A two-level optimization measurement planning method for complex model measurement tasks

Abstract

This invention relates to the field of 3D point cloud measurement and intelligent path planning technology, solving the technical problems of strong occlusion, large normal changes, limited accessibility, and difficult route execution in traditional point cloud measurement methods. Specifically, it relates to a two-layer optimization measurement planning method for complex model measurement tasks. Based on an adjacency graph of candidate viewpoints and a visibility evaluation mechanism, it calculates and caches the visible point set of each viewpoint under angle constraints and occlusion culling conditions. The inner layer optimization generates an initial measurement route and viewpoint set starting from the initial viewpoint, and reconstructs a continuously executable measurement route after satisfying the full coverage constraint. The outer layer optimization uses a genetic algorithm to perform a global search on the initial viewpoint, outputting the optimal measurement viewpoint set and measurement path with coverage and viewpoint size costs as fitness. This invention can achieve high coverage measurement planning under complex occlusion and strict viewpoint constraints, reducing the number of viewpoints and path redundancy, and improving planning efficiency and robustness.

Description

Technical Field

[0001] This invention relates to the field of 3D point cloud measurement and intelligent path planning technology, and in particular to a two-layer optimization measurement planning method for complex model measurement tasks. Background Technology

[0002] With the increasing demands for geometric accuracy and quality assessment of complex curved surface components in fields such as aerospace, precision manufacturing, and mold inspection, non-contact measurement based on point cloud data has become an important technical means for surface reconstruction, error analysis, and quality control. Compared with traditional contact measurement, this type of method can acquire large-scale spatial data in a shorter time, and has advantages such as high efficiency, large information capacity, and strong adaptability to workpiece surfaces. It can provide a reliable data foundation and analytical basis for the precision inspection and quality traceability of complex structural components.

[0003] In practical measurement processes, measurement planning typically involves selecting suitable measurement viewpoints from a limited set of candidate viewpoints and planning the movement routes of measurement equipment (such as scanning heads, robotic arms, or mobile platforms) to achieve comprehensive observation of the target area. However, complex models often exhibit significant occlusion, drastic changes in surface normals, and structural features such as local depressions, thin walls, and slits. These features result in strong spatial inhomogeneities in the visibility of the target point cloud from the viewpoint, significantly increasing the difficulty of visibility assessment and coverage completion. Simultaneously, the measurement process is also constrained by engineering constraints such as viewing angle range, incident angle (or normal angle), equipment mobility, path connectivity, and movement safety. This necessitates that the planning results not only achieve high coverage but also meet requirements such as route continuity, feasibility, and safe obstacle avoidance to be applicable in real-world scenarios.

[0004] Existing technical solutions typically employ local greedy algorithms or single-layer global search strategies. Local greedy methods generally select viewpoints step-by-step based on coverage gain, offering advantages such as simplicity and high computational efficiency. However, because decisions rely on local information, they are prone to creating coverage gaps in complex occlusion areas. Furthermore, subsequent coverage completion may introduce excessive redundant viewpoints, significantly increasing measurement time and path length. While single-layer global search (such as genetic algorithms and particle swarm optimization) possesses a certain degree of globality, without a coordinated mechanism with coverage construction, connectivity constraints, and occlusion detection, it often exhibits problems such as a large search space, slow convergence speed, low feasible solution ratio, and the output viewpoint sequence may suffer from strong jumps, disconnected routes, or excessive costs, thus affecting the stability and reproducibility of the planning scheme.

[0005] Furthermore, near complex structures, direct paths between candidate viewpoints may encounter risks such as clipping, collisions, or boundary violations. Without effective path clipping detection and detour mechanisms, even if the viewpoint set meets coverage requirements, the generated route may still be unsafe to execute, leading to extensive manual adjustments during engineering deployment and reducing the practical value and efficiency of automated measurement planning. Therefore, there is an urgent need for a measurement planning method that balances coverage, viewpoint scale, and route feasibility under realistic visibility constraints. This method should improve the robustness, efficiency, and engineering usability of measurement planning for complex models through collaborative optimization of global search and local construction, and by introducing redundancy suppression and safe path generation mechanisms. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a two-layer optimized measurement planning method for complex model measurement tasks. It solves the technical problems of strong occlusion, large normal changes, limited accessibility, and difficult route execution in point cloud measurement using traditional methods. It is particularly suitable for complex model measurement with significant occlusion, drastic changes in surface normals, and structural features such as local depressions, thin walls, and slits.

[0007] To address the aforementioned technical problems, this invention provides a two-layer optimization measurement planning method for complex model measurement tasks. This method achieves collaborative optimization of measurement viewpoints and measurement paths under realistic visibility constraints. By constructing an adjacency graph of candidate viewpoints and a visibility evaluation mechanism, the inner layer heuristically generates initial measurement paths and combines set coverage greedy supplementation and shortest path connectivity to achieve full coverage. Then, redundancy is eliminated by coverage counting and continuous executable measurement paths are reconstructed. Path clipping detection and relay point detour strategies are used to ensure safe execution. In the outer layer, a genetic algorithm is used to perform a global search of the starting viewpoints, outputting the optimal set of measurement viewpoints and the optimal measurement path with coverage and viewpoint size costs as fitness. This results in measurement planning results with high coverage, low redundancy, and engineering usability.

