Digital mapping method based on polygon modeling

Abstract

The present application belongs to the technical field of digital surveying and mapping, and relates to a digital surveying and mapping method based on polygon modeling. In view of the polygon grid topology hole problem caused by the sheltered area in the mobile surveying and mapping of urban environment, the present application proposes a constraint Delaunay hole repairing method based on scanning station visible cone and luminosity gradient guidance. The method constructs a scanning station visible cone for the hole in the initial triangulation grid, screens the visible station set by ray-triangle intersection test, calculates the multi-view fusion luminosity gradient field by using the visible station synchronous image, takes the gradient maximum value point as the Steiner constraint point to perform the constraint Delaunay triangulation, completes the topology repair while retaining the geometric mutation features such as building line feet, and writes the confidence attribute determined by the visible station coverage sufficiency and multi-view gradient consistency into each patch, so as to realize the patch-level quantification and hierarchical output of the repair quality.

Description

Technical Field

[0001] This invention relates to the field of digital surveying and mapping technology, and more particularly to a digital surveying and mapping method based on polygon modeling. Background Technology

[0002] In the field of polygon-based digital mapping, when mobile mapping platforms scan objects such as building facades, bridge piers, and road facilities in urban environments, cantilevered structures, vegetation canopies, and vehicle occlusion create systematic blind spots in the lidar point cloud, causing topological holes in the initial triangulated polygon mesh in the occluded areas. Existing Poisson surface reconstruction methods, with global smoothness as the optimization objective, over-smooth at geometrically abrupt locations such as building moldings and window sill edges when repairing holes, resulting in a systematic deviation of 2 to 5 millimeters between the repaired polygonal patches and the real geometric surfaces, leading to a loss of measurement accuracy. In areas of deep occlusion, spherical rotation algorithms generate non-manifold T-junctions or residual holes, producing topologically invalid meshes, causing subsequent volume calculations and texture mapping to fail.

[0003] Both of the aforementioned methods share a common drawback: they utilize only the geometric information of the hole boundaries during the restoration process, failing to incorporate the visibility information from the surveying instrument's scanning station and the synchronously acquired image data into the restoration constraints. This results in a lack of correspondence between the triangulation method in the restoration result and the geometric abrupt changes in the actual scene. Furthermore, existing methods do not attach any quality quantification attributes to the restored polygonal patches. Operators cannot determine, without field re-measurement, which restored patches have sufficient measurement reliability and which require supplementary data collection for verification. This renders the restoration quality untraceable throughout the entire deliverables process. Summary of the Invention

[0004] The purpose of this invention is to provide a digital mapping method based on polygon modeling to solve the problems existing in the prior art.

