Intelligent three-dimensional terrain modeling method and system based on surveying data

Abstract

This invention discloses an intelligent 3D terrain modeling method and system based on surveying and mapping data, relating to the field of surveying and mapping modeling technology. The method includes the following steps: using a drone equipped with high-definition cameras from multiple perspectives to capture terrain images; obtaining 3D point clouds of the terrain from the terrain images; acquiring the average density of the 3D point clouds; capturing suspicious areas in the 3D point clouds based on the average density of the 3D point clouds; capturing approximate point clouds of void areas in non-void areas using the point clouds from the terrain images, thereby determining whether cloud shadows exist in non-void areas. If cloud shadows similar to void areas exist, it indicates that the void areas are formed by cloud cover. Since light reflection has a specific angle, comparing images taken from different perspectives on the drone can determine whether there are changes in void areas, thus facilitating intelligent automated void repair judgment in 3D terrain modeling.

Description

Technical Field

[0001] This invention relates to the field of surveying and modeling technology, specifically to an intelligent three-dimensional terrain modeling method and system based on surveying data. Background Technology

[0002] 3D terrain modeling based on surveying data refers to the use of modern surveying technologies such as aerial photogrammetry, LiDAR, and satellite remote sensing to acquire high-precision geometric and image data of the Earth's surface. Through computer processing, a realistic, measurable, and analyzable digital 3D terrain model is reconstructed. The core of this technology lies in the intelligent processing of discrete, massive point cloud and image data, including steps such as dense matching, triangulation construction, and texture mapping, ultimately generating a structured model containing rich semantic information. It not only realizes an intuitive and realistic representation of terrain features from two-dimensional to three-dimensional, but also provides a powerful decision support and analysis platform for fields such as smart cities, land planning, disaster emergency response, and military simulation through deep integration with Geographic Information Systems (GIS) and simulation technology. It is the cornerstone for building a digital twin world.

[0003] Currently, 3D terrain modeling relies primarily on aerial photogrammetry for data acquisition. However, aerial photogrammetry also has its limitations. When converting images into 3D point clouds, holes often appear. There are two main reasons for this: First, when the drone was positioned above the clouds during image capture, some terrain features were obscured by clouds or other objects, resulting in missing points and holes. Second, due to image perspective and lighting conditions, the hole areas may be snowfields, water surfaces, or reflective rooftops, leading to insufficient image feature coverage and thus point cloud holes. Although both result in holes, their different causes affect the modeling process. The solutions for cavity repair also differ. Cavities caused by cloud cover can be filled in by using cameras from other perspectives on a drone to fill in the missing images. Cavities caused by insufficient features due to reflection can be repaired by lowering the matching threshold to accept matching points with lower confidence, forming a new point cloud with higher density. Therefore, it is crucial to capture cavity areas and analyze the causes of cavities after 3D terrain modeling. However, this part still relies heavily on manual judgment. As a result, there is a lack of intelligent and automated judgment in the process of 3D terrain modeling and post-modeling upgrade repair. To address this, we propose an intelligent 3D terrain modeling method and system based on surveying and mapping data. Summary of the Invention

[0004] (a) Technical problems to be solved

[0005] To address the shortcomings of existing technologies, this invention provides an intelligent three-dimensional terrain modeling method and system based on surveying and mapping data, thereby solving the aforementioned problems in existing technologies.

[0006] (II) Technical Solution

[0007] To achieve the above objectives, the present invention provides the following technical solution: an intelligent three-dimensional terrain modeling method and system based on surveying and mapping data, comprising the following steps:

[0008] S1: The terrain is captured by a drone equipped with a high-definition camera with multiple perspectives to obtain terrain images. The terrain images are used to obtain a 3D point cloud of the terrain. The average density of the 3D point cloud is obtained. Suspicious areas are captured in the 3D point cloud of the terrain based on the average density of the 3D point cloud.

[0009] S2: Based on the terrain images of the suspicious area, obtain the depth map of the terrain images of the entire suspicious area from all perspectives using a stereo vision matching algorithm. Obtain reference points in the 3D point cloud of the suspicious area. Obtain the depth map value of the reference points reprojected onto the camera of each perspective using the depth map of the terrain images of the suspicious area from all perspectives. Obtain the straight-line distance from the reference point to each perspective camera using the world coordinates of the reference point and mark it as the true distance of the reference point. Determine whether the suspicious area is a hollow area by comparing the depth map value of the entire perspective with the true distance of the reference point of the corresponding perspective.

[0010] S3: Obtain the cloud shadow matching degree of the cavity area through the terrain image, obtain the two-dimensional point cloud of the terrain image of all views, obtain the reference point matching degree by the reference point and the two-dimensional point cloud of the terrain image of all views, and determine whether the cavity area is formed by reflection or occlusion by the cloud shadow matching degree of the cavity area and the reference point matching degree.

