Research And Implementation Of Real-time Plane Detection Algorithms In LiDAR Point Clouds

With the continuous development and advancement of science and technology,various new types of sensing devices are emerging.Because of its small size and accurate detection,LiDAR has gradually become the mainstream environmental information acquisition equipment in the market.The planar detection technology for three-dimensional point cloud of lidar is a hot topic in the field of three-dimensional reconstruction and intelligent robot positioning and navigation.Because of the high complexity of LiDAR point cloud data and the high rate of change of data per frame for different scanning environments indoor and outdoor,the traditional three-dimensional point cloud plane detection algorithm is difficult to accurately and real-time detect the plane,and has poor adaptability to different indoor and outdoor environments.Based on the above problems,the GPU-based RANSAC plane detection algorithm is proposed in this paper.In order to reduce the complexity of scanning scene,the point cloud data is simplified by point cloud denoising and point cloud segmentation.In order to speed up the efficiency of plane detection,based on the traditional RANSAC algorithm,the LiDAR point cloud data is optimized according to its characteristics.Based on the traditional RANSAC algorithm and the characteristics of LiDAR point cloud data,a multi-plane random sampling consistency algorithm is proposed,and the algorithm is calculated by GPU.Finally,some planar models are merged.The overall algorithm uses CPU+GPU framework,which greatly improves the accuracy and efficiency of plane detection.Based on the GPU-based RANSAC plane detection algorithm,a real-time LiDAR point cloud plane detection system is studied.The main functions of the system include:point cloud data importation;LiDAR point cloud reduction;plane detection based on GPU multifaceted random sampling consistency algorithm;plane model merging;plane model visualization.The system realizes real-time plane detection for different types of LiDAR point clouds indoors and outdoors.At the same time,LiDAR point cloud plane detection method based on traditional RANSAC algorithm and LiDAR point cloud plane detection method based on three-dimensional Hough transform algorithm are studied and implemented.Through performance comparison and analysis,it is proved that our LiDAR point cloud real-time plane detection system has high accuracy and high efficiency.