Forest TLS point cloud automatic registration method and system based on ground overlap search

Through the method based on ground overlap search, the diamond area and point-surface ICP algorithm is used to achieve rapid, accurate and automatic registration of forest TLS point clouds, solving the problem of low accuracy of TLS point cloud registration in complex forest environments, and improving the Automation and robustness.

CN118229743BActive Publication Date: 2024-11-19WUHAN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410192282.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-02-21
Publication Date
2024-11-19
Estimated Expiration
2044-02-21

AI Technical Summary

Technical Problem

The existing forest TLS point cloud automatic registration method has low accuracy in complex forest environments, there are challenges in reliance on tree attribute acquisition, and manual target recognition and external information acquisition are limited, making it difficult to achieve fast and accurate automatic registration.

Method used

Using ground overlap search method, through data preprocessing, diamond area construction and point-surface ICP algorithm, automatic registration of forest TLS point cloud is realized, and ground points are used as registration primitives to eliminate the tree attribute extraction process, which improves the Registration efficiency and accuracy.

Benefits of technology

Fast, accurate and automatic registration of TLS point clouds is realized in complex forest environments, suitable for a wide range of forest types, improves automation and robustness, reduces manual intervention and hardware dependence, and is suitable for complex forest environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118229743B_ABST
    Figure CN118229743B_ABST
Patent Text Reader

Abstract

The present invention belongs to the technical field of point cloud registration, and discloses a method and system for automatic registration of forest TLS point clouds based on ground overlapping search, the method comprising: data preprocessing, reducing data redundancy and ground point extraction; coarse registration, diamond region construction, establishment of diamond region correspondence, overlapping diamond region search and transformation parameter estimation; fine registration, common ground point screening and point-surface ICP algorithm registration. The present invention is dedicated to proposing a method for automatic registration of forest TLS point clouds based on ground overlapping search, so as to quickly, accurately and automatically realize the registration of different TLS scanning point clouds in the forest, and promote the application of TLS in forest surveys.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of point cloud registration, and in particular relates to a forest TLS point cloud automatic registration method and system based on ground overlapping search. Background Art

[0002] As one of the key contents of forest resource survey, accurate representation of forest three-dimensional (3D) structural information is of great significance for ecological research and decision-making in forest management. Therefore, accurately and completely obtaining the three-dimensional structural information of forests is an important issue that needs to be solved urgently. Due to its penetrating and anti-interference capabilities, LiDAR technology has unparalleled advantages over optical remote sensing technology in obtaining forest vertical structural information. Therefore, LiDAR-based forest mapping has been widely used in forest resource surveys and has been used to extract various structural parameters of forests.

[0003] TLS, also known as ground-based lidar, can acquire details in the surrounding 3D scene with millimeter-level accuracy and support fast and automatic reconstruction of forest stand structure. The TLS-based 3D data acquisition modes in forestry resource surveys mainly include single-scan and multi-scan modes. In single-scan mode, the scanner is usually located in the center of the sample plot and only obtains data on one side of the visible tree. However, due to the occlusion of trees, the single-scan mode usually has a low detection rate for forest measurement. Therefore, multi-scan mode acquisition is usually required to measure forest land. In multi-scan mode, the scanner measures the sample plot at different locations. Although the multi-scan mode seems to be the most accurate method for forest mapping, it faces a problem that cannot be ignored, that is, the registration automation between different scan point clouds is low.

[0004] As a key step in point cloud data processing, point cloud registration has always been a focus of research. Some point cloud automatic registration algorithms have been proposed in existing studies. Most of these automatic registration algorithms are based on the geometric features (points, lines, and surfaces) of the point cloud model surface. However, these point cloud automatic registration methods are usually applicable to engineering surfaces. Due to the low overlap rate of forest point clouds and the high geometric similarity of tree trunks, they are often not directly applicable to the automatic registration of forest point clouds. However, research on forest point cloud registration is relatively limited.

[0005] Current forest point cloud registration methods can be divided into three categories: registration methods that rely on artificial targets, registration methods that rely on additional information, and registration methods that only rely on point clouds.

[0006] The most common method for TLS multi-scan data registration uses artificial reference targets placed in the forest to establish correspondence between different scans. This type of method requires a lot of work to select appropriate locations in the sample plot to place artificial targets (such as target balls), and then use commercial software packages to manually or automatically detect the targets as references for registration. These methods are effective in simple forest scenes, but in complex forest environments, manual registration will face some difficulties due to vegetation occlusion and terrain changes. When the software cannot detect the target, target identification requires additional user intervention.

[0007] In addition, additional information can be relied upon for point cloud data registration, which uses data provided by sensors such as inclinometers and visual inertial systems (VIS) to measure the relative positions between different scans. In situations with high forest density and complex terrain, the acquisition and use of external features will be limited. For example, it is challenging to maintain the stability of the Inertial Measurement Unit (IMU) in uneven terrain, and the effectiveness of VIS will be affected by tree occlusion and sudden changes in terrain conditions.