[0011] Preferably, in S1, the average density of the three-dimensional point cloud of the terrain is obtained, specifically as follows:

[0012] S101: Project the entire 3D point cloud vertically onto a plane to obtain a projection surface, and divide the projection surface into N equal squares;

[0013] S102: Statistically analyze the point cloud in each of the N squares, obtain the area of ​​each square, and obtain the single-cell density by dividing the number of point clouds in each square by the square area. Then, sum all the single-cell densities and take the average to obtain the average three-dimensional point cloud density.

[0014] Preferably, in S1, suspicious areas are captured in the terrain 3D point cloud based on the average density of the 3D point cloud, specifically as follows:

[0015] S103: Obtain the average density of three-dimensional point cloud and set a preset threshold for the percentage of point cloud density in suspicious areas;

[0016] S104: Obtain the extreme value of the number of point clouds in the suspicious area by multiplying the preset threshold of the point cloud density percentage of the suspicious area by the mean density of the three-dimensional point cloud. Set a single capture spherical space and capture areas in the three-dimensional point cloud with the single capture spherical space as the capture range. Capture areas with point cloud density lower than the extreme value of the number of point clouds in the suspicious area and mark them as suspicious areas.

[0017] Preferably, in S2, a reference point is obtained from the 3D point cloud of the suspected area. Specifically, within the single spherical capture range of the suspected area, the farthest straight-line distance between each point cloud and other point clouds is obtained. The two point clouds with the farthest straight-line distance are marked as candidate point clouds. The farthest distance between the two candidate point clouds and other point clouds is obtained from the other point clouds outside the two candidate point clouds. The candidate point cloud with the greater distance from the two candidate point clouds and other point clouds is selected as the reference point.

[0018] Preferably, in S2, the straight-line distance from the reference point to each viewpoint camera is obtained using the world coordinates of the reference point and marked as the true distance of the reference point, specifically:

[0019] S201: Obtain the world coordinates of the reference point, obtain the camera's rotation matrix, obtain the camera's plane vector, and establish a camera coordinate system with the camera's optical center position as the origin;

[0020] S202: The coordinates of the reference point in the camera coordinate system are obtained by using the collinearity equation through the world coordinates of the reference point, the rotation matrix of the camera, and the plane vector of the camera. The true distance of the reference point is obtained by using the Euclidean distance formula between the coordinates of the reference point in the camera coordinate system and the coordinates of the optical center of the camera in the camera coordinate system.

[0021] Preferably, in S2, the questionable area is determined to be a hole area by comparing the true distances between all viewpoint depth map values ​​and the corresponding viewpoint reference points. Specifically:

[0022] S203: Obtain the depth map values ​​of all viewpoint reference points, obtain the true distance of all viewpoint reference points, and obtain the error distance of the reference point by taking the absolute value of the difference between the true distance of the reference point and the depth map value of the corresponding viewpoint reference point.

[0023] S204: Set a preset threshold for error distance, and determine whether the error distance of each view reference point is less than the preset threshold for error distance. If the error distance of each view reference point is less than the preset threshold for error distance, mark the suspicious area as a non-hollow area. If there is one or more view reference points whose error distance is greater than or equal to the preset threshold for error distance, mark the suspicious area as a hollow area.

[0024] Preferably, in S3, the cloud shadow matching degree of the cavity area is obtained through terrain imagery, specifically as follows:

[0025] S301: Obtain the projection surface of the terrain image, mark the point cloud in all the hollow areas in the terrain image projection surface to obtain the marked point cloud, and delete the point cloud in all the hollow areas in the terrain image projection surface to obtain the comparison projection surface.

[0026] S302: Capture point clouds that are similar in shape to the marked point cloud but different in scale in the comparison projection plane to obtain multiple suspicious point clouds, and enlarge the suspicious point clouds proportionally to the same size as the marked point cloud;

[0027] S303: Count the number of points in the marked point cloud to obtain the total number of points, count the points in the suspicious point cloud that overlap with the marked point cloud to obtain the number of similar points, and obtain the cloud shadow matching degree of the hole area by dividing the number of similar points by the total number of points.

[0028] Preferably, in S3, a two-dimensional point cloud of terrain images from all viewpoints is acquired, and the matching degree of the reference point is obtained by comparing the reference point with the two-dimensional point cloud of terrain images from all viewpoints. Specifically:

[0029] S304: Obtain the two-dimensional point cloud of all viewpoint terrain images, and determine whether there is a reference point in the two-dimensional point cloud of each viewpoint terrain image. If there is a reference point in the two-dimensional point cloud of the viewpoint terrain image, mark the two-dimensional point cloud of the viewpoint terrain image as having a viewpoint. If there is no reference point in the two-dimensional point cloud of the viewpoint terrain image, mark the two-dimensional point cloud of the viewpoint terrain image as not having a viewpoint.

