uniformity, reflecting a more regular distribution of sampling points. Point cloud fitting residual refers to the root mean square value of the distance from each sampling point to the fitted surface after fitting the point cloud data subset to an ideal geometric surface (e.g., plane, curved surface). The smaller the residual, the smaller the deviation between the point cloud data and the real surface of the object, and the higher the data accuracy. In this way, by having the device perform scanning according to a preset strategy, while simultaneously collecting local data in real time and extracting quality assessment parameters, data support is provided for subsequent quality judgment, realizing the synchronous execution of scanning and quality monitoring.

[0134] Multiple point cloud density uniformity scores and multiple point cloud fitting residuals are weighted and calculated to obtain multiple real-time quality indicators. The specific calculation method is: Real-time Quality Indicator = (Point Cloud Density Uniformity Score × Uniformity Weight) + (Point Cloud Fitting Residual Score × Residual Weight). Here, the point cloud density uniformity score is a quantized value of 0-100 converted from the uniformity ratio (the smaller the ratio, the higher the score), and the point cloud fitting residual score is a quantized value of 0-100 converted from the ratio of the residual value to a preset allowable residual (the smaller the ratio, the higher the score). This integrates the two dimensions of quality assessment parameters into a unified quantitative indicator, simplifying the quality judgment logic and making subsequent threshold comparisons more intuitive and efficient.

[0135] Multiple real-time quality indicators are compared sequentially with preset quality thresholds, and the operating status of the scanning device is controlled based on the comparison results. The preset quality threshold is a quantified value (e.g., 80 points) set according to the accuracy requirements of the scanning task, determined by the user in conjunction with the application scenario: the threshold for high-precision scenarios can be set to 85 points or higher, and the threshold for ordinary precision scenarios can be set to 70 points or higher. When the real-time quality indicator is greater than or equal to the preset quality threshold, it indicates that the point cloud data quality of the current scanning path segment meets the requirements, and no adjustment of the scanning strategy is needed. The device continues scanning the next path segment until all path segments are scanned, resulting in the main scanning path, which is a complete path composed of all scanning path segments that meet the quality standards. When the real-time quality indicator is less than the preset quality threshold, it indicates that the point cloud data of the current scanning path segment has uneven density or insufficient accuracy, possibly due to environmental interference (e.g., reflection), slight device drift, or insufficient matching between the strategy parameters and the actual scene. In this case, it is necessary to obtain the specific location information (e.g., 3D coordinate range) and the corresponding scanning strategy matrix parameters of the path segment, and dynamically generate a locally compensated scanning sub-path based on this information. The logic for generating local compensation scan sub-paths is as follows: based on the spatial range of the current substandard path segment, the scan coverage is reduced (only covering the substandard area to avoid invalid scans), while the scan strategy parameters are optimized, increasing the target point density by 20%-30% (further increasing the sampling density) and reducing the target scan speed by 10%-20% (improving sampling stability), ensuring the quality of the supplementary scan data. In this way, the scanning process is dynamically controlled through real-time quality feedback, efficiently advancing the scan in the compliant area and accurately initiating supplementary scans in the substandard area, ensuring both overall scanning efficiency and without sacrificing local data quality.

[0136] The main scanning path and multiple local compensation scanning sub-paths are stitched together to obtain the scanning path result. The stitching process is achieved through coordinate alignment: the starting coordinates of the local compensation scanning sub-paths coincide with the starting coordinates of the substandard areas in the main scanning path, and the ending coordinates coincide with the ending coordinates of the substandard areas, ensuring seamless integration of the supplementary scanning data with the original data and avoiding data overlap or omissions. The final point cloud data corresponding to the scanning path result is a fusion of the high-quality point cloud data from the main scanning path and the supplementary scanning data from the local compensation scanning sub-paths. It includes complete scanning information for all areas and corrects local quality defects through supplementary scanning. For example, in the digitization scanning of cultural relics, the main scanning path covers most of the relics and acquires high-quality data, while the reflective areas at the edges of the relics are supplemented through local compensation scanning sub-paths, resulting in point cloud data with acceptable accuracy. After stitching, a complete and high-precision 3D point cloud data of the cultural relic is formed. This integration of the main scanning and supplementary scanning results forms a complete and high-quality scanning result, achieving a final balance between efficiency and accuracy.