[0008] Furthermore, the two-layer optimized measurement planning method includes the following steps:

[0009] S1. Obtain the point cloud data and candidate viewpoint data of the complex model to be tested, and generate the adjacency relationship of the candidate viewpoints based on the grid patches of the candidate viewpoints to construct the adjacency graph of the candidate viewpoints, which is used to constrain the reachability and connectivity of the measurement path.

[0010] S2. Establish a visibility evaluation mechanism for candidate viewpoints in the test point cloud data, which is used to calculate and cache the visible point set of each candidate viewpoint under the conditions of angle constraint and occlusion culling mechanism.

[0011] S3. Perform inner layer optimization. Under the adjacency graph constraint, start from the starting viewpoint and perform heuristic traversal to generate the initial measurement path and the initial viewpoint set. Then, use the set coverage greedy supplementation strategy to iteratively supplement the viewpoints in the uncovered areas. At the same time, connect and splice the supplemented viewpoints with the initial measurement path through the shortest path search to obtain the initial executable path that satisfies the full coverage constraint.

[0012] S4. After satisfying the full coverage constraint, redundant viewpoints are removed based on the coverage count and loop removal is performed to reconstruct a continuous and executable measurement path.

[0013] S5. Perform outer layer optimization. Use a genetic algorithm to perform a global search on the starting viewpoint. Output the optimal set of measurement viewpoints and the optimal measurement path with coverage and viewpoint size costs as fitness.

[0014] Further, step S1 includes:

[0015] The point cloud data to be measured includes the spatial coordinates of the points to be measured and their corresponding normal information;

[0016] Candidate viewpoint data includes viewpoint coordinates, viewpoint normals, and viewpoint mesh patch data;

[0017] Undirected edges are generated based on vertex pairs of faces, and after deduplication, an adjacency list is constructed as the adjacency graph.

[0018] Further, step S2 includes:

[0019] S21. Let the position of the candidate viewpoint be... The corresponding normal unit vector is The location of any point to be measured is Its normal unit vector is ;

[0020] S22. Calculate the unit vector of the line of sight direction. And based on the first angle threshold Second angle threshold Filtering the set of candidate visible points that satisfy the angle constraints ;

[0021] S23, For candidate visible point set Perform occlusion detection, and let the relative viewpoint vector ,distance Set the spherical flip radius ,in The preset scaling factor; Candidate Visible Point Set The set of indices;

[0022] S24. Perform a spherical flip mapping on each candidate visible point to obtain the points containing the mapping. Mapping point set The expression is:

[0023]

[0024] S25. Map the point set Together with the origin, they form a point set. Calculate its 3D convex hull and retain the candidate points corresponding to the convex hull vertices associated with the origin as the final visible points after occlusion culling, thus obtaining the final visible point set of the candidate viewpoint.

[0025] Further, step S22 includes:

[0026] Calculate the unit vector of the line of sight direction The calculation formula is:

[0027]

[0028] Based on the first angle threshold Determine whether the angular constraint between the line of sight direction and the viewpoint normal is satisfied, i.e., determine... ;

[0029] Based on the second angle threshold Determine whether the angular constraint between the line of sight direction and the viewpoint normal is satisfied, i.e., determine... ;

[0030] If and only if the angle constraint between the line of sight direction and the normal of the viewpoint simultaneously satisfies the first angle threshold. With the second angle threshold At that time, the point to be measured The candidate visible point is determined to be the candidate viewpoint.

[0031] Further, in step S3, generating the initial measurement path and the initial viewpoint set includes:

[0032] S311. Initialize the coverage set to record the points to be measured that are covered by the current viewpoint set; add the starting viewpoint to the measurement path and viewpoint set, and update the coverage set based on the visible point set of the starting viewpoint;

[0033] S312. Under the adjacency graph constraint, select candidate viewpoints from the set of adjacent viewpoints of the current viewpoint, and obtain the corresponding set of visible points for each candidate viewpoint;

[0034] S313. For each candidate viewpoint, calculate the unfamiliarity index of the candidate viewpoint based on the proportion of uncovered test points in its visible point set. The calculation formula is as follows:

[0035]

[0036] in, The set of visible points for candidate viewpoints; This represents the set of currently covered test points. The number of elements in the set;

[0037] When the unfamiliarity index is not less than a preset threshold, the candidate viewpoint is determined as a new final viewpoint, added to the viewpoint set, and the coverage set is updated according to its visible point set.

[0038] S314. Add the selected candidate viewpoints to the measurement path according to their adjacency relationship, update the current viewpoint to the added viewpoints, and proceed to the next round of candidate selection and evaluation.