[0030] S305: Sum the number of topographic images with existing viewpoints to obtain the total number of existing viewpoints, obtain the total number of viewpoints in all topographic images, and obtain the reference point matching degree by dividing the total number of existing viewpoints by the total number of viewpoints in all topographic images.

[0031] Preferably, in S3, the determination of whether a hole region is formed by reflection or occlusion is based on the matching degree of cloud shadow in the hole region and the matching degree of the reference point. Specifically:

[0032] S306: Obtain the cloud shadow matching degree of the hole area, set a preset threshold for the cloud shadow matching degree of the hole area, and determine whether the cloud shadow matching degree of the hole area is higher than the preset threshold for the cloud shadow matching degree of the hole area. If the cloud shadow matching degree of the hole area is higher than the preset threshold for the cloud shadow matching degree of the hole area, the determination result is that occlusion is formed. If the cloud shadow matching degree of the hole area is lower than or equal to the preset threshold for the cloud shadow matching degree of the hole area, then execute S307.

[0033] S307: Obtain the reference point matching degree, set the reference point matching degree preset threshold, and determine whether the reference point matching degree is lower than the reference point matching degree preset threshold. If the reference point matching degree is lower than the reference point matching degree preset threshold, the result is occlusion formation. If the reference point matching degree is higher than or equal to the reference point matching degree preset threshold, the result is reflection formation.

[0034] The intelligent 3D terrain modeling system based on surveying and mapping data includes the following modules:

[0035] The suspicious area capture module uses a drone equipped with high-definition cameras from multiple perspectives to capture terrain images, obtains a 3D point cloud of the terrain from the terrain images, acquires the average density of the 3D point cloud, and captures suspicious areas in the 3D point cloud based on the average density of the 3D point cloud.

[0036] The cavity region locking module obtains depth maps of all viewpoint terrain images of the suspicious region based on the terrain images of the suspicious region using a stereo vision matching algorithm. It obtains reference points in the 3D point cloud of the suspicious region and obtains the depth map values ​​of the reference points reprojected onto the camera of each viewpoint using the depth maps of the terrain images of the suspicious region from all viewpoints. It obtains the straight-line distance from the reference point to each viewpoint camera using the world coordinates of the reference point and marks it as the true distance of the reference point. It determines whether the suspicious region is a cavity region by comparing the depth map values ​​of all viewpoints with the true distance of the reference point of the corresponding viewpoint.

[0037] The cavity formation determination module obtains the cloud shadow matching degree of the cavity area through topographic images, obtains the two-dimensional point cloud of the topographic images from all perspectives, obtains the reference point matching degree by comparing the reference point with the two-dimensional point cloud of the topographic images from all perspectives, and determines whether the cavity area is formed by reflection or occlusion by combining the cloud shadow matching degree of the cavity area and the reference point matching degree.

[0038] (III) Beneficial Effects

[0039] This invention provides an intelligent 3D terrain modeling method and system based on surveying and mapping data, which has the following beneficial effects:

[0040] In this scheme, a depth map is generated from the terrain image, which facilitates the determination of the distance from the point cloud in the model to the camera. By setting and selecting a reference point, the world coordinates of the reference point and the camera coordinates are calculated to obtain the true distance from the reference point to the camera. The depth map value is then compared with the true distance to determine whether the error of the reference point is small enough. A sufficiently small error can achieve confidence, while a large error of the reference point results in low confidence. Suspicious areas already have a sparse point cloud. If the low confidence results in invalid point clouds, the actual point cloud in the suspicious area will be even sparser. Therefore, the suspicious area can be identified as a hole area, which facilitates subsequent analysis of the cause of the hole area.

[0041] In this solution, point clouds from terrain images are used to capture approximate point clouds of cavity areas in non-cavity areas, thereby determining whether cloud shadows exist in non-cavity areas. If cloud shadows similar to those of cavity areas exist, it indicates that the cavity areas are formed by cloud cover. Since light reflection has a specific angle, comparing images taken from different perspectives on the drone can determine whether there are changes in the cavity areas. By capturing reference points from images at different perspectives, it can be determined whether the cavity areas are formed due to reflection, thus facilitating intelligent automated cavity repair judgment in 3D terrain modeling. Attached Figure Description

[0042] Figure 1 This is a flowchart of the intelligent 3D terrain modeling method based on surveying data according to the present invention;

[0043] Figure 2 This is a schematic diagram of the module structure of the intelligent three-dimensional terrain modeling system based on surveying data according to the present invention. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] Please see Figures 1-2 This invention provides an intelligent 3D terrain modeling method and system based on surveying and mapping data, comprising the following steps:

[0046] S1: The terrain is captured by a drone equipped with a high-definition camera with multiple perspectives to obtain terrain images. The terrain images are used to obtain a 3D point cloud of the terrain. The average density of the 3D point cloud is obtained. Suspicious areas are captured in the 3D point cloud of the terrain based on the average density of the 3D point cloud.

