US-20260126550-A1 - UNKNOWN ENVIRONMENT MAPPING METHOD FOR SINGLE-BEAM LASER OF AIRCRAFT

US20260126550A1US 20260126550 A1US20260126550 A1US 20260126550A1US-20260126550-A1

Abstract

The invention discloses an unknown environment mapping method for a single-beam laser of aircraft, which belongs to the field of three-dimensional map reconstruction technology. Firstly, the data collected by a single-beam LiDAR is aligned with the timestamp of the stepper motor, and the obtained point cloud data is filtered and matched to obtain three-dimensional map information. The pose of the aircraft is estimated by the wheel odometer and IMU, and the loop detection is performed to reduce the cumulative error and update the three-dimensional map information in real time. The frontier search algorithm is used to determine the target point of the aircraft flight, the A* algorithm is used to plan the path trajectory, and the B-spline curve optimization algorithm is used to make the trajectory of the aircraft smoother. The control command is sent to the aircraft, and then the global three-dimensional map of the unknown environment is obtained.

Inventors

Jianwen HUO
Ping Zhang
Jingnan Peng
Lihui Zhang
Liguo TAN
Mingrun LING
Hua Zhang
Jun Sun

Assignees

SOUTHWEST UNIVERSITY OF SCIENCE AND TECHNOLOGY
HARBIN INSTITUTE OF TECHNOLOGY

Dates

Publication Date: 20260507
Application Date: 20251231
Priority Date: 20250225

Claims (10)