[0039] S315. When the total number of traversal steps reaches the preset limit, or when no new final viewpoint is generated for several consecutive steps and the preset limit is reached, the traversal is terminated, and the initial measurement path and the initial viewpoint set are output.

[0040] Further, in step S3, obtaining the initial executable path that satisfies the full coverage constraint using the set-cover greedy scan strategy includes:

[0041] S321. Determine the uncovered set, which is used to characterize the test points that are not yet covered by the current covered set;

[0042] S322. Using the union of the current measurement path and the viewpoint set as the seed set, perform K-ring neighborhood expansion on the adjacency graph of the candidate viewpoints to obtain a local candidate viewpoint set, and remove duplicates from the candidate viewpoints.

[0043] S323. When the local candidate viewpoint set is insufficient to supplement the uncovered set, a nearest neighbor search is performed in the candidate viewpoint coordinate space based on the uncovered test point to generate an extended candidate viewpoint set, which is then merged with the local candidate viewpoint set to form a supplementary scan candidate set.

[0044] S324. Select the candidate viewpoint that adds the largest number of new coverages to the uncovered set from the supplementary scan candidate set as the supplementary viewpoint, and add the supplementary viewpoint to the viewpoint set, while updating the coverage set;

[0045] S325, A is adopted * The shortest path search solves for the connected path from the end viewpoint of the current measurement path to the supplementary viewpoint in the adjacency graph, and then splices the connected path to the measurement path to ensure the route is reachable and continuous.

[0046] S326. Repeat S321 to S325 until the full coverage constraint is met or the preset scan iteration limit is reached.

[0047] Further, step S4 includes:

[0048] First, the coverage count is formed by counting the number of times the viewpoint set covers each test point. Under the premise of satisfying the full coverage constraint, viewpoints are tried to be deleted in reverse order. If the test points covered by a certain viewpoint are still covered by other viewpoints after the viewpoint is deleted, the viewpoint is deleted and the coverage count is updated synchronously, thereby reducing the final number of viewpoints without reducing the coverage rate.

[0049] The final set of viewpoints is then sorted and based on A * The shortest path search concatenates adjacent viewpoints into candidate paths, and performs loop removal while preserving the final viewpoint constraint to ensure that the measurement path is continuous and includes all final viewpoints.

[0050] The loop removal process includes detecting path clipping in the direct segments of adjacent final viewpoints in the candidate path. If clipping occurs, a relay point detour strategy is introduced and the detection is iterated until clipping is eliminated or the upper limit of recursion depth is reached.

[0051] Furthermore, the path penetration detection includes:

[0052] Let the coordinates of the adjacent final viewpoints be... Construct the parametric equations of line segments , sampling step size Generate sampling points ;

[0053] For each sampling point Calculate its distance to the point to be measured closest distance When sampling points exist Make the closest distance When this occurs, it is determined that the line segment has clipped through the ground, and the nearest distance is [not specified]. The calculation formula is:

[0054]

[0055] in, Let the set of points to be measured be _____. The preset collision distance threshold; It is the sampling point number when sampling along the line segment; It is a continuous parameter in the parametric equation of a line segment;

[0056] The relay point detour strategy includes:

[0057] In the generated measurement path A set of candidate relay points located within the corresponding position intervals of the two final viewpoints is selected, and relay points are selected according to the rule of prioritizing the midpoint of the interval or prioritizing distance-obstacle gain. , the original line segment Decompose into line segments With line segment and the decomposed line segments and line segments Detection of clipping in recursive execution paths;

[0058] When the recursion depth exceeds the upper limit If clipping still cannot be eliminated, it degenerates into using adjacency graph A. * The path is a continuous, executable measurement path.

[0059] Further, in step S5, the coverage rate is defined as the proportion of the number of covered test points to the total number of test points;

[0060] The viewpoint scale cost is used to characterize the size of the final set of measurement viewpoints selected in the measurement planning scheme, thereby reflecting the number of measurement stations and the resulting number of stops / triggers and measurement costs;

[0061] The fitness is based on a lexicographical order comparison criterion, prioritizing the maximization of coverage. When the coverage is the same or both meet the strict full coverage constraint, the viewpoint size cost and / or route cost are further minimized. The fitness can be synthesized into a second-level evaluation index using a weighted method.

[0062] The full coverage constraint is a strict full coverage constraint, that is, when the coverage rate... When this happens, the corresponding solution is determined to be infeasible or its fitness priority is reduced.

[0063] Furthermore, in step S5, the genetic algorithm uses the starting viewpoint as the individual encoding and generates multiple initial populations to enhance the global search capability.

[0064] By employing the above technical solution, the present invention provides a two-layer optimization measurement planning method for complex model measurement tasks, which has at least the following beneficial effects:

[0065] 1. Establish a visibility assessment mechanism that combines viewpoint and normal constraints with occlusion culling, and cache and reuse the visible point set to improve the reliability of visibility determination and reduce the overhead of repeated calculations.