[0047] S2: Based on the terrain images of the suspicious area, obtain the depth map of the terrain images of the entire suspicious area from all perspectives using a stereo vision matching algorithm. Obtain reference points in the 3D point cloud of the suspicious area. Obtain the depth map value of the reference points reprojected onto the camera of each perspective using the depth map of the terrain images of the suspicious area from all perspectives. Obtain the straight-line distance from the reference point to each perspective camera using the world coordinates of the reference point and mark it as the true distance of the reference point. Determine whether the suspicious area is a hollow area by comparing the depth map value of the entire perspective with the true distance of the reference point of the corresponding perspective.

[0048] S3: Obtain the cloud shadow matching degree of the cavity area through the terrain image, obtain the two-dimensional point cloud of the terrain image of all views, obtain the reference point matching degree by the reference point and the two-dimensional point cloud of the terrain image of all views, and determine whether the cavity area is formed by reflection or occlusion by the cloud shadow matching degree of the cavity area and the reference point matching degree.

[0049] In this embodiment, the terrain is photographed from the air by a drone equipped with cameras from multiple perspectives. Then, a three-dimensional point cloud is drawn based on the impact of the multi-perspective photography. Based on the density of the three-dimensional point cloud, areas with sparse density of the three-dimensional point cloud are initially marked as suspicious areas, which facilitates subsequent accurate judgment of suspicious areas, thereby reducing the judgment range and saving computing power.

[0050] In this scheme, a depth map is generated from the terrain image, which facilitates the determination of the distance from the point cloud in the model to the camera. By setting and selecting a reference point, the world coordinates of the reference point and the camera coordinates are calculated to obtain the true distance from the reference point to the camera. The depth map value is then compared with the true distance to determine whether the error of the reference point is small enough. A sufficiently small error can provide confidence, while a large error of the reference point results in low confidence. Suspicious areas already have a sparse point cloud. If the low confidence results in invalid point clouds, the actual point cloud in the suspicious area will be even sparser. Therefore, the suspicious area can be identified as a hole area, which facilitates the subsequent analysis of the cause of the hole area.

[0051] In this solution, the point cloud of the terrain image is used to capture the approximate point cloud of the cavity area in the non-cavity area, thereby determining whether there is a cloud shadow in the non-cavity area. If there is a cloud shadow similar to the cavity area, it indicates that the cavity area is formed by cloud cover. Since the reflection of light has a specific angle, the cavity area can be determined by comparing the images taken from different perspectives on the drone. The capture of the reference point by the images from different perspectives can be used to determine whether the cavity area is formed by reflection. Therefore, it is convenient to automatically determine the cavity repair in the intelligent 3D terrain model.

[0052] It is worth mentioning that in this scheme, multiple reference points can be selected for multiple comparisons as needed. The more reference points with lower confidence, the more likely the suspicious area is a real cavity area. The more reference points with higher confidence, the more likely the suspicious area is not a real cavity area, thus increasing the accuracy of the judgment results. The more independent reference points from different perspectives, the greater the probability of reflection. If there are a few independent reference points, it indicates that individual cameras are blocked by clouds or the reference points are blocked by the view of tall buildings, thus increasing the accuracy of the judgment results.

[0053] It is worth mentioning that the value of the preset threshold in this scheme can be obtained through weight analysis, which will not be elaborated on here.

[0054] In S1, the average density of the 3D point cloud of the terrain is obtained, specifically as follows:

[0055] S101: Project the entire 3D point cloud vertically onto a plane to obtain a projection surface, and divide the projection surface into N equal squares;

[0056] S102: Statistically analyze the point cloud in each of the N squares, obtain the area of ​​each square, and obtain the single-cell density by dividing the number of point clouds in each square by the square area. Then, sum all the single-cell densities and take the average to obtain the average three-dimensional point cloud density.

[0057] In S1, suspicious areas are captured in the 3D point cloud based on the average density of the 3D point cloud, specifically as follows:

[0058] S103: Obtain the average density of three-dimensional point cloud and set a preset threshold for the percentage of point cloud density in suspicious areas;

[0059] S104: Obtain the extreme value of the number of point clouds in the suspicious area by multiplying the preset threshold of the point cloud density percentage of the suspicious area by the mean density of the three-dimensional point cloud. Set a single capture spherical space and capture areas in the three-dimensional point cloud with the single capture spherical space as the capture range. Capture areas with point cloud density lower than the extreme value of the number of point clouds in the suspicious area and mark them as suspicious areas.

