Abstract:

The false detection rates of existing methods for object detection and tracking based on geometric characteristics are high, and missed detection can lead to wrong object association in the process of object tracking. Targeting at aforementioned problems, a vehicle detection and tracking method is presented based on range data using Light Detection And Ranging (LiDAR). In terms of the characteristics of the raw range data, the data is processed by a grid-based clustering algorithm with the parameter automation (PAG), where the line segments are extracted within each cluster to gain object features. On this basis vehicle targets are identified, and the position of the vehicles could be calculated. The object association and state estimation are accomplished by using a Kalman Filter combining with a filter management strategy. Finally, the proposed method is evaluated with the vehicle equipped with a forward-looking LiDAR sensor. The results show that the method proposed can detect and track multiple vehicle objects accurately, and wrong associations can be avoided.

Key words: intelligent transportation, detection and tracking, feature extraction, vehicle object, range data, Kalman Filter

摘要：

现有基于几何特征的目标检测与跟踪方法误检率较高，目标跟踪过程中的漏检易导致错误的目标关联.针对这些问题，本文提出了一种基于激光雷达(LiDAR)深度数据的车辆目标检测与跟踪方法.根据激光雷达深度数据特性，采用一种基于栅格的参数自动化聚类(PAG) 算法对原始数据进行处理，并在每个聚类中提取目标线段，获取目标特征.在此基础上对车辆目标进行识别，并计算得到目标的位置信息.采用卡尔曼滤波算法，制定滤波器管理策略，完成目标关联及状态估计.最后利用装备有一个前向激光雷达的实验车辆对提出的方法进行验证. 实验结果表明，本文提出的方法可准确识别并跟踪多个车辆目标，避免错误的目标关联.

关键词: 智能交通, 检测与跟踪, 特征提取, 车辆目标, 深度数据, 卡尔曼滤波