[0066] 2. Employing a "heuristic construction—set coverage and scan—A" approach. * The inner layer optimization of "connectivity" efficiently fills in uncovered areas and ensures route accessibility and continuity under strict full coverage constraints, thereby improving the feasibility of the project.

[0067] 3. By using coverage counting to remove redundancy and sorting and splicing to remove loops and reconstruct the measurement path, the number of viewpoints and path redundancy are reduced without reducing coverage, thereby improving measurement efficiency.

[0068] 4. Introduce path penetration detection and relay point detour strategies to correct potential collisions and penetration connections, thereby enhancing the safety and on-site usability of the measurement path.

[0069] 5. The outer layer optimization uses a genetic algorithm to search for the global starting viewpoint and constructs the fitness based on the inner layer optimization results, so as to achieve the coordination of global exploration and local construction and stably obtain a planning solution with high coverage and low cost.

[0070] 6. The framework is modular and scalable, adaptable to different complex models and equipment constraints, and can serve as the core algorithm support for automated measurement planning and digital inspection systems. Attached Figure Description

[0071] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0072] Figure 1 This is a flowchart of the two-layer optimized measurement planning method in this invention;

[0073] Figure 2 This is a schematic diagram illustrating the application of the present invention to a complex model in a real-world scenario;

[0074] Figure 3 This is a schematic diagram of the candidate viewpoints for the complex model corresponding to the present invention;

[0075] Figure 4 This is the initial executable path diagram for generating complex models in this invention;

[0076] Figure 5 This is the optimal measurement path diagram for generating the miscellaneous model in this invention. Detailed Implementation

[0077] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This will allow for a full understanding of how the present application uses technical means to solve technical problems and achieve technical effects, and to facilitate its implementation.

[0078] This embodiment proposes a two-layer optimization measurement planning method for complex model measurement tasks. Taking candidate viewpoint sets and point cloud data of the model to be measured as input, it combines inner-layer construction and outer-layer optimization to output an optimal measurement path that satisfies coverage constraints and is executable. This method can achieve high-coverage measurement planning under complex occlusion and strict viewpoint constraints, reducing the number of viewpoints and path redundancy, and improving planning efficiency and robustness. Figure 1 As shown, the method includes the following steps:

[0079] S1. Obtain the point cloud data and candidate viewpoint data of the complex model under test, such as... Figure 2 and Figure 3 As shown, an adjacency graph of candidate viewpoints is constructed by generating viewpoint adjacency relationships based on the mesh patches of candidate viewpoints, which is used to constrain the reachability and connectivity of the measurement path. The point cloud data to be measured includes the spatial coordinates and corresponding normal information of the points to be measured; the candidate viewpoint data includes viewpoint coordinates, viewpoint normals, and viewpoint mesh patch data. More specifically, in this embodiment, undirected edges are generated based on patch vertex pairs, and after deduplication, an adjacency list is constructed as the adjacency graph.

[0080] S2. Establish a visibility evaluation mechanism for candidate viewpoints in the point cloud data under test, to calculate and cache the visible point set of each viewpoint under angle constraints and occlusion culling mechanisms. This embodiment filters candidate visible points while satisfying viewpoint / normal constraints, and obtains the true visible point set by combining the occlusion culling mechanism; the visible point set of each candidate viewpoint is cached in index form to support efficient reuse in subsequent inner layer construction, rescanning, and outer layer fitness evaluation. As a preferred example, the specific details are as follows:

[0081] S21. Let the position of the candidate viewpoint be... The corresponding normal unit vector is The location of any point to be measured is Its normal unit vector is ;

[0082] S22. Calculate the unit vector of the line of sight direction. And based on the first angle threshold Second angle threshold Filtering the set of candidate visible points that satisfy the angle constraints Specifically, it includes:

[0083] Calculate the unit vector of the line of sight direction The calculation formula is:

[0084]

[0085] Based on the first angle threshold Determine whether the angular constraint between the line of sight direction and the viewpoint normal is satisfied, i.e., determine... ;

[0086] Based on the second angle threshold Determine whether the angular constraint between the line of sight direction and the viewpoint normal is satisfied, i.e., determine... ;

[0087] If and only if the angle constraint between the line of sight direction and the normal of the viewpoint simultaneously satisfies the first angle threshold. With the second angle threshold At that time, the point to be measured The candidate visible point is determined as the candidate viewpoint;

[0088] S23, For candidate visible point set Perform occlusion detection, and let the relative viewpoint vector ,distance Set the spherical flip radius ,in The preset scaling factor; Candidate Visible Point Set The set of indices.