[0060] In this embodiment, regions with sparse density in the 3D point cloud are initially marked as suspicious areas based on the density of the 3D point cloud. This facilitates a second, more accurate judgment of the suspicious areas, thereby reducing the judgment range and saving computing power.

[0061] In S2, a reference point is obtained from the 3D point cloud of the suspicious area. Specifically, within the single spherical capture range of the suspicious area, the farthest straight-line distance between each point cloud and other point clouds is obtained. The two point clouds with the farthest straight-line distance are marked as candidate point clouds. Among the other point clouds outside the two candidate point clouds, the farthest distance between the two candidate point clouds and other point clouds is obtained. Among the farthest distances between the two candidate point clouds and other point clouds, the candidate point cloud with the greater distance is selected as the reference point.

[0062] In this embodiment, multiple reference points can be selected for multiple comparisons as needed. The more reference points with lower confidence, the more likely the suspicious area is a real void area; the more reference points with higher confidence, the more likely the suspicious area is not a real void area. This helps to increase the accuracy of the judgment results. The more independent reference points there are from different perspectives, the greater the probability of reflection as the cause. If there are a few independent reference points, it indicates that individual cameras are blocked by clouds or that the reference points are blocked by the view of tall buildings. This helps to increase the accuracy of the judgment results.

[0063] In S2, the straight-line distance from the reference point to each viewpoint camera is obtained using the world coordinates of the reference point and marked as the true distance of the reference point. Specifically:

[0064] S201: Obtain the world coordinates (X, Y, Z) of the reference point, obtain the camera's rotation matrix R, obtain the camera's plane vector T, and establish a camera coordinate system with the camera's optical center position as the origin;

[0065] S202: Obtain the coordinates of the reference point in the camera coordinate system (X×R+T, Y×R+T, Z×R+T) by multiplying the world coordinates of the reference point with the rotation matrix of the camera and then summing the result with the plane vector of the camera. Then, use the Euclidean distance formula to obtain the true distance of the reference point by multiplying the world coordinates of the reference point in the camera coordinate system with the optical center coordinates of the camera in the camera coordinate system.

[0066] In this embodiment, the coordinates of the reference point in the camera coordinate system are calculated by collinearity equation, and the distance from the reference point to the camera optical center is calculated by Euclidean distance formula. This makes it easier to obtain the true distance of the reference point, which in turn makes it easier to compare the true distance of the reference point with the depth map value, thereby making it easier to judge the matching degree of the reference point and also to judge the confidence degree of the reference point in the hole area.

[0067] In S2, the questionable area is determined to be a hole area by comparing the true distances between all viewpoint depth map values ​​and the corresponding viewpoint reference points. Specifically:

[0068] S203: Obtain the depth map values ​​of all viewpoint reference points, obtain the true distance of all viewpoint reference points, and obtain the error distance of the reference point by taking the absolute value of the difference between the true distance of the reference point and the depth map value of the corresponding viewpoint reference point.

[0069] S204: Set a preset threshold for error distance, and determine whether the error distance of each view reference point is less than the preset threshold for error distance. If the error distance of each view reference point is less than the preset threshold for error distance, mark the suspicious area as a non-hollow area. If there is one or more view reference points whose error distance is greater than or equal to the preset threshold for error distance, mark the suspicious area as a hollow area.

[0070] In this embodiment, the true distance from the reference point to the camera is calculated by using the world coordinates of the reference point and the camera coordinates. The depth map value is then compared with the true distance to determine whether the error of the reference point is small enough. A sufficiently small error can be considered as having confidence, while a large error of the reference point results in low confidence. Suspicious areas already have a sparse point cloud. If the low confidence results in invalid point clouds, the actual point cloud in the suspicious area will be even sparser. Therefore, the suspicious area can be identified as a hole area, which facilitates subsequent analysis of the cause of the hole area.

[0071] In S3, the cloud shadow matching degree of the hollow area is obtained through terrain imagery, specifically as follows:

[0072] S301: Obtain the projection surface of the terrain image, mark the point cloud in all the hollow areas in the terrain image projection surface to obtain the marked point cloud, and delete the point cloud in all the hollow areas in the terrain image projection surface to obtain the comparison projection surface.

[0073] S302: Capture point clouds that are similar in shape to the marked point cloud but different in scale in the comparison projection plane to obtain multiple suspicious point clouds, and enlarge the suspicious point clouds proportionally to the same size as the marked point cloud;

[0074] S303: Count the number of points in the marked point cloud to obtain the total number of points, count the points in the suspicious point cloud that overlap with the marked point cloud to obtain the number of similar points, and obtain the cloud shadow matching degree of the hole area by dividing the number of similar points by the total number of points.