[0008] Existing methods for automatic registration of multi-scan TLS forest point clouds mainly rely on tree attributes. However, it is challenging to accurately obtain attributes such as DBH, tree height, and tree position in situations with high tree density, complex terrain, and irregular tree growth. In addition, tree trunks and canopies sway in the wind, which in turn affects the reliability and accuracy of tree attributes obtained in different scans at different times.

[0009] The present invention is devoted to proposing a forest TLS point cloud automatic registration method based on ground overlapping search, so as to quickly and accurately realize the registration of different TLS scanning point clouds in the forest and promote the application of TLS in forest surveys. Summary of the invention

[0010] In view of the problems existing in the prior art, the present invention provides a forest TLS point cloud automatic registration method and system based on ground overlapping search.

[0011] The present invention is implemented as follows: a forest TLS point cloud automatic registration method based on ground overlap search, the method comprising:

[0012] S1: Data preprocessing, reducing data redundancy and ground point extraction;

[0013] S2: coarse registration, diamond region construction, diamond region correspondence establishment, overlapping diamond region search and transformation parameter estimation;

[0014] S3: Precise registration, common ground point screening and point-to-surface ICP algorithm registration.

[0015] Furthermore, in S1, reducing data redundancy specifically includes:

[0016] Calculate the horizontal distance from each point to the center of the scanner. If it is greater than a certain threshold, it will be eliminated. The threshold is usually set to 15m, and the user can set the threshold size according to different forest scenes.

[0017] The ground point extraction specifically includes:

[0018] The ground point cloud data is extracted by morphological filtering. All points in the point cloud are converted into a two-dimensional grid space; the lowest point in each pixel is selected as the seed point for clustering to obtain the ground point.

[0019] Further, in S2, the diamond region construction specifically includes:

[0020] In the XOY coordinate plane, a rhombus region with a side length of Rmax is constructed. One vertex of the rhombus is located at the origin O(0,0), and the polar angles of the two sides passing through the vertex are θrad and θ+2π / 3rad respectively. Points whose horizontal positions are within the rhombus region are extracted from the ground point cloud, and the horizontal distances from these points to the two sides are calculated using the following formula:

[0021]

[0022] In the formula, p = (x, y) represents the horizontal coordinates of the point, m = (cosθ, sinθ) and n = (cos(θ+2π / 3), sin(θ+2π / 3)) represent the direction vectors on both sides of the diamond area passing through the origin.

[0023] Further, in S2, the corresponding relationship of the diamond regions is established, which specifically includes:

[0024] Find the diamond area (whose diagonal is formed by the horizontal position of the scanner in the two scans) and establish the correspondence between the two scans. In addition, since the diamond area is located in the area between the two scan positions in the plot, the TLS target point cloud and the TLS reference point cloud usually have better overlap in this area than in other areas; in the diamond area, a pair of corresponding points should satisfy the following relationship:

[0025]

[0026] Where dr and Dr refer to the horizontal distances from point pr in the diamond region in the TLS reference point cloud to the two sides of the diamond passing through the origin; where dt and Dt refer to the horizontal distances from point pt in the diamond region in the TLS target point cloud to the two sides of the diamond passing through the origin; point pr and point pt are a pair of corresponding points; R represents the horizontal distance between the centers of the two scanners;

[0027] In order to quickly and accurately establish paired correspondences between points, the rhombic region correspondence principle (RRCP) based on formula (3) is proposed. First, the rhombic region data in the TLS reference point cloud and the TLS target point cloud are rasterized. Then, the grid point with index (i, j) in the TLS reference point cloud should correspond to the grid point with index (int(R / gridsize)+1-i,int(R / gridsize)+1-j) or (int(R / gridsize)-i,int(R / gridsize)-j) in the TLS target point cloud. Finally, error=|ds+dt-Ds-Dt| is calculated for these two correspondences respectively. The case with the smaller error is the correct correspondence.

[0028] Furthermore, in S2, the overlapping diamond area search specifically includes:

[0029] Through the above steps, θ takes values ​​at intervals of 2π / l in [0,2π), that is, θ=k*2π / l, k∈[0,l), and l diamond regions are generated for the TLS reference point cloud and the TLS target point cloud respectively; since the scanner has been leveled, the standard deviation of the Z coordinate difference (i.e., height difference) between corresponding points should be 0 in theory, and this method uses this feature to establish the true correspondence between the diamond regions;

[0030] The rhombus regions generated in the TLS reference point cloud and the TLS target point cloud are freely combined. Considering that the horizontal distance R between the centers of the two scanners is unknown, the values ​​are taken at intervals of Rstp (set to 0.2 m in the present invention) within the range [Rmin, Rmax], and the initial correspondence relationship N is repeatedly established. count =int((Rmax-Rmin) / R stp ) times; the settings of Rmin and Rmax depend on the empirical values ​​of data collection, where Rmin is set to a value less than R and Rmax is set to a value greater than R; the mean and standard deviation of the height differences of corresponding point pairs are calculated for all combinations using RRCP; the correct corresponding diamond area, R and average height difference dH are determined by selecting the combination with the smallest standard deviation.

