摘要
为了检测与跟踪道路上的目标,提出了一种基于多层激光雷达的目标检测与跟踪算法。利用地平面拟合算法将点云数据进行地面滤除,将过滤后的点栅格化并利用区域增长算法进行目标聚类,根据RANSAC算法构造的BBox得到的目标信息,利用改进的匈牙利算法建立对应的跟踪器。利用卡尔曼滤波算法实时更新目标的状态。改进的匈牙利算法能够解决非平衡目标指派到的问题,适应环境中目标突然增多或者减少的情况,以数据集论证了该算法的可靠性。
In order to detect and track the target on road,an algorithm of target detection and tracking based on multi-layer lidar is proposed. Collected points cloud data is filtered out using ground plane fitting algorithm,the filtered points are rasterized and clustered by region growing algorithm. The information of targets is obtained from bounding box(BBox) constructed by random sample consensus(RANSAC) algorithm and build corresponding tracker by improved Hungarian algorithm. Updating the state of the target using the Kalman filtering algorithm. The improved Hungarian algorithm can solve the problem of non-equilibrium target assignment and adapt to the sudden increase or decrease of targets in the environment. The reliability of the algorithm is demonstrated by the KITTI data set.
作者
李帅印
段建民
冉旭辉
LI Shuaiyin;DUAN Jianmin;RAN Xuhui(Department of Information Science,Beijing University of Technology,Beijing 100124,China)
出处
《传感器与微系统》
CSCD
2020年第6期123-126,共4页
Transducer and Microsystem Technologies
基金
北京市教委基金资助项目(JJ002790200802)
北京市属高等学校人才强教计划资助项目(038000543115025)。
关键词
目标检测与跟踪
地平面拟合算法
RANSAC算法
匈牙利算法
卡尔曼滤波
target detection and tracking
ground plane fitting algorithm
random sample consensus(RANSAC)algorithm
Hungarian algorithm
Kalman filtering