[0075] In this embodiment, the point cloud of the terrain image captures the approximate point cloud of the cavity area in the non-cavity area, thereby determining whether there is a cloud shadow in the non-cavity area. If there is a cloud shadow similar to the cavity area, it indicates that the cavity area is formed by cloud cover.

[0076] In S3, two-dimensional point clouds of terrain images from all viewpoints are acquired. The matching degree of the reference point is obtained by comparing the reference point with the two-dimensional point clouds of terrain images from all viewpoints. Specifically:

[0077] S304: Obtain the two-dimensional point cloud of all viewpoint terrain images, and determine whether there is a reference point in the two-dimensional point cloud of each viewpoint terrain image. If there is a reference point in the two-dimensional point cloud of the viewpoint terrain image, mark the two-dimensional point cloud of the viewpoint terrain image as having a viewpoint. If there is no reference point in the two-dimensional point cloud of the viewpoint terrain image, mark the two-dimensional point cloud of the viewpoint terrain image as not having a viewpoint.

[0078] S305: Sum the number of topographic images with existing viewpoints to obtain the total number of existing viewpoints, obtain the total number of viewpoints in all topographic images, and obtain the reference point matching degree by dividing the total number of existing viewpoints by the total number of viewpoints in all topographic images.

[0079] In this embodiment, since light reflection has a specific angle, it is possible to determine whether there is a change in the cavity area by comparing images taken from different perspectives on the drone. By judging whether the cavity area is formed due to reflection by capturing the reference point from different perspective images, it is easy to automatically determine the cavity repair in the three-dimensional terrain model.

[0080] In S3, the matching degree of cloud shadow in the hole area and the matching degree of the reference point are used to determine whether the hole area is formed by reflection or occlusion. Specifically:

[0081] S306: Obtain the cloud shadow matching degree of the hole area, set a preset threshold for the cloud shadow matching degree of the hole area, and determine whether the cloud shadow matching degree of the hole area is higher than the preset threshold for the cloud shadow matching degree of the hole area. If the cloud shadow matching degree of the hole area is higher than the preset threshold for the cloud shadow matching degree of the hole area, the determination result is that occlusion is formed. If the cloud shadow matching degree of the hole area is lower than or equal to the preset threshold for the cloud shadow matching degree of the hole area, then execute S307.

[0082] S307: Obtain the reference point matching degree, set the reference point matching degree preset threshold, and determine whether the reference point matching degree is lower than the reference point matching degree preset threshold. If the reference point matching degree is lower than the reference point matching degree preset threshold, the result is occlusion formation. If the reference point matching degree is higher than or equal to the reference point matching degree preset threshold, the result is reflection formation.

[0083] In this embodiment, the point cloud of the terrain image captures the approximate point cloud of the cavity area in the non-cavity area, thereby determining whether there is a cloud shadow in the non-cavity area. If there is a cloud shadow similar to the cavity area, it indicates that the cavity area is formed by cloud cover. Since light reflection has a specific angle, it can be determined whether there is a change in the cavity area by comparing images taken from different perspectives on the drone. The capture of the reference point by images from different perspectives determines whether the cavity area is formed by reflection, thus facilitating intelligent automated cavity repair judgment for 3D terrain modeling.

[0084] Please see Figures 1-2 This invention provides an intelligent 3D terrain modeling method and system based on surveying and mapping data, comprising the following modules:

[0085] The suspicious area capture module uses a drone equipped with high-definition cameras from multiple perspectives to capture terrain images, obtains a 3D point cloud of the terrain from the terrain images, acquires the average density of the 3D point cloud, and captures suspicious areas in the 3D point cloud based on the average density of the 3D point cloud.

[0086] The cavity region locking module obtains depth maps of all viewpoint terrain images of the suspicious region based on the terrain images of the suspicious region using a stereo vision matching algorithm. It obtains reference points in the 3D point cloud of the suspicious region and obtains the depth map values ​​of the reference points reprojected onto the camera of each viewpoint using the depth maps of the terrain images of the suspicious region from all viewpoints. It obtains the straight-line distance from the reference point to each viewpoint camera using the world coordinates of the reference point and marks it as the true distance of the reference point. It determines whether the suspicious region is a cavity region by comparing the depth map values ​​of all viewpoints with the true distance of the reference point of the corresponding viewpoint.