[0031] Further, in S2, the transformation parameter estimation specifically includes:

[0032] First, the Fast Point Feature Histograms (FPFH) descriptor is used to establish initial correspondences for points in the overlapping diamond regions of the TLS reference point cloud and the TLS target point cloud. Subsequently, multiple constraints are introduced to eliminate false matches and obtain accurate correspondences:

[0033]

[0034] Among them, σ, δ and There are three different thresholds, which are set to 0.2m, 0.1m and π / 6rad respectively; angle1 represents the angle between the normal vector of a point in the reference point cloud and the Z axis; angle2 represents the angle between the normal vector of a point in the target point cloud and the Z axis;

[0035] Finally, two correspondences are randomly selected to estimate the transformation parameters; this process is repeated a certain number of times, and the case with the minimum root mean square error after registration is output, thus achieving coarse registration.

[0036] Furthermore, in S3, the common ground point screening specifically includes:

[0037] There are often bushes, rocks, stumps and dead branches on the forest floor, some of which are not correctly removed when extracting ground point data; the ground points in the overlapping diamond area are filtered to reduce noise and increase the overlap between data sets; after coarse registration, the ground points of the reference point cloud and the target point cloud in the overlapping diamond area are rasterized; in the grids with the same index, if the number of ground points of the reference point cloud and the target point cloud is not less than a certain value (set to 20 in the present invention), and the Z coordinate difference between the centroids of the points in the grid is less than a certain threshold (set to 0.1 meters), then the ground points in the area are considered to be common ground points between the two scans;

[0038] The point-to-surface ICP (Iterative Closest Point) algorithm registration specifically includes:

[0039] The ground is usually composed of many planes or curved surfaces. After obtaining the common ground points in the overlapping diamond area, the point-to-surface ICP algorithm is applied for precise registration to find the best registration result.

[0040] Another object of the present invention is to provide a forest TLS point cloud automatic registration system based on ground overlap search for implementing the forest TLS point cloud automatic registration method based on ground overlap search, the system comprising:

[0041] Data preprocessing module, used to reduce data redundancy and extract ground points;

[0042] Coarse registration module, used for rhombus region construction, rhombus region correspondence establishment, overlapping rhombus region search and transformation parameter estimation;

[0043] The precise registration module is used for common ground point screening and point-to-surface ICP algorithm registration.

[0044] Another object of the present invention is to provide a computer device, characterized in that the computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the forest TLS point cloud automatic registration method based on ground overlapping search.

[0045] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to perform the steps of the forest TLS point cloud automatic registration method based on ground overlap search.

[0046] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0047] First, the existing forest TLS point cloud automatic registration method mainly relies on the location of trees. However, in a complex forest environment, it is challenging to accurately obtain the location of trees, resulting in registration failure and low accuracy, which in turn limits the application of TLS in forest surveys. To solve this problem, the present invention proposes a new method for automatic registration of TLS point clouds in a forest environment, which searches for ground points in overlapping diamond areas between different measuring stations, designs a method for establishing correspondence between diamond areas, and establishes correspondence based on the graphic constraints of the diamond areas and the FPFH features to complete the registration.

[0048] This method is the first attempt to automatically register forest point clouds relying only on ground points, without the need to extract tree attributes, eliminating many complex and time-consuming steps. Another advantage of this method is that ground points are widely present in the original forest point cloud and are largely unaffected by natural conditions such as wind. In addition, there are a variety of tree types in the forest, requiring different methods to estimate tree attributes, but the ground always has similar characteristics. This inherent similarity makes the proposed method applicable to a wide range of forest types.

[0049] Second, the present invention is committed to proposing a forest TLS point cloud automatic registration method based on ground overlap search, so as to quickly and accurately automatically realize the registration of different TLS scanning point clouds in the forest, and promote the application of TLS in forest surveys. Compared with the forest TLS point cloud automatic registration method that relies on tree attributes, the present invention uses ground points as registration primitives and has the following advantages:

[0050] Wide range of applications: Ground points are widely present in various types of forests and always have similar characteristics in various forest environments. Different tree types often require different tree attribute extraction methods. In complex forest environments, it is challenging to accurately obtain tree locations.

[0051] Efficient: The process of tree attribute extraction often requires ground extraction first. Therefore, directly using ground points for registration eliminates the process of tree attribute extraction.

[0052] High precision: The accuracy of tree attribute extraction is often greatly affected by terrain, tree species, natural environment and other conditions. However, the ground point extraction algorithm is mature and has good extraction effect.