[0137] This application also provides a 3D scanning efficiency and accuracy balancing system based on an adaptive sampling strategy, including:

[0138] The first acquisition module is used to acquire the three-dimensional scan data of the target object, wherein the three-dimensional scan data includes the prior data of the target object and the scan state data;

[0139] The second acquisition module is used to acquire the initial three-dimensional point cloud data, multispectral reflectance image data and reflectance intensity information of the target object based on the prior data of the target object, and to acquire the surface curvature distribution characteristics of the target object based on the initial three-dimensional point cloud data and the multispectral reflectance image data.

[0140] The third acquisition module is used to acquire a sequence of local geometric feature vectors based on the surface curvature distribution characteristics and the reflection intensity information;

[0141] The fourth acquisition module is used to acquire an initial scanning path based on the initial 3D point cloud data, and to mark the initial scanning path based on the local geometric feature vector sequence to obtain a marked scanning path;

[0142] The fifth acquisition module is used to acquire geometric positioning accuracy parameters and maximum scanning speed based on the scanning state data, and to construct a scanning strategy matrix based on the local geometric feature vector sequence, the geometric positioning accuracy parameters, and the maximum scanning speed;

[0143] The scanning module is used to perform a balanced scanning of the marked scanning path based on the scanning strategy matrix, and output the scanning results.

[0144] In one embodiment, the second acquisition module includes:

[0145] The first acquisition unit is used to acquire the spatial grid corresponding to the target object based on the initial three-dimensional point cloud data, and to divide the spatial grid into non-overlapping parts based on a preset local region to obtain multiple local three-dimensional grid units.

[0146] The second acquisition unit is used to acquire geometric feature values ​​within each local three-dimensional mesh cell based on the plurality of local three-dimensional mesh cells, wherein the geometric feature values ​​include local principal curvature and normal vector change rate;

[0147] The third acquisition unit is used to acquire multiple local curvature feature values ​​based on multiple local principal curvatures and multiple normal vector change rates;

[0148] The registration unit is used to register the initial three-dimensional point cloud data and the multispectral reflectance image data to obtain three-dimensional point cloud-multispectral reflectance image reference data, and to assign corresponding local spectral features to each local three-dimensional mesh unit according to the three-dimensional point cloud-multispectral reflectance image reference data.

[0149] The correction unit is used to correct the multiple local curvature feature values ​​according to the multiple local spectral features, to obtain multiple corrected local curvature feature values, and to use the multiple corrected local curvature feature values ​​as surface curvature distribution features.

[0150] This application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0151] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0152] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in this application and in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0153] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0154] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for balancing efficiency and accuracy in three-dimensional scanning based on an adaptive sampling strategy, characterized in that, include: Acquire 3D scan data of the target object, which includes prior data of the target object and scan state data; The initial three-dimensional point cloud data, multispectral reflectance image data, and reflectance intensity information of the target object are obtained based on the prior data of the target object. The surface curvature distribution characteristics of the target object are then obtained based on the initial three-dimensional point cloud data and the multispectral reflectance image data. The step of obtaining the surface curvature distribution characteristics of the target object based on the initial three-dimensional point cloud data and the multispectral reflectance image data includes: The spatial mesh corresponding to the target object is obtained based on the initial three-dimensional point cloud data, and the spatial mesh is divided into non-overlapping parts based on a preset local region to obtain multiple local three-dimensional mesh units. The geometric feature values ​​within each local three-dimensional mesh cell are obtained based on the plurality of local three-dimensional mesh cells, wherein the geometric feature values ​​include the local principal curvature and the rate of change of the normal vector; Multiple local curvature feature values ​​are obtained based on multiple local principal curvatures and multiple normal vector change rates; The initial 3D point cloud data and the multispectral reflectance image data are registered to obtain 3D point cloud-multispectral reflectance image reference data, and each local 3D mesh cell is assigned a corresponding local spectral feature based on the 3D point cloud-multispectral reflectance image reference data. The local curvature feature values ​​are corrected based on the local spectral features to obtain the corrected local curvature feature values, and the corrected local curvature feature values ​​are used as surface curvature distribution features. A sequence of local geometric feature vectors is obtained based on the surface curvature distribution characteristics and the reflection intensity information; An initial scanning path is obtained based on the initial 3D point cloud data, and the initial scanning path is marked based on the local geometric feature vector sequence to obtain a marked scanning path; The geometric positioning accuracy parameters and maximum scanning speed are obtained based on the scanning state data, and a scanning strategy matrix is ​​constructed based on the local geometric feature vector sequence, the geometric positioning accuracy parameters, and the maximum scanning speed. The marked scanning path is scanned in a balance between efficiency and accuracy according to the scanning strategy matrix, and the scanning results are output.

