新能源汽车激光雷达传感器缺失数据填补方法研究

Research on Filling Method for Missing Data of LiDAR Sensor in New Energy Vehicle

为了增强车辆激光雷达传感器数据采集的全面性,研究新能源汽车激光雷达传感器缺失数据填补方法。利用数据融合的点云采集技术和中值滤波算法,预处理点云数据。采用改进的噪声密度聚类算法构建点云超体素块,建立图模型,并利用图割算法进行全局聚类。结合典型地物特征提取地物信息,并利用全景图像进行密集匹配填补缺失区域,以完成点云数据中空洞区域的填补。实验结果表明,该方法能够有效实现缺失数据的填补,并且填补效果良好。填补后的点云数据与缺失区域原始点云在深度方向上的分布状况几乎一致。

In order to enhance the comprehensiveness of vehicle LiDAR sensor data collection,a method for filling missing data in new energy vehicle LiDAR sensors is proposed.The method utilizes data fusion-based point cloud acquisition techniques and median filtering for preprocessing the point cloud data.An improved noise density clustering algorithm is employed to construct point cloud supervoxels,build a graph model,and perform global clustering using graph cuts.By extracting typical features of objects and densely matching the missing regions using panoramic images,the missing areas in the point cloud data are filled.Experimental results demonstrate that the proposed method effectively fills missing data with good results.The filled point cloud data closely approximates the depth distribution of the original point cloud in the missing regions.