[0053] Third, the expected benefits and commercial value of the technical solution of the present invention after transformation are as follows: The present invention proposes a method and system for automatic registration of forest TLS point clouds based on ground overlapping search, which provides a new solution for forest surveys, and can automatically, efficiently and accurately realize the registration of different scanned TLS point clouds, providing support for the automatic acquisition of forest parameters. The present invention can provide technical support for forestry resource surveys, promote the development of the forestry economy, help achieve the dual carbon goals, and promote international cooperation.

[0054] The technical solution of the present invention fills the technical gap in the industry at home and abroad: the technical solution of the present invention innovatively proposes to use ground points as registration primitives, and designs a method for establishing the correspondence relationship of diamond regions, and completes the automatic registration of forest TLS point clouds according to the graphic constraints and FPFH features of the diamond regions. It fills the technical gap in the industry at home and abroad in this regard and provides a new technical solution for data collection in forestry resource surveys.

[0055] The technical solution of the present invention solves the technical problems that people have been eager to solve but have never been successful: the forest point cloud registration method that relies on artificial reference targets requires a lot of manpower and time costs to place artificial targets; in complex forest environments, artificial target recognition will be affected. The forest point cloud registration method that relies on external information is limited by specific hardware and has high costs; in complex forest environments, the acquisition and use of external information will be limited. In complex forest environments, it is challenging to accurately obtain tree attributes; using tree attributes for registration has a large time complexity. Therefore, the current forest point cloud registration method is often more suitable for simple forest environments. In addition, due to the complexity of the forest environment, it is a difficult task to quickly and accurately realize the automatic registration of forest TLS point clouds, which needs to be solved urgently. The forest TLS point cloud automatic registration method and system based on ground overlapping search proposed in the present invention use ground points widely present in various forest environments as registration primitives, successfully solving the technical problem of low degree of registration automation between different TLS scanning point clouds in forest environments, and can be applied to complex forest environments.

[0056] Fourth, the forest TLS point cloud automatic registration method based on ground overlap search proposed in the present invention embodies significant technical progress, which is mainly reflected in the following aspects:

[0057] 1. Optimization of data preprocessing

[0058] Technological progress: The data preprocessing step effectively reduces data redundancy and accurately extracts ground points, which provides a clearer and more accurate data basis for the subsequent registration process. In a forest environment, due to the occlusion of trees, shrubs, etc., the extraction of ground points is particularly critical and challenging. The data preprocessing optimization of this method ensures the high-quality extraction of ground points, laying the foundation for improving the registration accuracy.

[0059] 2. Innovative methods for coarse registration

[0060] Technological progress: In the coarse registration stage, an innovative registration method is used to perform overlapping diamond region search and transformation parameter estimation by constructing diamond regions and establishing correspondences between them. The construction of diamond regions and the establishment of correspondences provide a new perspective for processing TLS point cloud data in forest environments, especially when the point cloud overlap is low or the quality is complex. This method improves the efficiency and reliability of coarse registration.

[0061] 3. Efficient Algorithm for Precise Registration

[0062] Technological progress: The fine registration stage uses common ground point screening and point-to-surface ICP algorithm registration. This method effectively utilizes the characteristics of ground points and improves the accuracy of registration. The point-to-surface ICP algorithm has high efficiency and accuracy when processing large amounts of point cloud data, especially for complex forest terrain. The application of this step significantly improves the accuracy and robustness of the entire registration process.

[0063] 4. Automation and efficiency of the overall process

[0064] Technological progress: The entire registration method from data preprocessing to fine registration is designed to be highly automated, reducing the need for manual intervention and improving processing speed and efficiency. In the processing of forest TLS point cloud data, the increased degree of automation has greatly saved time and labor costs, which is particularly important in large-scale or time-sensitive applications.

[0065] 5. Specific optimization for forest environment

[0066] Technological progress: This registration method is specifically optimized for the special challenges of point cloud data processing in forest environments, taking into account the complexity of forest environments, such as occlusion by trees and irregularities of the ground, to ensure the applicability and effectiveness of the method.

[0067] The forest TLS point cloud automatic registration method based on ground overlapping search provided by the present invention demonstrates significant technological progress by optimizing data preprocessing, innovating coarse registration methods, applying efficient and precise registration algorithms, realizing process automation, and performing specific optimization for forest environments, thereby improving the efficiency, accuracy, and robustness of TLS point cloud data registration in forest environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a flow chart of a forest TLS point cloud automatic registration method based on ground overlap search provided by an embodiment of the present invention;

[0069] Figure 2 is a schematic diagram of rhombus region rasterization provided by an embodiment of the present invention;

[0070] Figure 3 is a schematic diagram of the corresponding relationship of diamond-shaped areas provided in an embodiment of the present invention;

[0071] Figure 4 is a schematic diagram of a sample site scanning position provided by an embodiment of the present invention;

[0072] Figure 5 Schematic diagram of ground point extraction provided by an embodiment of the present invention, (a) top view, (b) front view, (c) left view;

[0073] Figure 6 is a schematic diagram of a registration result provided by an embodiment of the present invention, including (a) a sample plot after registration, (b) a cross-sectional view, and (c) a partially enlarged view of the cross-sectional view;

[0074] Figure 7 It is a structural diagram of a forest TLS point cloud automatic registration system based on ground overlapping search provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0075] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0076] Example 1: Forest Park TLS point cloud data registration

[0077] To conduct topographic mapping and tree growth monitoring in forest parks, it is necessary to accurately align the terrestrial laser scanning (TLS) point cloud data collected at different time points or locations to analyze terrain changes and tree growth.

