摘要
针对激光雷达在复杂场景中检测多目标时所遇到的传感器本身数据抖动、目标相互遮挡导致目标检测准确率不高的问题,提出了一种基于置信滤波的多目标检测方法。本方法将Velodyne点云数据投影到地面,使用置信滤波方法将聚集区域边缘的不稳定栅格滤除,并将栅格分为动态与静态两类。最后使用改进的DBSCAN聚类算法,对动、静态栅格进行聚类处理,得出运动目标与静态目标。实验结果表明:该方法能够较大地提高对复杂场景下多目标的检测准确度,是一种高效的激光雷达多目标检测方法。
Aiming at the questions happened when laser radar detecting multi-target in the com- plex environment that the date of laser sensor itself jittered and the accuracy of the target detec- tion result caused by the mutual occlusion of the target is not high, a multi-target detection method based on the confidence filter is presented. The method projects the Velodyne point data to the ground, using the theory of the confidence filtering to filter the unstable grids in the edge of the point clusters, then dividing the grids into two types： dynamic and static. Fi- nally, an improved DBSCAN clustering algorithm is used to cluster the dynamic and static grids separately to obtain the dynamic target and the static target. The experiment results show that the proposed method can improve the detection accuracy of multi-targets under complex terrain scenes. It is an efficient laser radar-based multi-target detection method.
出处
《制导与引信》
2017年第2期25-29,共5页
Guidance & Fuze
基金
2016上海航天科技创新基金资助项目
2015航天支撑技术基金资助项目