[0005] This invention provides a digital mapping method based on polygon modeling, comprising the following steps: Perform a topological traversal on the initial triangulated polygon mesh, mark the half-edges referenced by only one triangle facet as hole boundary edges, record the closed loops formed by connecting the first and last hole boundary edges as holes H, and extract the ordered sequence of boundary vertices of holes H and the centroid coordinates cH of holes. Traverse the scanning station pose sequence recorded by the mapping platform during the acquisition process, emit a detection ray for each scanning station with cH as the target point, perform an intersection test between the ray and the triangulated mesh surface, and add the scanning station whose detection ray has not intersected with any mesh surface before reaching cH to the set of visible stations VH. For each scanning station in the visible station set VH, the aperture boundary vertex is projected onto the synchronous image plane using the pre-calibrated camera intrinsic parameter matrix Ak and extrinsic parameter pose matrix [Rk|tk]. The region enclosed by the aperture boundary projection polygon is taken as the region of interest. The horizontal and vertical image gradient components are calculated using the 5×5 Sobel operator. The gradient magnitude Gk is obtained by taking the square root of the sum of the squares of the two components. The image coordinates are then mapped back to the three-dimensional coordinates q of the aperture plane through inverse projection. The multi-view fusion photometric gradient field is calculated as follows: G(q)=(1 / |VH|)×ΣGk(πk(q)), k∈VH, where πk is the projection function of the k-th station projecting the three-dimensional point to the two-dimensional image coordinates, |VH| is the number of elements in the visible station set, and Gk(πk(q)) is the Sobel gradient magnitude of the k-th station image at the corresponding projection coordinates q. The gradient field G is binarized using 1.5 times the median gradient magnitude of the whole image in the region of interest as the threshold τG. Local gradient maxima are extracted using the 3×3 neighborhood maxima suppression operator. Each maxima is mapped back to the three-dimensional coordinates of the hole plane through inverse projection to obtain the Steiner constraint point set sm. Using the ordered vertex sequence of the hole boundary ring as the external constraint boundary and the Steiner constraint point set sm as the internal forced insertion point, perform constrained Delaunay triangulation to generate a patch face set TH. The triangulation result guarantees that the circumcircle of any generated triangle does not contain other constraint points, and all Steiner points appear in the vertices of the triangulated triangle. For each triangular facet f in the patch facet set TH, calculate the confidence score cf using the following formula and write it into its attribute field: cf=min(1,|VH| / Nref)×exp(-σ 2 G(f) / σ 2 0), where Nref is the preset number of reference visible stations, σ 2 G(f) is the variance among the back projection gradient magnitudes of each visible station within the coverage area of ​​patch f, σ 2 0 is a normalization constant determined by the mean variance of the hole gradients in the entire dataset; Patches with a confidence level below the preset lower limit cmin are stored in the verification layer and are not included in the volume and area measurement calculations.

[0006] The above scheme constructs a visible cone for each hole and performs a ray-triangle intersection test to select truly visible scanning stations as effective information sources, avoiding interference from image data of occluded scanning stations on gradient field calculations. By performing Sobel gradient calculations on the synchronous images of visible stations and back-projecting and fusing the gradient field onto the hole plane, the repair process can perceive the true distribution of geometrical abrupt changes such as building moldings and welds within the hole area. By using the maximum point of the photometric gradient as the Steiner insertion point to constrain the Delaunay triangulation, the triangulation is forced to generate natural boundaries at geometrical abrupt changes, thereby completing topological repair while preserving key measurement features. By writing a confidence attribute determined by the sufficiency of visible station coverage and the consistency of multi-view gradients to each patch, the repair quality is quantitatively expressed at the patch granularity, providing a traceable data foundation for subsequent measurement statistics and supplementary acquisition decisions.

[0007] Furthermore, after constructing the visible station set VH for the hole H, if VH is empty, the hole is marked as a hard occlusion type and the process is switched to geometric interpolation based on boundary curvature. Only when VH is not empty does the hole enter the multi-view fusion photometric gradient field calculation process.

[0008] Furthermore, each scanning station in the scanning station pose sequence consists of three-dimensional center coordinates Ok and a main view unit vector dk. The ray-triangle intersection test is performed one by one on the detection ray emitted along the direction Ok→cH and on all the triangulated mesh patches.

[0009] Furthermore, the patches in the patch set TH are divided into three levels according to their confidence level cf: patches with a confidence level of not less than 0.7 are marked as high confidence level and incorporated into the main mapping result grid; patches with a confidence level between 0.3 and 0.7 are marked as unverified level, stored in an independent layer and a verification task work order is generated for the corresponding location; patches with a confidence level of less than 0.3 are marked as low confidence level, retained only as a visualization reference and not included in any measurement statistics.

[0010] Furthermore, the normalization constant σ 2 0 is determined by the mean of the variance of the gradient magnitude obtained from the multi-view fusion photometric gradient field calculation of all holes in the entire dataset.

[0011] Furthermore, when low-confidence patches form continuous regions in the spatial distribution, the system generates a supplementary acquisition suggestion report based on the spatial coordinate distribution and normal vector orientation of the low-confidence patches. After accessing the supplementary scanning data, the system re-executes the complete process of visible station set construction, multi-view fusion photometric gradient field calculation, constrained Delaunay triangulation, and confidence calculation for the corresponding holes.