[0078] 1. Data preprocessing: Use a high-pass filter to remove noise points, and use voxel grid downsampling method to reduce the amount of data and retain important ground features.

[0079] A ground point extraction algorithm is applied to extract ground points, excluding non-ground objects such as trees and shrubs.

[0080] 2. Coarse registration: Construct a diamond region grid covering the entire scan area in each TLS point cloud dataset. Establish the correspondence between diamond regions by comparing the geometric features and spatial positions of diamond regions in different point cloud datasets. Use the overlapping diamond region search algorithm to determine the matching area and estimate the preliminary transformation parameters.

[0081] 3. Fine registration: Common ground points are selected from the coarse registration results for use in the fine registration process.

[0082] The point-to-plane ICP algorithm is applied to iteratively align ground points with their corresponding ground planes until a predetermined error threshold or number of iterations is reached.

[0083] Application scenarios:

[0084] It is suitable for topographic mapping, ecological monitoring and long-term ecological research in forest parks, especially for forest environments with complex terrain and dense cover.

[0085] Example 2: TLS point cloud data registration for urban forest ecological monitoring

[0086] To conduct ecological monitoring of urban forest areas, it is necessary to align TLS point cloud data collected in different seasons to analyze the seasonal changes of trees and the dynamics of the ecological environment.

[0087] 1. Data preprocessing:

[0088] The original TLS point cloud data is filtered to remove noise points generated by dynamic objects such as branches caused by wind.

[0089] The ground point extraction algorithm is used to separate the ground point cloud to provide an accurate benchmark for the subsequent registration process.

[0090] 2. Coarse registration: A regular diamond area grid is constructed based on the ground points in each TLS point cloud dataset as the basic unit of coarse registration.

[0091] By analyzing the characteristics of overlapping diamond regions, such as the density and distribution of ground points, the best match between the diamond regions is determined.

[0092] The preliminary transformation matrix is ​​calculated based on the matched diamond regions to achieve a rough alignment of the point cloud dataset.

[0093] 3. Precision registration:

[0094] Based on the rough registration, common ground points in the overlapping area are selected for further fine registration.

[0095] The improved point-to-surface ICP algorithm is used to accurately align the point cloud data with the ground model to ensure high accuracy and stability of the registration results.

[0096] Application scenarios:

[0097] It is suitable for urban forest ecological monitoring, tree health assessment and ecological environment change analysis, and is particularly suitable for research projects that require long-term monitoring across seasons.

[0098] These two embodiments demonstrate the specific implementation schemes of the forest TLS point cloud automatic registration method based on ground overlapping search in different application scenarios, reflecting the flexibility and practicality of the method.

[0099] like Figure 1 As shown, a forest TLS point cloud automatic registration method based on ground overlap search, the method comprising:

[0100] S1: Data preprocessing, reducing data redundancy and ground point extraction;

[0101] S2: coarse registration, diamond region construction, diamond region correspondence establishment, overlapping diamond region search and transformation parameter estimation;

[0102] S3: Precise registration, common ground point screening and point-to-surface ICP algorithm registration.

[0103] The specific implementation steps are as follows:

[0104] Step 1: Data Preprocessing

[0105] (1) Reduce data redundancy

[0106] Calculate the horizontal distance from each point to the center of the scanner, and if it is greater than a certain threshold, it will be eliminated. The threshold is usually set to 15m, and the user can set the threshold size according to different forest scenes.

[0107] (2) Ground point extraction

[0108] The ground point cloud data is extracted by morphological filtering. All points in the point cloud are converted into a two-dimensional grid space. The lowest point in each pixel is selected as the seed point for clustering to obtain the ground point.

[0109] Step 2: Rough registration

[0110] (1) Diamond area construction

[0111] In the XOY coordinate plane, a rhombus region with a side length of Rmax is constructed. One vertex of the rhombus is located at the origin O(0,0), and the polar angles of the two sides passing through the vertex are θrad and θ+2π / 3rad respectively. Points whose horizontal positions are within the rhombus region are extracted from the ground point cloud, and the horizontal distances from these points to the two sides are calculated as follows: Figure 2 As shown in (a), the calculation formula is as follows:

[0112]

[0113] In the formula, p = (x, y) represents the horizontal coordinates of the point, m = (cosθ, sinθ) and n = (cos(θ+2π / 3), sin(θ+2π / 3)) represent the direction vectors on both sides of the diamond area passing through the origin.