1 . An unknown environment mapping method for an aircraft single-beam laser, wherein the aircraft is configured with a single-beam laser detection and ranging (LiDAR), a stepper motor, and a binocular camera, and wherein point cloud data of the single-beam LiDAR is aligned with stepper motor data on a timestamp; wherein the method comprises the following steps: S 1 , denoising point cloud data collected at a current position, and adjusting position and attitude information of an aircraft by using denoised point cloud data, so as to unify a single-beam LiDAR coordinate system and the aircraft; S 2 , based on a unified coordinate system, aligning the point cloud data at adjacent positions initially, and performing a feature extraction, a point cloud alignment and a point cloud stitching in turn to generate a unified point cloud map; S 3 , for point cloud data in the unified point cloud map, establishing a local neighborhood of each point cloud according to a K-nearest neighbors search algorithm, and determining a local tangent space of each point cloud in a local neighborhood by a principal component analysis method, using a multi-dimensional scaling analysis method to integrate information of all local tangent spaces, and obtaining a global parameterization of the point cloud data on a Riemannian manifold; S 4 , based on point cloud data represented by the global parameterization, generating a continuous three-dimensional manifold surface according to a Delaunay triangulation method, and generating a local map of the current position after a smooth and curvature adjustment of the continuous three-dimensional manifold surface; S 5 , in a process of aircraft data acquisition, using a loop detection algorithm to correct a pose of the aircraft, and then optimizing a local map to obtain an optimized local map; and S 6 , taking a current optimized local map as a starting point, performing a boundary exploration and a path planning, and collecting data according to a planned path, repeating S 1 -S 5 to gradually generate an optimized local map, and then completing an unknown environment mapping.
2 . The unknown environment mapping method for the aircraft single-beam laser according to claim 1 , wherein S 1 comprises the following steps: S 11 , calculating a distance between the point cloud data collected at the current position and its adjacent point cloud; S 12 , according to an average value and standard deviation of each distance, determining a determination threshold of an outlier point cloud; S 13 , removing outlier point cloud data according to the determination threshold of the outlier point cloud, and completing a point cloud noise reduction; S 14 , extracting feature points from the point cloud data after a noise reduction; S 15 , according to a type of feature points and a nearest point search method, establishing a geometric relationship between feature points after noise reduction at different times, and then calculating a pose estimation of a single-beam LiDAR during this period; S 16 , according to an estimated pose estimation, using the point cloud data from different perspectives for three-dimensional reconstruction, and then determining an initial pose of the aircraft; S 17 , according to the initial pose of the aircraft, using a Kalman-based fusion matching algorithm to predict a next state estimation of the aircraft, and correcting the state estimation according to a difference between a predicted value and an actual observed value to obtain a next pose of the aircraft; S 18 , weighting and averaging the initial pose and the next pose of the aircraft to adjust the attitude information of the aircraft, and then unifying a single-beam LiDAR coordinate system of the aircraft.
3 . The unknown environment mapping method for the aircraft single-beam laser according to claim 1 , wherein S 2 comprises the following steps: S 21 , according to a random sampling consistency algorithm, using a registration point to unify a coordinate system of two points to be configured in the point cloud data at adjacent positions, and completing an initial alignment to obtain coarsely calibrated point cloud data; S 22 , clustering continuous point cloud data collected by a single-beam LiDAR to extract feature points; wherein types of feature points include scatter points, corner points, and breakpoints; S 23 , using an improved iterative adaptive point algorithm to screen the corner points, using a continuous edge tracking algorithm to screen the breakpoints, and then screening out all the real corner points and breakpoints, and filtering out scattered points; S 24 , using a least square method to perform a feature line fitting of selected corner points and breakpoints to determine a feature line segment; S 25 , using an ICP algorithm to coarsely calibrate point cloud data and feature line segments for fine registration; and S 26 , fusing finely registered point cloud data into a unified voxel grid, and performing a point cloud stitching to generate a unified point cloud map.
4 . The unknown environment mapping method for the aircraft single-beam laser according to claim 3 , wherein in S 22 , the method of clustering continuous point cloud data is as follows: taking a position of a single-beam LiDAR as an origin O, scanning a straight line L representing a detected object at an equal interval angle Δθ, and then obtaining a continuous point cloud data Q i ; a length distance between the origin and each point cloud data is expressed as d i , and then calculating a slope value of an i-th point cloud data; when a slope value difference between adjacent point cloud data is less than a predetermined threshold, it is determined to be on the same straight line; otherwise, it is determined to be a mutation point and used as a feature point; where a slope value k i is expressed as: k i = d i + 1 - d i d i ⁢ Δ ⁢ θ .
5 . The unknown environment mapping method for the aircraft single-beam laser according to claim 1 , wherein S 3 comprises the following steps: S 31 , using the K-nearest neighbors search algorithm to search each point cloud in the point cloud data of the unified point cloud map to find a corresponding K-nearest neighbors point cloud; S 32 , using a principal component analysis network to perform a principal component analysis on a neighborhood of a searched point cloud, calculating a covariance matrix and extracting eigenvalues and eigenvectors, and then obtaining a local tangent space of each point cloud in a corresponding neighborhood; S 33 , calculating a Riemannian metric matrix of each point cloud according to the local tangent space; and S 34 , using a multidimensional scale analysis method to integrate the Riemannian metric matrix of all point clouds to obtain a global parameter representation of point cloud data on the Riemannian manifold.
6 . The unknown environment mapping method for the aircraft single-beam laser according to claim 5 , wherein S 32 is specified as follows: using a principal component analysis network to perform a principal component analysis on the neighborhood of the point cloud obtained by searching; decentralizing point cloud data obtained by analysis, and calculating a covariance matrix of decentralized neighborhood point cloud data; decomposing the covariance matrix to obtain the eigenvalues and eigenvectors, and calculating the local tangent space of each point cloud in its field according to the eigenvalues and eigenvectors; and in a three-dimensional space where the local tangent space is located, a feature vector V j corresponds to a main direction of a local geometry, where a feature vector V 1 denotes a main direction of a point cloud data distribution, a feature vector V 2 denotes a secondary direction, and a feature vector V 3 denotes a local normal vector, that is, a normal vector of the point cloud surface; the feature vectors V 1 and V 2 form the local tangent space, and the feature vector V 3 is perpendicular to the local tangent space.
7 . The unknown environment mapping method for the aircraft single-beam laser according to claim 1 , wherein S 4 comprises the following steps: S 41 , using the Delaunay triangulation method to generate a triangular mesh in a manifold space where globally parameterized point cloud data is located, and generating a continuous three-dimensional manifold surface from the triangular mesh; S 42 , extracting a topological structure of the three-dimensional manifold surface; S 43 , extracting topological structure characteristics of the three-dimensional manifold surface by using a discrete Morse theory algorithm, and optimizing the topological structure; S 44 , using an optimization method of graph cuts to perform a smooth and curvature adjustment for the three-dimensional manifold surface according to an optimized topology; and S 45 , performing a three-dimensional visualization of the three-dimensional manifold surface after the smooth and curvature adjustment to generate a local map of a current position.
8 . The unknown environment mapping method for the aircraft single-beam laser according to claim 1 , wherein S 5 comprises the following steps: S 51 , in a process of local map generation, collecting aircraft odometer data continuously to obtain real-time estimated pose information and determine a corresponding current position; S 52 , when an Euclidean distance between the current position and a historical position of the aircraft is less than the predetermined threshold, a loopback detection is triggered; S 53 , extracting feature points from collected data of a binocular camera configured for the aircraft; S 54 , performing a feature point matching and data splicing on extracted feature points in turn, and then fusing with the point cloud data of a LiDAR in the current position; S 55 , according to fused data, calculating a pose error of the aircraft between a current frame and a historical frame, and according to aircraft pose information after optimization and adjustment, and then fusing the point cloud data corresponding to the optimized aircraft pose information into the local map; and S 56 , performing a smooth and curvature adjustment of the three-dimensional manifold surface corresponding to the local map after a fusion of point cloud data to obtain an optimized local map.
9 . The unknown environment mapping method for the aircraft single-beam laser according to claim 1 , wherein S 6 comprises the following steps: S 61 , detecting a boundary of an unexplored area in the current optimized local map, and selecting a target point closest to the aircraft; S 62 , using an A* path planning algorithm to calculate an optimal path from a current boundary point to the target point; S 63 , performing a uniform sampling for the determined optimal path, using selected control points, curve orders, and node vectors to calculate a B-spline basis function, and then subdividing a B-spline curve into several discrete points as passing points of the aircraft; S 64 , performing S 1 -S 5 to gradually generate an optimized local map at each passing point, and then completing an unknown environment mapping.
10 . The unknown environment mapping method for the aircraft single-beam laser according to claim 1 , wherein in S 63 , under a selected set of control points {p 0 , p 1 , . . . , p n }, the B-spline curve is represented as: C ⁡ ( T ) = ∑ n = 0 N N n , k ( t ) ⁢ P n where C(t) denotes a point of the B-spline curve at a parameter t, P n denotes an n-th control point, and N n,k (t) denotes a k-th-order B-spline basis function corresponding to an n-th control point, it is expressed as: N n , k ( t ) = t - t n t n + k - t n ⁢ N n , k - 1 ( t ) + t n + k + 1 - t t n + k + 1 - t n + 1 ⁢ N n + 1 , k - 1 ( t ) ; where t n denotes the n-th node value in the spline knot vector.