2. The method for balancing efficiency and accuracy in three-dimensional scanning based on an adaptive sampling strategy according to claim 1, characterized in that, The step of obtaining a local geometric feature vector sequence based on the surface curvature distribution characteristics and the reflection intensity information includes: Based on the surface curvature distribution characteristics, the curvature extreme points of each region on the surface of the target object are extracted, and multiple curvature extreme points are subjected to region growing to generate multiple initial geometric feature regions, wherein each initial geometric feature region corresponds to an average curvature value. Based on the reflection intensity information, multiple reflection intensities of the target object within a preset time period are obtained, and an average reflection intensity is obtained based on the multiple reflection intensities. A standard reflection intensity difference is obtained based on the average reflection intensity and the multiple reflection intensities. The reflection intensity feature value is obtained based on the difference between the average reflection intensity and the standard reflection intensity, and the reflection intensity feature value is sequentially weighted and fused with multiple average curvature values ​​to obtain multiple local feature fusion coefficients. The local feature fusion coefficients are concatenated into an incremental vector to obtain a sequence of local geometric feature vectors.

3. The method for balancing efficiency and accuracy in three-dimensional scanning based on an adaptive sampling strategy according to claim 1, characterized in that, The step of obtaining an initial scan path based on the initial 3D point cloud data and marking the initial scan path based on the local geometric feature vector sequence to obtain a marked scan path includes: Based on the initial 3D point cloud data, the corresponding spatial bounding box and normal vector field are obtained, and an initial scanning path covering all visible surfaces of the target object is generated by the spatial bounding box and the normal vector field based on the spatial filling curve algorithm. The initial scanning path includes multiple path points. The local geometric feature vector sequence is mapped to each path point of the initial scan path, and a feature label is assigned to each path point, wherein the feature label contains a sequence priority value; The relationship between multiple sequence priority values ​​and a preset sequence priority value range is determined sequentially, and corresponding scan action identifiers are assigned to the multiple sequence priority values. The scan action identifiers include fast scan identifiers, standard scan identifiers, and fine scan identifiers. When the sequence priority value is less than the minimum value of the preset sequence priority value range, the path point corresponding to the sequence priority value is mapped to a fast scan identifier. When the sequence priority value is within the preset sequence priority value range, the path point corresponding to the sequence priority value is mapped to a standard scan identifier. When the sequence priority value is greater than the maximum value of the preset sequence priority value range, the path point corresponding to the sequence priority value is mapped to a fine scan identifier. The path points assigned to the scan action identifier in the multiple sequence priority values ​​are optimized by interval merging, and adjacent path points with the same scan identifier are merged into the same scan path segment. A marked scan path is generated based on multiple scan path segments and their corresponding scan action identifiers.

4. The method for balancing efficiency and accuracy in three-dimensional scanning based on an adaptive sampling strategy according to claim 1, characterized in that, The step of constructing a scanning strategy matrix based on the local geometric feature vector sequence, the geometric positioning accuracy parameter, and the maximum scanning speed includes: The minimum feasible point distance is obtained based on the geometric positioning accuracy parameters. Multiple sampling densities are obtained based on the local geometric feature vector sequence and the minimum feasible point distance; Obtain the total task pre-constraint time, and construct a scanning strategy matrix based on the total task pre-constraint time, multiple sampling densities, and the maximum scanning speed.