[0114] The point cloud of the diamond area is rasterized. The coordinates of each grid are represented by the centroid of all points in the grid. The index (i, j) of the grid where the point p(x, y, z) is located is as follows: Figure 2 As shown in (b), the calculation formula is as follows:

[0115]

[0116] In the formula, int means rounding the number to the nearest integer; gridsize represents the grid side length, which is usually set to 0.2m.

[0117] (2) Establishment of the correspondence relationship between diamond regions

[0118] Or and Ot represent the horizontal positions of the reference scanner and the target scanner in the plot, respectively. θr and θr+2π / 3 represent the polar angles of the two sides of the diamond area in the reference scanner coordinate system, respectively. θt and θt+2π / 3 represent the polar angles of the two sides of the diamond area in the target scanner coordinate system, respectively. Figure 3 (a) and (b) show overlapping diamond areas in the plots.

[0119] The method aims to find the rhombus-shaped area whose diagonals are formed by the horizontal positions of the scanners in the two scans, i.e. Figure 3 In addition, since the diamond-shaped area is located between the two scan positions in the plot, the TLS target point cloud and the TLS reference point cloud usually have better overlap in this area than in other areas. Figure 3 As shown in (a), in the rhombus region, a pair of corresponding points should satisfy the following relationship:

[0120]

[0121] Among them, dr and Dr refer to the horizontal distances from point pr in the diamond area in the TLS reference point cloud to the two sides of the diamond passing through the origin; among them, dt and Dt refer to the horizontal distances from point pt in the diamond area in the TLS target point cloud to the two sides of the diamond passing through the origin; point pr and point pt are a pair of corresponding points; R represents the horizontal distance between the centers of the two scanners, such as Figure 3 shown.

[0122] In order to quickly and accurately establish the pairwise correspondence between points, we propose the rhombic region correspondence principle (RRCP) based on formula (3). First, the rhombic region data in the TLS reference point cloud and the TLS target point cloud are rasterized, as follows: Figure 2 (b) As shown. Then, the grid point with index (i, j) in the TLS reference point cloud should correspond to the grid point with index (int(R / gridsize)+1-i,int(R / gridsize)+1-j) or (int(R / gridsize)-i,int(R / gridsize)-j) in the TLS target point cloud. Finally, for these two correspondences, error = |ds+dt-Ds-Dt| is calculated respectively. The case with the smaller error is the correct correspondence ( Figure 3 (b)).

[0123] (3) Overlapping diamond area search

[0124] Through the above steps, θ takes values ​​at intervals of 2π / l in [0,2π), that is, θ=k*2π / l, k∈[0,l), and l (set to 360 in the present invention) diamond regions are generated for the TLS reference point cloud and the TLS target point cloud respectively. Since the scanner has been leveled, the standard deviation of the Z coordinate difference (i.e., height difference) between corresponding points should be 0 in theory. This method uses this feature to establish the true correspondence between the diamond regions.

[0125] The rhombus regions generated in the TLS reference point cloud and the TLS target point cloud are freely combined. Considering that the horizontal distance R between the centers of the two scanners is unknown, the values ​​are taken at intervals of Rstp (set to 0.2 m in the present invention) within the range [Rmin, Rmax], and the initial correspondence relationship N is repeatedly established. count =int((Rmax-Rmin) / R stp) times. The settings of Rmin and Rmax depend on the empirical values ​​of data collection, where Rmin is set to a value less than R and Rmax is set to a value greater than R. The mean and standard deviation of the height difference of corresponding point pairs are calculated for all combinations using RRCP. The correct corresponding diamond area, R, and average height difference dH are determined by selecting the combination with the smallest standard deviation.

[0126] (4) Transformation parameter estimation

[0127] First, the Fast Point Feature Histograms (FPFH) feature descriptor is used to establish initial correspondences for points in the TLS reference point cloud and the TLS target point cloud that are located in the overlapping diamond region. Subsequently, the present invention introduces multiple constraints to eliminate false matches and obtain accurate correspondences:

[0128]

[0129] Among them, σ, δ and There are three different thresholds, which are set to 0.2m, 0.1m and π / 6rad respectively; angle1 represents the angle between the normal vector of a point in the reference point cloud and the Z axis; angle2 represents the angle between the normal vector of a point in the target point cloud and the Z axis.

[0130] Finally, two correspondences are randomly selected to estimate the transformation parameters. This process is repeated a certain number of times, and the case with the smallest root mean square error (RMSE) after registration is output, thus achieving coarse registration.