[0089] During the occlusion culling process, the position of the candidate viewpoint is used. Construct an inversion sphere for spherical inversion around the center of the sphere; its radius is called the spherical inversion radius. For the measurement points Calculate its relative viewpoint vector and distance To ensure that the scale of the inverted sphere matches the scale of the point set and to guarantee the numerical stability of the flip mapping, the spherical flip radius is taken as... This is used to determine the scale of the inverted sphere and control the intensity of the spherical flip mapping, so that the occlusion visibility discrimination can be achieved through the convex hull analysis of the point set formed by the flipped point set and the origin.

[0090] S24. Perform a spherical flip mapping on each candidate visible point to obtain the points containing the mapping. Mapping point set The expression is:

[0091]

[0092] S25. Map the point set Together with the origin, they form a point set. Calculate its 3D convex hull and retain the candidate points corresponding to the convex hull vertices associated with the origin as the final visible points after occlusion culling, thus obtaining the final visible point set of the candidate viewpoint.

[0093] As a preferred example, the set of visible points for each candidate viewpoint is cached as follows:

[0094] For each candidate viewpoint, a visible point index cache corresponding to its index is established. When the same viewpoint is repeatedly called during the fitness evaluation process of inner or outer optimization, the cached result is read directly to avoid repeated visibility calculation.

[0095] S3. Perform inner-layer optimization. Under the adjacency graph constraint, start from the initial viewpoint and perform heuristic traversal to generate the initial measurement path and the initial viewpoint set. Then, use the set coverage greedy supplementation strategy to iteratively supplement the viewpoints in the uncovered areas. At the same time, connect and splice the supplemented viewpoints with the initial measurement path through the shortest path search to obtain the initial executable path that satisfies the full coverage constraint.

[0096] This embodiment includes inner-layer optimization of initial path generation and coverage completion. Under the constraint of the adjacency graph, a heuristic traversal is performed starting from the initial viewpoint to generate the initial measurement path and initial viewpoint set, and the viewpoint set is gradually expanded using "new coverage contribution" as the criterion. For uncovered areas, a greedy sweep strategy for set coverage is adopted. Among the local neighborhood candidates and the expanded candidates guided by the uncovered area, the supplementary viewpoint with the largest increase in coverage is iteratively selected, and then processed by A. * The shortest path search will connect and concatenate the supplementary viewpoint with the existing path to obtain an initial executable path that satisfies the full coverage constraint, such as... Figure 4 The initial executable path is shown.

[0097] As a preferred example, the generation of the initial measurement path and initial viewpoint set includes:

[0098] S311. Initialize the coverage set to record the points to be measured that are covered by the current viewpoint set; add the starting viewpoint to the measurement path and viewpoint set, and update the coverage set based on the visible point set of the starting viewpoint;

[0099] S312. Under the adjacency graph constraint, select candidate viewpoints from the set of adjacent viewpoints of the current viewpoint, and obtain the corresponding set of visible points for each candidate viewpoint;

[0100] S313. For each candidate viewpoint, calculate the unfamiliarity index based on the proportion of uncovered test points in its visible point set. The calculation formula is:

[0101]

[0102] in, The set of visible points for candidate viewpoints; This represents the set of currently covered test points. The number of elements in the set.

[0103] When the unfamiliarity index is not less than the preset threshold, the candidate viewpoint is determined as the new final viewpoint, added to the viewpoint set, and the coverage set is updated according to its visible point set.

[0104] S314. Add the selected candidate viewpoints to the measurement path according to their adjacency relationship, update the current viewpoint to the added viewpoints, and proceed to the next round of candidate selection and evaluation.

[0105] S315. When the total number of traversal steps reaches the preset limit, or when no new final viewpoint is generated for several consecutive steps and the preset limit is reached, the traversal is terminated, and the initial measurement path and the initial viewpoint set are output.

[0106] As a preferred example, the initial executable path that satisfies the full coverage constraint using a set-coverage greedy scan strategy includes:

[0107] S321. Determine the uncovered set, which is used to characterize the test points that are not yet covered by the current covered set;

[0108] S322. Using the union of the current measurement path and the viewpoint set as the seed set, perform K-ring neighborhood expansion on the adjacency graph of the candidate viewpoints to obtain a local candidate viewpoint set, and remove duplicates from the candidate viewpoints.

[0109] S323. When the local candidate viewpoint set is insufficient to supplement the uncovered set, perform nearest neighbor search in the candidate viewpoint coordinate space based on the uncovered test point to generate an expanded candidate viewpoint set, and merge it with the local candidate viewpoint set to form a supplementary scan candidate set.

[0110] S324. Select the candidate viewpoint that adds the largest amount of coverage to the uncovered set from the supplementary scan candidate set as the supplementary viewpoint, add the supplementary viewpoint to the viewpoint set, and update the coverage set at the same time;

[0111] S325, A is adopted * The shortest path search solves for the connected path from the end viewpoint of the current measurement path to the supplementary viewpoint in the adjacency graph, and then splices the connected path to the measurement path to ensure the route is reachable and continuous.

[0112] S326. Repeat S321 to S325 until the full coverage constraint is met or the preset scan iteration limit is reached.

