基于机器学习的单线激光雷达进行车辆识别与跟踪方法研究

Machine Learning Based Vehicle Recognition and Tracking Using Single Layer LIDAR

针对多线激光雷达应用于无人驾驶车辆的高额造价问题,提出了一种基于单线激光雷达作为主传感器,并利用机器学习方法实现车辆识别与跟踪任务。首先通过激光雷达扫描并获取空间轮廓数据,对每帧数据采用层次聚类算法进行目标分割;然后对每个目标进行特征提取,通过交叉验证和网格搜索对支持向量机的参数进行优化以实现更好的分类效果,采用卡尔曼滤波实现目标车辆的跟踪;最后在城市公路和高架桥上开展了数据采集的工作,使用了装备单线激光雷达和相机的乘用车作为实验平台,实验结果表明,所提取六种目标车辆特征值组成特征向量并配合参数优化后的支持向量机可以实现较高识别率,并可实现对目标车辆的稳定跟踪。

Considering the high cost of multi-layer LIDAR application for unmanned vehicles,in this paper,a vehicle detection and tracking method based on single-layer LIDAR using machine learning techniques is proposed. Firstly,the spatial profile data is acquired by LIDAR scaning. And then hierarchical clustering algorithm is applied to divide the target data for each frame. Furthermore,the features are extracted and the optimized parameters of Support Vector Machine(SVM) are found through cross validation and grid search to achieve better classification results. The Kalman filter is used to track the target vehicle. Finally,a passenger vehicle OptoBot-IV which is equipped with a single-layer LIDAR and a camera is driven on the urban area for the data collection. The experimental results show that this approach can achieve a reasonable high recognition accuracy and achieve stable tracking of the object vehicles.