[0131] Step 3: Fine alignment

[0132] (1) Common ground point screening

[0133] There are often bushes, rocks, stumps and dead branches on the forest floor, some of which are not correctly removed when extracting ground point data. Therefore, the present invention filters the ground points in the overlapping diamond area to reduce noise and increase the overlap between data sets. After coarse alignment, the ground points of the reference point cloud and the target point cloud in the overlapping diamond area are rasterized. In the grids with the same index, if the number of ground points of the reference point cloud and the target point cloud is not less than a certain value (set to 20 in the present invention), and the Z coordinate difference between the centroids of the points in the grid is less than a certain threshold (set to 0.1 meters), the ground points in the area are considered to be common ground points between the two scans.

[0134] (2) Point-to-surface ICP (Iterative Closest Point) algorithm registration

[0135] The ground is usually composed of many planes or curved surfaces. Therefore, after obtaining the common ground points in the overlapping diamond regions, the point-surface ICP algorithm is applied for precise registration to find the best registration result.

[0136] like Figure 7 As shown, an embodiment of the present invention provides a forest TLS point cloud automatic registration system based on ground overlap search for implementing the forest TLS point cloud automatic registration method based on ground overlap search, the system comprising:

[0137] Data preprocessing module, used to reduce data redundancy and extract ground points;

[0138] Coarse registration module, used for rhombus region construction, rhombus region correspondence establishment, overlapping rhombus region search and transformation parameter estimation;

[0139] The precise registration module is used for common ground point screening and point-to-surface ICP algorithm registration.

[0140] An embodiment of the present invention provides a computer device, characterized in that the computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the forest TLS point cloud automatic registration method based on ground overlapping search.

[0141] An embodiment of the present invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the forest TLS point cloud automatic registration method based on ground overlap search.

[0142] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0143] Taking a forest plot as an example, the technical solution of the forest TLS point cloud automatic registration method based on ground overlapping search is as follows.

[0144] (1) Sample site TLS data collection

[0145] In the forest plot, the operator used a terrestrial laser scanner to collect data at multiple scanning locations to ensure complete coverage of the forest. The scanning locations should try to avoid obstructions such as adjacent trees. A total of 4 scans were performed at different locations in the plot, such as Figure 4 S1-S4 indicate scanning positions.

[0146] (2) Ground point extraction

[0147] For different scanning data of the sample site, morphological filtering is used to extract ground points. The ground point extraction results of a certain scanning TLS point cloud of the sample site are displayed, as shown in the figure below. Figure 5 shown.

[0148] (3) Point cloud registration

[0149] After extracting the ground points of different scans of the sample plot, the forest TLS point cloud automatic registration method based on ground overlapping search is used to find the ground points in the overlapping diamond area located in the middle of the different measuring stations to be registered. The correspondence between the graphic constraints of the diamond area and the FPFH features is established to complete the coarse registration, and then the point-to-surface ICP algorithm is used to optimize the coarse registration results.

[0150] A forest plot is used as an example to demonstrate the technical effect. The four scanned point clouds of the plot are registered using the forest TLS point cloud automatic registration method based on ground overlap search. Scans S2, S3, and S4 are registered to scan S1 ( Figure 4 ). The visual registration effect is as follows Figure 6 As shown in the figure, the point clouds of different colors represent the TLS point clouds of different scans of the sample site; the different TLS scans are correctly registered, and the registration accuracy is sufficient to describe the structural characteristics of branches and trunks. The error statistics of the registration results are shown in Table 1. The average translation error, average rotation error and average point position error of the registration results of the sample site are 2.68cm, 1.79mrad and 3.55cm respectively.

[0151] Table 1. Accuracy evaluation of registration results

[0152]