[0113] S4. After satisfying the full coverage constraint, redundant viewpoints are removed based on the coverage count, and loop removal is performed to reconstruct a continuous and executable measurement path. As a preferred example, this embodiment also includes inner-layer optimized redundancy removal and loop removal reconstruction, specifically as follows:

[0114] First, the coverage count is generated by counting the number of times the viewpoint set covers each test point. Under the premise of satisfying the full coverage constraint, viewpoints are tried to be deleted in reverse order. If the test points covered by a certain viewpoint are still covered by other viewpoints after the viewpoint is deleted, the viewpoint is deleted and the coverage count is updated synchronously, thereby reducing the final number of viewpoints without reducing the coverage rate.

[0115] The final set of viewpoints is then sorted and based on A *The shortest path search concatenates adjacent viewpoints into candidate paths, and performs loop removal while preserving the final viewpoint constraint to ensure that the measurement path is continuous and includes all final viewpoints.

[0116] To improve the feasibility of route engineering, loop removal processing includes path clipping detection for direct segments between adjacent final viewpoints in candidate paths. If clipping is detected, a relay point detour strategy is introduced and the detection is iterated until clipping is eliminated or the recursion depth limit is reached. Specifically, this includes:

[0117] Let the coordinates of the adjacent final viewpoints be... Construct the parametric equations of line segments , ; with sampling step size Generate sampling points For each sampling point Calculate its distance to the point to be measured closest distance When sampling points exist Make the closest distance When this occurs, it is determined that the line segment has experienced clipping; nearest distance The calculation formula is:

[0118]

[0119] in, Let the set of points to be measured be _____. The preset collision distance threshold; It is the sampling point number (discrete index) when sampling along the line segment; It is a continuous parameter in the parametric equation of a line segment.

[0120] Relay point detour strategies include:

[0121] In the generated measurement path A set of candidate relay points located within the corresponding position intervals of the two final viewpoints is selected, and relay points are selected according to the rule of prioritizing the midpoint of the interval or prioritizing distance-obstacle gain. , the original line segment Decompose into line segments With line segment and the decomposed line segments and line segments Detection of clipping in recursive execution paths; when the recursion depth exceeds the upper limit. If clipping still cannot be eliminated, it degenerates into using adjacency graph A. * The path is a continuous, executable measurement path.

[0122] This embodiment, after satisfying the full coverage constraint, counts the coverage times of each test point and performs redundancy removal on the viewpoint set accordingly, reducing the final number of viewpoints without reducing coverage; then, the final viewpoints are sorted and based on A * The shortest path search is performed by concatenating segments to form candidate paths. Loop removal is then performed on the candidate paths while ensuring that the final viewpoint is not deleted, thereby reducing path loops and redundancy and improving path continuity and execution efficiency.

[0123] S5. Perform outer-layer optimization by using a genetic algorithm to perform a global search on the starting viewpoint. Output the optimal set of measurement viewpoints and the optimal measurement path using coverage and viewpoint size costs as fitness. Coverage is defined as the proportion of covered measurement points to the total number of measurement points.

[0124] Viewpoint size cost characterizes the size of the final set of measurement viewpoints selected in the measurement planning scheme, thus reflecting the number of measurement stations and the resulting number of stops / triggers and measurement costs. Let the final set of measurement viewpoints be... Then the viewpoint scale cost Defined as the final number of measurement viewpoints, i.e.:

[0125]

[0126] in, Represents the final set of measurement viewpoints The number of elements.

[0127] Preferably, the final measurement viewpoint set The final set of viewpoints, obtained after supplementary scanning and redundancy removal, after satisfying coverage constraints (preferably strict full coverage), therefore the viewpoint size cost... Minimizing is equivalent to reducing the number of measurement viewpoints while meeting coverage requirements.

[0128] Viewpoint size cost when candidate viewpoints have different measurement costs This can be expanded into a weighted summation form, namely:

[0129]

[0130] in, From the perspective The preset cost coefficient; when Time degenerates into .

[0131] The fitness criterion is based on lexicographical order comparison, prioritizing maximizing coverage. When coverage is the same or both meet the strict full coverage constraint, the viewpoint size cost and / or route cost are further minimized, and a weighted approach can be used to synthesize them into a second-level evaluation index.

[0132] As a preferred example, this embodiment uses a genetic algorithm to perform a global search on the starting viewpoint, that is, using the starting viewpoint as the individual code and generating multiple initial populations to enhance the global search capability.

[0133] As a preferred example, the full coverage constraint is a strict full coverage constraint, that is, when the coverage rate... When this happens, the corresponding solution is determined to be infeasible or its fitness priority is reduced.

[0134] As a preferred example, this embodiment writes the optimal measurement path and the optimal set of measurement viewpoints output in step S5 into a point cloud file. The point cloud file contains at least viewpoint coordinates and viewpoint normal fields, so that the measurement equipment or measurement planning system can directly call them.