[0012] Furthermore, the surveying platform is a vehicle-mounted mobile surveying system. The scanning station pose sequence consists of multiple pose nodes continuously recorded during the vehicle's movement. The synchronous camera is installed on the vehicle and faces at least two different directions. The camera's internal and external parameters are calibrated before leaving the warehouse.

[0013] Furthermore, the surveying platform is a ground tripod-mounted static ground lidar. The pose sequence of the scanning stations consists of the poses of multiple independent scanning stations deployed in the survey area. Each scanning station is equipped with a fisheye camera. After scanning is completed, the fisheye camera synchronously acquires panoramic images. The joint calibration residual between the fisheye camera and the lidar coordinate system does not exceed 0.3 pixels.

[0014] The beneficial effects of this invention are: First, this invention introduces the visible cone constraint of the scanning station into the hole repair process. The scanning station with true visibility of the hole is selected by ray-triangle intersection test, and the maximum point in the multi-view fusion photometric gradient field of the synchronous image is used as the Steiner forced insertion point for constraining Delaunay triangulation. This ensures that the boundary of the repaired triangular facet corresponds spatially with the geometric abrupt changes such as building moldings and window sills in the scene. This overcomes the defect of the Poisson surface reconstruction method that produces excessive smoothing at geometric abrupt changes, resulting in loss of measurement accuracy.

[0015] Second, this invention calculates the confidence attribute of each patch triangle obtained from the repair according to two dimensions: the sufficiency of visible station coverage and the consistency of multi-view gradient field variance, and writes it into the patch attribute field. This quantifies the repair quality to the patch granularity, which solves the problem that existing hole repair methods cannot classify the quality of repair results and that operators cannot judge the reliability of patches without field retesting. This achieves the traceability of repair quality throughout the entire process of deliverables.

[0016] Third, when a low-confidence patch forms a continuous region, the present invention automatically generates a supplementary acquisition suggestion report based on the spatial coordinate distribution and normal vector orientation of the low-confidence patch. After the supplementary scan data is accessed, the corresponding hole is re-executed with a complete repair process, forming an iterative acquisition closed loop driven by confidence. This enables surveying personnel to increase the proportion of high-confidence patches in the target area to a level that meets measurement requirements with minimal supplementary scanning workload. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1This is a flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the visible cone construction of the present invention; Figure 3 This is a schematic diagram of the constrained Delaunay triangulation of the present invention; Figure 4 This is the closed-loop diagram of the confidence level output of the present invention. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should also be noted that, to make the embodiments more comprehensive, the following embodiments are the best and preferred embodiments, and those skilled in the art can use other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0020] It should be noted that the use of terms such as "an embodiment," "an embodiment," "an exemplary embodiment," and "some embodiments" in the specification indicates that the described embodiment may include a specific feature, structure, or characteristic, but not every embodiment necessarily includes that specific feature, structure, or characteristic. Furthermore, when a specific feature, structure, or characteristic is described in connection with an embodiment, implementing such a feature, structure, or characteristic in conjunction with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the art.

[0021] Generally, terms can be understood at least partly from their use in context. For example, depending at least partly on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in a singular sense, or a combination of features, structures, or characteristics in a plural sense. Additionally, the term "based on" can be understood not necessarily to convey an exclusive set of factors, but rather, alternatively, depending at least partly on the context, to allow for the presence of other factors that are not necessarily explicitly described.

[0022] See Figures 1 to 4 As shown This invention relates to a digital mapping method based on polygon modeling. Addressing the problem of polygon mesh topological holes caused by occlusion areas in urban environmental mobile mapping, a constrained Delaunay hole repair method based on the visible cone of the scanning station and photometric gradient guidance is proposed. The specific implementation is described below.

[0023] Hole identification and classification After the mobile mapping platform completes the scanning, the system performs a topological traversal on the initial triangulated polygon mesh, identifies all open boundary loops, marks each half-edge referenced by only one triangular facet as a hole boundary edge, and records the closed loop formed by connecting the ends of such half-edges as a hole H, extracting its ordered sequence of boundary vertices and the centroid coordinates of the hole. .