Description

TECHNICAL FIELD The invention belongs to the field of autonomous navigation robots and LiDAR three-dimensional space reconstruction technology, and specifically relates to an unknown environment mapping method for a single-beam laser of aircraft. BACKGROUND Traditional surveying and mapping methods mainly rely on manual measurements conducted by surveying engineers. However, the measurement process is cumbersome, time-consuming, and wastes considerable manpower and material resources. Moreover, the outcome of such measurements is often less than ideal, with errors that are difficult to control, resulting in relatively low accuracy. With the advancement of 3D spatial reconstruction technology, surveying personnel now only need to traverse the environment to be mapped using equipment equipped with surveying tools. This allows environmental structure information to be acquired more easily, greatly reducing errors caused by manual measurement. Three-dimensional reconstruction technology serves as an important technical means for aircraft to perceive their surrounding environment. By utilizing various sensors onboard, the aircraft can actively capture characteristic information from its surroundings. However, multi-beam LiDAR, which offers better mapping performance on the market, is expensive and may not be cost-effective in certain environmental measurement scenarios. To overcome the drawbacks of traditional single-beam LiDAR 3D reconstruction methods and to enhance portability, measurement accuracy, and cost-effectiveness, the invention patent designs a single-beam laser-based mapping system for unknown environments. This system primarily uses a single-beam radar to accomplish 3D environmental reconstruction. It retains the advantages of multi-beam radar, such as long measurement range, insensitivity to light intensity and weather conditions, and the ability to obtain high-precision three-dimensional environmental information. SUMMARY In view of the above shortcomings in the existing technology, an unknown environment mapping method for a single-beam laser of aircraft provided by the invention solves the problems of insufficient measurement accuracy, poor portability, and economy of the existing related methods. It employs a single-beam laser to perform a three-dimensional environmental reconstruction, while retaining the advantages typical of multi-beam radar, such as an extensive measurement range, immunity to variations in light intensity and weather conditions, and the capability to acquire highly accurate three-dimensional environmental data. In order to achieve the above invention purpose, the technical scheme adopted in this invention is: An unknown environment mapping method for a single-beam laser of aircraft, the aircraft is configured with a single-beam LiDAR, a stepper motor, and a binocular camera, and point cloud data of the single-beam LiDAR is aligned with stepper motor data on a timestamp. The method includes the following steps: S1, denoising point cloud data collected at a current position, and adjusting position and attitude information of an aircraft by using denoised point cloud data, so as to unify a single-beam LiDAR coordinate system and the aircraft;S2, based on a unified coordinate system, aligning the point cloud data at adjacent positions initially, and performing a feature extraction, a point cloud alignment and a point cloud stitching in turn to generate a unified point cloud map;S3, for point cloud data in the unified point cloud map, establishing a local neighborhood of each point cloud according to a K-nearest neighbors search algorithm, and determining a local tangent space of each point cloud in a local neighborhood by a principal component analysis method, using a multi-dimensional scaling analysis method to integrate information of all local tangent spaces, and obtaining a global parameterization of the point cloud data on a Riemannian manifold;S4, based on point cloud data represented by the global parameterization, generating a continuous three-dimensional manifold surface according to a Delaunay triangulation method, and generating a local map of the current position after a smooth and curvature adjustment of the continuous three-dimensional manifold surface;S5, in a process of aircraft data acquisition, using a loop detection algorithm to correct a pose of the aircraft, and then optimizing a local map to obtain an optimized local map;S6, taking a current optimized local map as a starting point, performing a boundary exploration and a path planning, and collecting data according to a planned path, repeating S1-S5 to gradually generate an optimized local map, and then completing an unknown environment mapping. In some embodiments, S1 includes the following steps: S11, calculating a distance between the point cloud data collected at the current position and its adjacent point cloud;S12, according to an average value and standard deviation of each distance, determining a