[0087] The cavity formation determination module obtains the cloud shadow matching degree of the cavity area through terrain imagery, acquires the two-dimensional point cloud of terrain imagery from all perspectives, obtains the reference point matching degree by comparing the reference point with the two-dimensional point cloud of terrain imagery from all perspectives, and determines whether the cavity area is formed by reflection or occlusion by combining the cloud shadow matching degree of the cavity area and the reference point matching degree.

[0088] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0089] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0090] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. An intelligent 3D terrain modeling method based on surveying and mapping data, characterized in that, Includes the following steps: S1: The terrain is captured by a drone equipped with a high-definition camera with multiple perspectives to obtain terrain images. The terrain images are used to obtain a 3D point cloud of the terrain. The average density of the 3D point cloud is obtained. Suspicious areas are captured in the 3D point cloud of the terrain based on the average density of the 3D point cloud. S2: Based on the terrain images of the suspicious area, obtain the depth map of the terrain images of the entire suspicious area from all perspectives using a stereo vision matching algorithm. Obtain reference points in the 3D point cloud of the suspicious area. Obtain the depth map value of the reference points reprojected onto the camera of each perspective using the depth map of the terrain images of the suspicious area from all perspectives. Obtain the straight-line distance from the reference point to each perspective camera using the world coordinates of the reference point and mark it as the true distance of the reference point. Determine whether the suspicious area is a hollow area by comparing the depth map value of the entire perspective with the true distance of the reference point of the corresponding perspective. S3: Obtain the cloud shadow matching degree of the cavity area through the terrain image, obtain the two-dimensional point cloud of the terrain image of all views, obtain the reference point matching degree by the reference point and the two-dimensional point cloud of the terrain image of all views, and determine whether the cavity area is formed by reflection or occlusion by the cloud shadow matching degree of the cavity area and the reference point matching degree. In S1, the average density of the 3D point cloud of the terrain is obtained, specifically as follows: S101: Project the entire 3D point cloud vertically onto a plane to obtain a projection surface, and divide the projection surface into N equal squares; S102: Statistically analyze the point cloud in each of the N squares, obtain the area of ​​each square, and obtain the single-cell density by dividing the number of points in each square by the square area. Then, sum all the single-cell densities and take the average to obtain the average three-dimensional point cloud density. In S3, the cloud shadow matching degree of the hollow area is obtained through terrain imagery, specifically as follows: S301: Obtain the projection surface of the terrain image, mark the point cloud in all the hollow areas in the terrain image projection surface to obtain the marked point cloud, and delete the point cloud in all the hollow areas in the terrain image projection surface to obtain the comparison projection surface. S302: Capture point clouds that are similar in shape to the marked point cloud but different in scale in the comparison projection plane to obtain multiple suspicious point clouds, and enlarge the suspicious point clouds proportionally to the same size as the marked point cloud; S303: Count the number of points in the marked point cloud to obtain the total number of points, count the points in the suspicious point cloud that overlap with the marked point cloud to obtain the number of similar points, and obtain the cloud shadow matching degree of the hole area by dividing the number of similar points by the total number of points. In S3, two-dimensional point clouds of terrain images from all viewpoints are acquired. The matching degree of the reference point is obtained by comparing the reference point with the two-dimensional point clouds of terrain images from all viewpoints. Specifically: S304: Obtain the two-dimensional point cloud of all viewpoint terrain images, and determine whether there is a reference point in the two-dimensional point cloud of each viewpoint terrain image. If there is a reference point in the two-dimensional point cloud of the viewpoint terrain image, mark the two-dimensional point cloud of the viewpoint terrain image as having a viewpoint. If there is no reference point in the two-dimensional point cloud of the viewpoint terrain image, mark the two-dimensional point cloud of the viewpoint terrain image as not having a viewpoint. S305: Sum the number of topographic images with existing viewpoints to obtain the total number of existing viewpoints, obtain the total number of viewpoints in all topographic images, and obtain the reference point matching degree by dividing the total number of existing viewpoints by the total number of viewpoints in all topographic images.

2. The intelligent 3D terrain modeling method based on mapping data according to claim 1, characterized in that: In S1, suspicious areas are captured in the 3D terrain point cloud based on the average density of the 3D point cloud, specifically as follows: S103: Obtain the average density of three-dimensional point cloud and set a preset threshold for the percentage of point cloud density in suspicious areas; S104: Obtain the extreme value of the number of point clouds in the suspicious area by multiplying the preset threshold of the point cloud density percentage of the suspicious area by the mean density of the three-dimensional point cloud. Set a single capture spherical space and capture areas in the three-dimensional point cloud with the single capture spherical space as the capture range. Capture areas with point cloud density lower than the extreme value of the number of point clouds in the suspicious area and mark them as suspicious areas.