[0135] This embodiment employs a genetic algorithm to perform a global search of the starting viewpoint. The coverage and viewpoint size / path cost output from the inner layer construction are used as fitness evaluation criteria to iteratively obtain better initial conditions and drive the improvement of the overall solution. During path generation, path crossing detection is further performed on the direct connection segments between adjacent final viewpoints. If crossing is detected, a relay point is introduced for detour and iterative verification continues until the non-crossing constraint is met or a preset termination condition is triggered, thereby outputting the optimal measurement path. Figure 5 The optimal measurement path.

[0136] Through the above steps, the present invention achieves coordinated optimization of measurement viewpoint selection and measurement path planning under complex occlusion and strict visibility constraints, and can stably obtain measurement planning results with high coverage, low redundancy and engineering feasibility.

[0137] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0138] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. Since the above embodiments are substantially similar to the method embodiments, their descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0139] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A two-level optimization measurement planning method for complex model measurement tasks, characterized in that, The method includes the following steps: S1. Obtain the point cloud data and candidate viewpoint data of the complex model to be tested, and generate the adjacency relationship of the candidate viewpoints based on the grid patches of the candidate viewpoints to construct the adjacency graph of the candidate viewpoints, which is used to constrain the reachability and connectivity of the measurement path. S2. Establish a visibility evaluation mechanism for candidate viewpoints in the test point cloud data, which is used to calculate and cache the visible point set of each candidate viewpoint under the conditions of angle constraint and occlusion culling mechanism. S3. Perform inner-layer optimization. Under the adjacency graph constraint, start from the initial viewpoint and perform a heuristic traversal to generate the initial measurement path and initial viewpoint set. Then, use a greedy sweep strategy with set coverage to iteratively supplement viewpoints in the uncovered areas. At the same time, connect and splice the supplemented viewpoints with the initial measurement path through shortest path search to obtain an initial executable path that satisfies the full coverage constraint, including: S321. Determine the uncovered set, which is used to characterize the test points that are not yet covered by the current covered set; S322. Using the union of the current measurement path and the viewpoint set as the seed set, perform K-ring neighborhood expansion on the adjacency graph of the candidate viewpoints to obtain a local candidate viewpoint set, and remove duplicates from the candidate viewpoints. S323. When the local candidate viewpoint set is insufficient to supplement the uncovered set, a nearest neighbor search is performed in the candidate viewpoint coordinate space based on the uncovered test point to generate an extended candidate viewpoint set, which is then merged with the local candidate viewpoint set to form a supplementary scan candidate set. S324. Select the candidate viewpoint that adds the largest number of new coverages to the uncovered set from the supplementary scan candidate set as the supplementary viewpoint, and add the supplementary viewpoint to the viewpoint set, while updating the coverage set; S325, A is adopted * The shortest path search solves for the connected path from the end viewpoint of the current measurement path to the supplementary viewpoint in the adjacency graph, and then splices the connected path to the measurement path to ensure the route is reachable and continuous. S326. Repeat S321 to S325 until the full coverage constraint is met or the preset scan iteration limit is reached. S4. After satisfying the full coverage constraint, redundant viewpoints are removed based on the coverage count, and loop removal is performed to reconstruct a continuous and executable measurement path, including: First, the coverage count is formed by counting the number of times the viewpoint set covers each test point. Under the premise of satisfying the full coverage constraint, viewpoints are tried to be deleted in reverse order. If the test points covered by a certain viewpoint are still covered by other viewpoints after the viewpoint is deleted, the viewpoint is deleted and the coverage count is updated synchronously, thereby reducing the final number of viewpoints without reducing the coverage rate. The final set of viewpoints is then sorted and based on A * The shortest path search concatenates adjacent viewpoints into candidate paths, and performs loop removal while preserving the final viewpoint constraint to ensure that the measurement path is continuous and includes all final viewpoints. The loop removal process includes performing path clipping detection on the direct segments of adjacent final viewpoints in the candidate path. If clipping is found, a relay point detour strategy is introduced and the detection is iterated until clipping is no longer found or the upper limit of recursion depth is reached. S5. Perform outer layer optimization. Use a genetic algorithm to perform a global search on the starting viewpoint. Output the optimal set of measurement viewpoints and the optimal measurement path with coverage and viewpoint size costs as fitness.

2. The dual-layer optimized measurement planning method according to claim 1, characterized in that, Step S1 includes: The point cloud data to be measured includes the spatial coordinates of the points to be measured and their corresponding normal information; Candidate viewpoint data includes viewpoint coordinates, viewpoint normals, and viewpoint mesh patch data; Undirected edges are generated based on vertex pairs of faces, and after deduplication, an adjacency list is constructed as the adjacency graph.