5. The method for balancing efficiency and accuracy in three-dimensional scanning based on an adaptive sampling strategy according to claim 1, characterized in that, The step of performing an efficiency- and accuracy-balanced scan of the marked scan path according to the scan strategy matrix and outputting the scan results includes: Multiple spatial coordinates are obtained according to the marked scanning path, and the multiple spatial coordinates are paired according to the scanning strategy matrix to obtain multiple pairing strategies, wherein the pairing strategy includes target scanning speed and target point density; The preset 3D scanning device is started and moves along the marked scanning path. The marked scanning path is scanned in a balance between efficiency and accuracy according to the multiple target scanning speeds and multiple target point densities to obtain the initial scanning result. The initial scanning result includes real-time acquisition of multiple point cloud data subsets acquired in the current scanning path segment, and acquisition of the point cloud density uniformity and point cloud fitting residual of each point cloud data subset. Multiple real-time quality indicators are obtained by weighting the density uniformity of multiple points cloud and the fitting residuals of multiple points cloud; The real-time quality indicators are compared sequentially with preset quality thresholds; If the real-time quality index is greater than or equal to the preset quality threshold, then the preset 3D scanning device is controlled to continue scanning the next path segment until the main body path scanning is completed, and the main body scanning path is obtained. If the real-time quality index is less than the preset quality threshold, the position of the current scanning path segment and the scanning strategy matrix corresponding to the real-time quality index are obtained, and a local compensation scanning sub-path is dynamically generated according to the position of the current scanning path segment and the scanning strategy matrix. The main scanning path and multiple local compensation scanning sub-paths are concatenated to obtain the scanning path result.

6. A three-dimensional scanning efficiency and accuracy balancing system based on an adaptive sampling strategy, characterized in that, include: The first acquisition module is used to acquire the three-dimensional scan data of the target object, wherein the three-dimensional scan data includes the prior data of the target object and the scan state data; The second acquisition module is used to acquire initial three-dimensional point cloud data, multispectral reflectance image data, and reflectance intensity information of the target object based on the prior data of the target object, and to acquire the surface curvature distribution characteristics of the target object based on the initial three-dimensional point cloud data and the multispectral reflectance image data. The step of acquiring the surface curvature distribution characteristics of the target object based on the initial three-dimensional point cloud data and the multispectral reflectance image data includes: The spatial mesh corresponding to the target object is obtained based on the initial three-dimensional point cloud data, and the spatial mesh is divided into non-overlapping parts based on a preset local region to obtain multiple local three-dimensional mesh units. The geometric feature values ​​within each local three-dimensional mesh cell are obtained based on the plurality of local three-dimensional mesh cells, wherein the geometric feature values ​​include the local principal curvature and the rate of change of the normal vector; Multiple local curvature feature values ​​are obtained based on multiple local principal curvatures and multiple normal vector change rates; The initial 3D point cloud data and the multispectral reflectance image data are registered to obtain 3D point cloud-multispectral reflectance image reference data, and each local 3D mesh cell is assigned a corresponding local spectral feature based on the 3D point cloud-multispectral reflectance image reference data. The local curvature feature values ​​are corrected based on the local spectral features to obtain the corrected local curvature feature values, and the corrected local curvature feature values ​​are used as surface curvature distribution features. The third acquisition module is used to acquire a sequence of local geometric feature vectors based on the surface curvature distribution characteristics and the reflection intensity information; The fourth acquisition module is used to acquire an initial scanning path based on the initial 3D point cloud data, and to mark the initial scanning path based on the local geometric feature vector sequence to obtain a marked scanning path; The fifth acquisition module is used to acquire geometric positioning accuracy parameters and maximum scanning speed based on the scanning state data, and to construct a scanning strategy matrix based on the local geometric feature vector sequence, the geometric positioning accuracy parameters, and the maximum scanning speed; The scanning module is used to perform a balanced scanning of the marked scanning path based on the scanning strategy matrix, and output the scanning results.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images