LiDAR Dynamic Target Detection Based on Multidimensional Features

Aigong Xu,Jiaxin Gao,Xin Sui,Changqiang Wang,Zhengxu Shi

DOI: https://doi.org/10.3390/s24051369

IF: 3.9

2024-02-22

Sensors

Abstract:To address the limitations of LiDAR dynamic target detection methods, which often require heuristic thresholding, indirect computational assistance, supplementary sensor data, or postdetection, we propose an innovative method based on multidimensional features. Using the differences between the positions and geometric structures of point cloud clusters scanned by the same target in adjacent frame point clouds, the motion states of the point cloud clusters are comprehensively evaluated. To enable the automatic precision pairing of point cloud clusters from adjacent frames of the same target, a double registration algorithm is proposed for point cloud cluster centroids. The iterative closest point (ICP) algorithm is employed for approximate interframe pose estimation during coarse registration. The random sample consensus (RANSAC) and four-parameter transformation algorithms are employed to obtain precise interframe pose relations during fine registration. These processes standardize the coordinate systems of adjacent point clouds and facilitate the association of point cloud clusters from the same target. Based on the paired point cloud cluster, a classification feature system is used to construct the XGBoost decision tree. To enhance the XGBoost training efficiency, a Spearman's rank correlation coefficient-bidirectional search for a dimensionality reduction algorithm is proposed to expedite the optimal classification feature subset construction. After preliminary outcomes are generated by XGBoost, a double Boyer–Moore voting-sliding window algorithm is proposed to refine the final LiDAR dynamic target detection accuracy. To validate the efficacy and efficiency of our method in LiDAR dynamic target detection, an experimental platform is established. Real-world data are collected and pertinent experiments are designed. The experimental results illustrate the soundness of our method. The LiDAR dynamic target correct detection rate is 92.41%, the static target error detection rate is 1.43%, and the detection efficiency is 0.0299 s. Our method exhibits notable advantages over open-source comparative methods, achieving highly efficient and precise LiDAR dynamic target detection.

engineering, electrical & electronic,chemistry, analytical,instruments & instrumentation

What problem does this paper attempt to address?

This paper aims to address several key issues in LiDAR dynamic object detection. Existing LiDAR dynamic object detection methods typically rely on heuristic thresholds, indirect computation aids, supplementary sensor data, or post-processing techniques. These methods require different thresholds to be set in different scenarios to obtain effective detection results and often have certain limitations. To overcome these problems, the authors propose an innovative method based on multi-dimensional features.

Specifically, this method comprehensively evaluates the motion state of point cloud clusters by analyzing the differences in the position and geometric structure of the same target scanned in point cloud clusters in adjacent frames. To achieve automatic and precise pairing of point cloud clusters, the paper proposes a dual registration algorithm for the registration of point cloud cluster centroids. In the coarse registration stage, the Iterative Closest Point (ICP) algorithm is used for approximate inter-frame pose estimation; in the fine registration stage, the Random Sample Consensus (RANSAC) and four-parameter transformation algorithms are used to obtain the precise inter-frame pose relationship, thereby standardizing the coordinate systems of adjacent point clouds and promoting the association of point cloud clusters from the same target.

Based on the paired point cloud clusters, a classification feature system is constructed and trained using the XGBoost decision tree. To improve the training efficiency of XGBoost, the paper proposes a bidirectional search dimensionality reduction algorithm based on the Spearman rank correlation coefficient to accelerate the construction of the optimal classification feature subset. Additionally, to further enhance the final LiDAR dynamic object detection accuracy, the paper designs a dual correction scheme based on the sliding window strategy and Boyer-Moore voting strategy to perform secondary corrections on the preliminary detection results of XGBoost.

Experimental results show that this method has a high correct detection rate (92.41%), a low false detection rate of static objects (1.43%), and efficient detection efficiency (0.0299 seconds), demonstrating significant advantages compared to open-source comparison methods.