[0024] Visible cone construction at the scanning station For each hole H, the system iterates through the sequence of scanning station poses recorded by the mapping platform during the acquisition process. The k-th scanning station in this sequence consists of the three-dimensional center coordinates Ok and the main view unit vector d̂k. The system emits a detection ray along the direction Ok→cH with cH as the target point and performs a ray-triangle intersection test with the triangulated mesh facets. If the detection ray does not intersect with any facet before reaching cH, the k-th scanning station is determined to be visible to the hole H and is added to the set of visible stations VH. If all scanning stations are not visible, the hole is marked as a hard occlusion type and the process is switched to the standby process based on boundary curvature geometric interpolation. If VH is not empty, the hole enters the photometric gradient guidance process.

[0025] Epipolar projection of multiview images and calculation of photometric gradient field right For each visible scanning station k, the system uses a pre-calibrated camera intrinsic parameter matrix Ak and extrinsic pose matrix. The three-dimensional coordinates of the aperture boundary vertex are projected onto the synchronous image plane of the station. The region enclosed by the projected polygon of the aperture boundary is taken as the region of interest. Within the region of interest, the image gradients in the horizontal and vertical directions are calculated using a 5×5 Sobel operator. The square root of the sum of the squares of the two gradient components is taken to obtain the gradient magnitude of the station at image coordinate x. Then, the image coordinates are mapped back to the three-dimensional coordinates q of the aperture plane through inverse projection. Finally, the gradient magnitudes of all visible stations are arithmetically averaged on the aperture plane to obtain the multi-view fused photometric gradient field, as shown in the following formula:

[0026] Where q represents a three-dimensional sampling point within the plane of the hole. For the first The projection function that projects 3D points onto 2D image coordinates. The number of elements in the set of visible stations. Let be the Sobel gradient magnitude of the image at station k at the corresponding projection coordinate q.

[0027] Steiner point extraction The system binarizes the fused gradient field G with a threshold τG, where τG is 1.5 times the median gradient magnitude of the entire image within the region of interest. Local gradient maxima are extracted from the binarized result using a 3×3 neighborhood maxima suppression operator. Each maxima is then mapped back to the 3D coordinates of the hole plane via inverse projection, resulting in a set of Steiner constraint points. These points physically correspond to geometric abrupt changes in the area of ​​the hole, such as building moldings and window sill edges.

[0028] Constrained Delaunay triangulation The system uses the polygon formed by the ordered sequence of vertices of the hole boundary loop as the external constraint boundary, and the Steiner point set as the external constraint boundary. As an internal forced insertion point, constrained Delaunay triangulation is performed on the hole plane. The triangulation result ensures that the circumcircle of any generated triangle does not contain other constraint points, and all Steiner points must appear in the vertices of the triangulated triangles. This ensures that the triangulation produces natural boundaries at geometric abrupt changes, avoids interpolation across discontinuous surfaces, and preserves key metric features such as acute angles of the building. The set of patch faces generated by the triangulation is denoted as . .

[0029] Face-by-face confidence attribute writing For each newly generated triangle facet f in the patch facet set TH, the system calculates and writes the confidence score cf to its attribute field. The calculation formula is as follows:

[0030] Where Nref is the system's preset number of reference visible stations, used to normalize the coverage of visible stations to the range of 0 to 1; The variance of the back-projection gradient magnitudes of each visible station within the coverage area of ​​patch f reflects the consistency of the gradient field from multiple perspectives. This is a normalization constant, determined by the mean of the variance of the hole gradients across the entire dataset. The first term in the formula measures the adequacy of scan coverage, and the second term measures the consistency of multi-view gradients. The multiplication of these two terms causes the confidence level to decrease simultaneously when there is insufficient coverage or inconsistency between viewpoints. Patches with a confidence level below the preset lower limit cmin are stored separately in the verification layer and are not included in subsequent volume and area measurement calculations, but are retained in the model file for operator review.