[0153] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A forest TLS point cloud automatic registration method based on ground overlap search, characterized in that: The method includes: S1: Data preprocessing: reducing data redundancy and extracting ground points; S2: Rough registration: diamond region construction, diamond region correspondence establishment, overlapping diamond region search and transformation parameter estimation; S3: Precision registration: common ground point screening and point-surface ICP algorithm registration; In S2, the diamond-shaped area is constructed, specifically including: In the XOY coordinate plane, a rhombus region with a side length of Rmax is constructed. One vertex of the rhombus is located at the origin O(0,0), and the polar angles of the two sides passing through the vertex are θrad and θ+2π / 3rad respectively. Points whose horizontal positions are within the rhombus region are extracted from the ground point cloud, and the horizontal distances from these points to the two sides are calculated. The calculation formula is as follows: Where p = (x, y) represents the horizontal coordinates of the point, m = (cosθ, sinθ) and n = (cos(θ+2π / 3), sin(θ+2π / 3)) represent the direction vectors of the two sides of the diamond area passing through the origin respectively; In S2, the corresponding relationship of the diamond regions is established, which specifically includes: Find the diamond-shaped area and establish the correspondence between the two scans; the diamond-shaped area is located between the two scan positions in the sample plot; in the diamond-shaped area, a pair of corresponding points should satisfy the following relationship: Where dr and Dr refer to the horizontal distances from point pr in the diamond region in the TLS reference point cloud to the two sides of the diamond passing through the origin; where dt and Dt refer to the horizontal distances from point pt in the diamond region in the TLS target point cloud to the two sides of the diamond passing through the origin; point pr and point pt are a pair of corresponding points; R represents the horizontal distance between the centers of the two scanners; First, the diamond area data in the TLS reference point cloud and the TLS target point cloud are rasterized; then, the grid point with index (i, j) in the TLS reference point cloud should correspond to the grid point with index (int(R / gridsize)+1-i,int(R / gridsize)+1-j) or (int(R / gridsize)-i,int(R / gridsize)-j) in the TLS target point cloud; finally, error=|dr+dt-Dr-Dt| is calculated for these two correspondences respectively, and the case with smaller error is the correct correspondence; In S2, overlapping diamond area search specifically includes: θ takes values ​​at intervals of 2π / l in [0,2π), that is, θ=k*2π / l, k∈[0,l), to generate l diamond-shaped areas for the TLS reference point cloud and the TLS target point cloud respectively; the scanner has been leveled, so the standard deviation of the Z coordinate difference between corresponding points should theoretically be 0; The rhombus regions generated in the TLS reference point cloud and the TLS target point cloud are freely combined; considering that the horizontal distance R between the centers of the two scanners is unknown, the values ​​are taken at intervals of Rstp within the range [Rmin, Rmax], and the initial correspondence relationship N is repeatedly established count =int((Rmax-Rmin) / R stp ) times; the settings of Rmin and Rmax depend on the empirical values ​​of data collection, where Rmin is set to a value less than R and Rmax is set to a value greater than R; the mean and standard deviation of the height differences of corresponding point pairs are calculated for all combinations using RRCP; the correct corresponding diamond area, R and average height difference dH are determined by selecting the combination with the smallest standard deviation.

2. The forest TLS point cloud automatic registration method based on ground overlap search as claimed in claim 1, characterized in that: In S1, reducing data redundancy specifically includes: Calculate the horizontal distance from each point to the center of the scanner. If it is greater than a certain threshold, it will be eliminated. The threshold is usually set to 15m, and the user can set the threshold size according to different forest scenes. The ground point extraction specifically includes: The ground point cloud data is extracted through morphological filtering; all points in the point cloud are converted into a two-dimensional grid space; the lowest point in each pixel is selected as the seed point for clustering to obtain the ground point.

3. The forest TLS point cloud automatic registration method based on ground overlap search as claimed in claim 1, characterized in that: In S2, the transformation parameter estimation specifically includes: First, the fast point feature histogram feature descriptor is used to establish initial correspondences for points in the overlapping diamond regions of the TLS reference point cloud and the TLS target point cloud; then, several constraints are introduced to eliminate false matches and obtain accurate correspondences: Among them, σ, δ and are three different thresholds, set to 0.2m, 0.1m and π / 6rad respectively; angle1 represents the angle between the normal vector of a point in the reference point cloud and the Z axis; angle2 represents the angle between the normal vector of a point in the target point cloud and the Z axis; dz represents the Z coordinate difference between a set of initial corresponding points in the overlapping diamond area in the TLS reference point cloud and the TLS target point cloud, that is, the height difference; Finally, two correspondences are randomly selected to estimate the transformation parameters; this process is repeated a certain number of times, and the case with the minimum root mean square error after registration is output.

4. The forest TLS point cloud automatic registration method based on ground overlap search as claimed in claim 1, characterized in that: In S3, the common ground point screening specifically includes: There are often bushes, rocks, stumps and dead branches on the forest floor, some of which are not correctly removed when extracting ground point data; the ground points in the overlapping diamond area are filtered to reduce noise and increase the overlap between data sets; after coarse registration, the ground points of the reference point cloud and the target point cloud in the overlapping diamond area are rasterized; in the grids with the same index, if the number of ground points in the reference point cloud and the target point cloud is not less than a certain value, and the Z coordinate difference between the centroids of the points in the grid is less than a certain threshold, the ground points in the area are considered to be common ground points between the two scans; The point-surface ICP algorithm registration specifically includes: The ground is usually composed of many planes or curved surfaces. After obtaining the common ground points in the overlapping diamond area, the point-to-surface ICP algorithm is applied for precise registration to find the best registration result.

5. A forest TLS point cloud automatic registration system based on ground overlap search, which implements the forest TLS point cloud automatic registration method based on ground overlap search as claimed in any one of claims 1 to 4, characterized in that: The system includes: Data preprocessing module, used to reduce data redundancy and extract ground points; Coarse registration module, used for rhombus region construction, rhombus region correspondence establishment, overlapping rhombus region search and transformation parameter estimation; The precise registration module is used for common ground point screening and point-to-surface ICP algorithm registration.

6. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the forest TLS point cloud automatic registration method based on ground overlapping search as described in any one of claims 1-4.

7. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the forest TLS point cloud automatic registration method based on ground overlap search as described in any one of claims 1 to 4.