3. The intelligent 3D terrain modeling method based on mapping data according to claim 1, characterized in that: In S2, a reference point is obtained from the 3D point cloud of the suspicious area. Specifically, within the single spherical capture range of the suspicious area, the farthest straight-line distance between each point cloud and other point clouds is obtained. The two point clouds with the farthest straight-line distance are marked as candidate point clouds. Among the other point clouds outside the two candidate point clouds, the farthest distance between the two candidate point clouds and other point clouds is obtained. Among the farthest distances between the two candidate point clouds and other point clouds, the candidate point cloud with the greater distance is selected as the reference point.

4. The intelligent 3D terrain modeling method based on mapping data according to claim 1, characterized in that: In S2, the straight-line distance from the reference point to each viewpoint camera is obtained using the world coordinates of the reference point and marked as the true distance of the reference point. Specifically: S201: Obtain the world coordinates of the reference point, obtain the camera's rotation matrix, obtain the camera's plane vector, and establish a camera coordinate system with the camera's optical center position as the origin; S202: The coordinates of the reference point in the camera coordinate system are obtained by using the collinearity equation through the world coordinates of the reference point, the rotation matrix of the camera, and the plane vector of the camera. The true distance of the reference point is obtained by using the Euclidean distance formula between the coordinates of the reference point in the camera coordinate system and the coordinates of the optical center of the camera in the camera coordinate system.

5. The intelligent 3D terrain modeling method based on mapping data according to claim 1, wherein: In S2, the questionable area is determined to be a hole area by comparing the true distances between all viewpoint depth map values ​​and the corresponding viewpoint reference points. Specifically: S203: Obtain the depth map values ​​of all viewpoint reference points, obtain the true distance of all viewpoint reference points, and obtain the error distance of the reference point by taking the absolute value of the difference between the true distance of the reference point and the depth map value of the corresponding viewpoint reference point. S204: Set a preset threshold for error distance, and determine whether the error distance of each view reference point is less than the preset threshold for error distance. If the error distance of each view reference point is less than the preset threshold for error distance, mark the suspicious area as a non-hollow area. If there is one or more view reference points whose error distance is greater than or equal to the preset threshold for error distance, mark the suspicious area as a hollow area.

6. The intelligent 3D terrain modeling method based on mapping data according to claim 1, wherein: In S3, the matching degree of cloud shadow in the hole area and the matching degree of the reference point are used to determine whether the hole area is formed by reflection or occlusion. Specifically: S306: Obtain the cloud shadow matching degree of the hole area, set a preset threshold for the cloud shadow matching degree of the hole area, and determine whether the cloud shadow matching degree of the hole area is higher than the preset threshold for the cloud shadow matching degree of the hole area. If the cloud shadow matching degree of the hole area is higher than the preset threshold for the cloud shadow matching degree of the hole area, the determination result is that occlusion is formed. If the cloud shadow matching degree of the hole area is lower than or equal to the preset threshold for the cloud shadow matching degree of the hole area, then execute S307. S307: Obtain the reference point matching degree, set the reference point matching degree preset threshold, and determine whether the reference point matching degree is lower than the reference point matching degree preset threshold. If the reference point matching degree is lower than the reference point matching degree preset threshold, the result is occlusion formation. If the reference point matching degree is higher than or equal to the reference point matching degree preset threshold, the result is reflection formation.

7. An intelligent 3D terrain modeling system based on surveying data, applied to the intelligent 3D terrain modeling method based on surveying data as described in any one of claims 1-6, characterized in that, Includes the following modules: The suspicious area capture module uses a drone equipped with high-definition cameras from multiple perspectives to capture terrain images, obtains a 3D point cloud of the terrain from the terrain images, acquires the average density of the 3D point cloud, and captures suspicious areas in the 3D point cloud based on the average density of the 3D point cloud. The cavity region locking module obtains depth maps of all viewpoint terrain images of the suspicious region based on the terrain images of the suspicious region using a stereo vision matching algorithm. It obtains reference points in the 3D point cloud of the suspicious region and obtains the depth map values ​​of the reference points reprojected onto the camera of each viewpoint using the depth maps of the terrain images of the suspicious region from all viewpoints. It obtains the straight-line distance from the reference point to each viewpoint camera using the world coordinates of the reference point and marks it as the true distance of the reference point. It determines whether the suspicious region is a cavity region by comparing the depth map values ​​of all viewpoints with the true distance of the reference point of the corresponding viewpoint. The cavity formation determination module obtains the cloud shadow matching degree of the cavity area through topographic images, obtains the two-dimensional point cloud of the topographic images from all perspectives, obtains the reference point matching degree by comparing the reference point with the two-dimensional point cloud of the topographic images from all perspectives, and determines whether the cavity area is formed by reflection or occlusion by combining the cloud shadow matching degree of the cavity area and the reference point matching degree.

Images