[0031] Confidence level output and subsequent processing After the repair is completed, the system classifies all patch areas into three levels based on their confidence level: patches with a confidence level of not less than 0.7 are marked as high confidence level and directly incorporated into the main mapping result grid; patches with a confidence level between 0.3 and 0.7 are marked as pending verification level, stored in an independent layer, and a verification task work order is generated for the corresponding location; patches with a confidence level below 0.3 are marked as low confidence level, retained only as a visual reference, and not included in any measurement statistics, thereby achieving full traceability of the repair quality.

[0032] Example 1 This embodiment uses digital mapping of elevated bridge piers on urban expressways as the application scenario. The scanning platform is a vehicle-mounted mobile mapping system equipped with a rotating multi-line lidar and four synchronous cameras. The four cameras are respectively oriented towards the left front, right front, left rear, and right rear directions of the vehicle's travel direction. The internal and external parameters are calibrated before leaving the warehouse. The scanning vehicle travels along the lanes on both sides of the bridge pier once, forming two scanning tracks. Each track records 500 pose nodes, and the track spacing is about 6 meters.

[0033] After initial triangulation, the system identified 14 holes at the junction of the pier root and the ground. These holes were caused by the obstruction of the lidar by the steel casing at the bottom of the pier. Among them, 12 holes had at least one scanning station that passed the visibility test and were included in the photometric gradient guidance process. The remaining 2 holes were marked as hard obstruction type because the trajectories on both sides were completely obstructed by the casing, and were transferred to the standby process of geometric interpolation based on boundary curvature.

[0034] Taking the largest hole (No. 1) among the 14 holes as an example, the boundary loop of this hole consists of 38 vertices, and its centroid coordinates are... Located at the bottom of the arc at the base of the bridge pier. The system performed ray-triangle intersection tests on each of the 1000 scanning stations along the two trajectories, detecting that 127 scanning stations were visible to the hole, forming a set of visible stations. The number of visible stations The system defaults to 127, and the number of reference visible stations Nref is 150.

[0035] The system extracts the Sobel gradient magnitudes of the regions of interest from synchronized images of 127 visible stations, backprojects the gradient fields onto the hole plane, and performs an arithmetic mean to obtain the fused gradient field G. A threshold of 1.5 times the median gradient magnitude of this region is then used. Nine Steiner points were extracted from the binarization results. These nine points were confirmed by manual sampling to fall on the edge of the reserved holes in the steel bars at the root of the pier and the location of the weld seam of the casing.

[0036] Constrained Delaunay triangulation was performed using 38 boundary vertices and 9 Steiner points, generating 64 patch triangles. The confidence score for each patch was calculated, including the gradient variance. The normalization constant for the entire dataset is calculated from the gradient magnitudes of 127 viewpoints within the area of ​​this patch. With a value of 120, the confidence levels of the 64 patches are distributed as follows: 52 patches have a confidence level of not less than 0.7 and are classified as high confidence level; 11 patches have a confidence level between 0.3 and 0.7 and are classified as unverified level; and 1 patch, due to falling in the geometric dead corner of two scanning trajectories, has a confidence level of 0.18 and is classified as low confidence level.

[0037] After this embodiment was completed, the high-reliability surface patch of the repair area of ​​hole No. 1 was verified by field retesting. The root mean square angle error between the surface patch normal vector and the measured surface normal vector was 1.3 degrees, which is better than the 3.7 degrees of traditional Poisson reconstruction in the same scene. Moreover, the arc-straight line intersection edge at the root of the pier was completely preserved in the repair result, and the smooth transition of 2 to 5 mm radius produced by traditional Poisson reconstruction at this position was completely eliminated.

[0038] Example 2 This embodiment uses digital mapping of the building facades in a commercial district as the application scenario. The scanning platform is a ground tripod-type static ground lidar. A total of 8 scanning stations are set up on both sides of the street, with a station spacing of about 15 meters. After each scanning station completes a panoramic 360-degree scan, it simultaneously triggers the fisheye camera mounted on the top of the scanner to acquire a panoramic image. The 8 fisheye cameras complete joint calibration with the lidar coordinate system before scanning, and the calibration residual does not exceed 0.3 pixels.