3. The dual-layer optimized measurement planning method according to claim 1, characterized in that, Step S2 includes: S21. Let the position of the candidate viewpoint be... The corresponding normal unit vector is The location of any point to be measured is Its normal unit vector is ; S22. Calculate the unit vector of the line of sight direction. And based on the first angle threshold Second angle threshold Filtering the set of candidate visible points that satisfy the angle constraints ; S23, For candidate visible point set Perform occlusion detection, and let the relative viewpoint vector ,distance Set the spherical flip radius ,in The preset scaling factor; For candidate visible point set The set of indices; S24. Perform a spherical flip mapping on each candidate visible point to obtain the points containing the mapping. mapping point set The expression is: ； S25. Map the point set Together with the origin, they form a point set. Calculate its 3D convex hull and retain the candidate points corresponding to the convex hull vertices associated with the origin as the final visible points after occlusion culling, thus obtaining the final visible point set of the candidate viewpoint.

4. The dual-layer optimized measurement planning method according to claim 3, characterized in that, Step S22 includes: Calculate the unit vector of the line of sight direction The calculation formula is: ； Based on the first angle threshold Determine whether the angular constraint between the line of sight direction and the viewpoint normal is satisfied, i.e., determine... ; Based on the second angle threshold Determine whether the angular constraint between the line of sight direction and the viewpoint normal is satisfied, i.e., determine... ; If and only if the angle constraint between the line of sight direction and the normal of the viewpoint simultaneously satisfies the first angle threshold. With the second angle threshold At that time, the point to be measured The candidate visible point is determined to be the candidate viewpoint.

5. The dual-layer optimized measurement planning method according to claim 1, characterized in that, In step S3, generating the initial measurement path and the initial viewpoint set includes: S311. Initialize the coverage set to record the points to be measured that are covered by the current viewpoint set; add the starting viewpoint to the measurement path and viewpoint set, and update the coverage set based on the visible point set of the starting viewpoint; S312. Under the adjacency graph constraint, select candidate viewpoints from the set of adjacent viewpoints of the current viewpoint, and obtain the corresponding set of visible points for each candidate viewpoint; S313. For each candidate viewpoint, calculate the unfamiliarity index of the candidate viewpoint based on the proportion of uncovered test points in its visible point set. The calculation formula is as follows: ； in, The set of visible points for candidate viewpoints; This represents the set of currently covered test points. The number of elements in the set; When the unfamiliarity index is not less than a preset threshold, the candidate viewpoint is determined as a new final viewpoint, added to the viewpoint set, and the coverage set is updated according to its visible point set. S314. Add the selected candidate viewpoints to the measurement path according to their adjacency relationship, update the current viewpoint to the added viewpoints, and proceed to the next round of candidate selection and evaluation. S315. When the total number of traversal steps reaches the preset limit, or when no new final viewpoint is generated for several consecutive steps and the preset limit is reached, the traversal is terminated, and the initial measurement path and the initial viewpoint set are output.

6. The dual-layer optimized measurement planning method according to claim 1, characterized in that, The path penetration detection includes: Let the coordinates of the adjacent final viewpoints be... Construct the parametric equations of line segments , sampling step size Generate sampling points ; For each sampling point Calculate its distance to the point to be measured closest distance When sampling points exist Make the closest distance When this occurs, it is determined that the line segment has clipped through the ground, and the nearest distance is [not specified]. The calculation formula is: ； in, Let the set of points to be measured be _____. The preset collision distance threshold; It is the sampling point number when sampling along the line segment; It is a continuous parameter in the parametric equation of a line segment; The relay point detour strategy includes: In the generated measurement path A set of candidate relay points located within the corresponding position intervals of the two final viewpoints is selected, and relay points are selected according to the rule of prioritizing the midpoint of the interval or prioritizing distance-obstacle gain. , the original line segment Decompose into line segments With line segment and the decomposed line segments and line segments Detection of clipping in recursive execution paths; When the recursion depth exceeds the upper limit If clipping still cannot be eliminated, it degenerates into using adjacency graph A. * The path is a continuous, executable measurement path.

7. The dual-layer optimized measurement planning method according to claim 1, characterized in that, In step S5, the coverage rate is defined as the proportion of the number of covered test points to the total number of test points; The viewpoint scale cost is used to characterize the size of the final set of measurement viewpoints selected in the measurement planning scheme, thereby reflecting the number of measurement stations and the resulting number of stops / triggers and measurement costs; The fitness is based on a lexicographical order comparison criterion, prioritizing the maximization of coverage. When the coverage is the same or both meet the strict full coverage constraint, the viewpoint size cost and / or route cost are further minimized. The fitness can be synthesized into a second-level evaluation index using a weighted method. The full coverage constraint is a strict full coverage constraint, that is, when the coverage rate... When this happens, the corresponding solution is determined to be infeasible or its fitness priority is reduced.

8. The dual-layer optimized measurement planning method according to claim 1, characterized in that, In step S5, the genetic algorithm uses the starting viewpoint as the individual encoding and generates multiple initial populations to enhance the global search capability.

Images