[0039] Due to the presence of multiple protruding canopies and air conditioner unit brackets on the target building's facade, the system identified 31 holes on the facade after initial triangulation. Nine of these holes were located beneath the canopy at the first-floor entrance, caused by the canopy systematically obstructing multiple scanning stations. The system performed a ray-triangle visibility check on all eight scanning stations. Seven of the nine holes had at least one visible station, while the remaining two holes were invisible to all eight scanning stations due to their recessed entrance structure, and were then moved to a geometric interpolation backup process.

[0040] Taking the No. 3 hole under the canopy on the right side of the entrance as an example, the boundary ring of this hole is composed of 22 vertices. The system detected that 3 scanning stations are visible to this hole, with the number of visible stations |VH| being 3. The system presets the number of reference visible stations Nref to be 8. The panoramic fisheye images corresponding to the 3 visible stations are from scanning stations No. 3, 4, and 5 on the same side of the street, respectively.

[0041] Since the fisheye camera has a field of view of 185 degrees, the images from the three visible stations completely cover the projection area of ​​the hole boundary. The system performs Sobel gradient calculation and back-projection fusion on the three fisheye images respectively. Using 1.5 times the median gradient magnitude as the threshold τG, a total of 5 Steiner points are extracted. These 5 points correspond to the welding node positions of the canopy-type steel frame and the horizontal mortar joint positions of the wall brickwork, respectively, after coordinate verification.

[0042] Constrained Delaunay triangulation was performed using 22 boundary vertices and 5 Steiner points, generating 37 patch triangle faces. In the confidence calculation, due to the limited number of visible stations (3) and reference stations (8), the first term min(1,3 / 8) in the formula was set to 0.375, significantly suppressing the overall confidence. Of the 37 faces, only 14 had a confidence level of at least 0.7, 17 had a confidence level between 0.3 and 0.7, and 6 had a confidence level below 0.3, all of which were assigned to their respective hierarchical layers.

[0043] Unlike Example 1, this example triggers the system's supplementary data collection suggestion module when the overall confidence level is low. Based on the spatial distribution and orientation of the low-confidence patches, the module automatically generates a supplementary scanning suggestion report, indicating that a temporary scanning station should be added about 5 meters in front of the building entrance, which can increase the number of visible stations in the area to 6, and is estimated to improve the confidence level of all 37 patches to above 0.5.

[0044] After the construction team adopted the suggestion, the system re-executed the complete visible cone construction, photometric gradient fusion, constrained Delaunay partitioning, and confidence calculation for hole No. 3 after the new scanning station data was accessed. After the supplementary scanning, the visible station set VH expanded to 6, and the number of high-confidence patches among the 37 patches increased to 29. Compared with the initial repair results, the proportion of high-confidence patches increased from 37.8% to 78.4%, which verified the practical application value of the present invention in confidence-driven iterative acquisition optimization.

[0045] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0046] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A digital mapping method based on polygon modeling, characterized in that, Includes the following steps: Perform a topological traversal on the initial triangulated polygon mesh, marking the half-edges referenced by only one triangular facet as hole boundary edges, recording the closed loops formed by connecting the head and tail of the hole boundary edges as holes H, and extracting the ordered sequence of boundary vertices of holes H and the centroid coordinates of the holes. ; Iterate through the sequence of scanning station poses recorded by the surveying platform during the data acquisition process, and perform a search on each scanning station. A detection ray is emitted to the target point, and an intersection test is performed between the ray and the triangulated mesh facet. The detection ray is then positioned at the target point. Scan stations that did not previously intersect with any mesh patches are added to the set of visible stations. ; For the set of visible stations Each scanning station uses a pre-calibrated camera intrinsic parameter matrix. With extrinsic pose matrix The vertices of the hole boundary are projected onto the synchronous image plane. The region enclosed by the projected polygon of the hole boundary is the region of interest. The horizontal and vertical image gradient components are calculated using a 5×5 Sobel operator. The square root of the sum of the squares of the two components is taken to obtain the gradient magnitude. Then, the image coordinates are mapped back to the three-dimensional coordinates q of the aperture plane through inverse projection, and the multi-view fusion photometric gradient field is calculated using the following formula: ,in Let be the projection function that projects a 3D point to a 2D image coordinate system at station k. The number of elements in the set of visible stations. Let $\frac{k}{q}$ be the Sobel gradient magnitude of the image at the corresponding projection coordinates of station $k$. The threshold is set to 1.5 times the median gradient magnitude of the entire image within the region of interest. The gradient field G is binarized, and local gradient maxima are extracted using a 3×3 neighborhood maxima suppression operator. Each maxima is then mapped back to the 3D coordinates of the hole plane via inverse projection to obtain the set of Steiner constraint points. ; Using the ordered sequence of vertices of the hole boundary ring as the external constraint boundary and the Steiner constraint point set as the external constraint boundary. Perform constrained Delaunay triangulation on the internal forced insertion points to generate a set of patch faces. The partitioning result guarantees that the circumcircle of any generated triangle does not contain other constraint points, and all Steiner points appear in the vertices of the partitioned triangle. For patch set The confidence level of each triangular facet f is calculated using the following formula. And write it into its attribute field: ,in The preset number of reference visible stations, Let f be the variance among the back-projection gradient magnitudes of each visible station within the coverage area of ​​patch f. The normalization constant is determined by the mean variance of the hole gradients in the entire dataset; Confidence level below the preset lower limit The surface patches are stored in the verification layer and are not included in the volume and area measurement calculations.

2. The digital mapping method based on polygon modeling according to claim 1, characterized in that, Construct a visible station set for hole H. Afterwards, if If empty, the hole is marked as a hard occlusion type and the process is switched to geometric interpolation based on boundary curvature, only if... When not empty, the hole enters the multi-view fusion photometric gradient field calculation process.

3. The digital mapping method based on polygon modeling according to claim 1, characterized in that, Each scanning station in the scanning station pose sequence is defined by three-dimensional center coordinates. and the main view unit vector Composition, ray-triangle intersection test targets along the direction The emitted detection rays are executed one by one with all the triangulated mesh faces.

4. The digital mapping method based on polygon modeling according to claim 1, characterized in that, The patch set Pieces in the middle according to confidence level The results are divided into three levels: patches with a confidence level of not less than 0.7 are marked as high confidence level and incorporated into the main mapping result grid; patches with a confidence level between 0.3 and 0.7 are marked as unverified level, stored in an independent layer and a verification task work order is generated for the corresponding location. Patches with a confidence level below 0.3 are marked as low confidence level, retained only for visualization reference, and not included in any measurement statistics.

5. The digital mapping method based on polygon modeling according to claim 1, characterized in that, The normalization constant It is determined by the mean of the variance of the gradient magnitude obtained from the multi-view fusion photometric gradient field calculation of all holes in the entire dataset.

6. The digital mapping method based on polygon modeling according to claim 4, characterized in that, When low-confidence patches form continuous regions in the spatial distribution, the system generates a supplementary acquisition suggestion report based on the spatial coordinate distribution and normal vector orientation of the low-confidence patches. After accessing the supplementary scanning data, the system re-executes the complete process of visible station set construction, multi-view fusion photometric gradient field calculation, constrained Delaunay triangulation, and confidence calculation for the corresponding holes.

7. The digital mapping method based on polygon modeling according to claim 1, characterized in that, The surveying platform is a vehicle-mounted mobile surveying system. The scanning station pose sequence consists of multiple pose nodes continuously recorded during the vehicle's movement. The synchronous camera is installed on the vehicle and faces at least two different directions. The camera's internal and external parameters are calibrated before leaving the warehouse.

8. The digital mapping method based on polygon modeling according to claim 1, characterized in that, The surveying platform is a ground tripod-mounted static ground lidar. The pose sequence of the scanning stations consists of the poses of multiple independent scanning stations deployed in the survey area. Each scanning station is equipped with a fisheye camera. After scanning, the fisheye camera synchronously acquires panoramic images. The joint calibration residual between the fisheye camera and the lidar coordinate system does not exceed 0.3 pixels.

Images