联合时空特征的车辆跟踪  

Tracking Vehicles Based on Spatio-Temporal Analysis

作　　者：陈柏生[1] 陈锻生[1] 

机构地区：[1]华侨大学信息科学与工程学院,福建泉州362021

出　　处：《华侨大学学报（自然科学版）》2008年第2期222-224,共3页Journal of Huaqiao University(Natural Science)

基　　金：福建省自然科学基金资助项目(A0510020);福建省青年科技人才创新项目(2006F3086)

摘　　要：提出一种联合空间特征和时间特征的匹配准则,以实现对道路车辆的跟踪.计算目标区域与候选区域的加权累积灰度分布差异,结合运动的路径连续性特征来构造运动模式相关函数,以此作为目标匹配的判据.引入虚轨迹点补偿由于目标遮挡和检测失败等导致丢失的运动轨迹点,以实现对目标连续的、完整的跟踪.实验结果表明,算法能够实现多车辆的鲁棒跟踪,并且能够较好地解决目标相互遮挡下的跟踪问题.A scheme combining temporal and spatial features to track vehicles is proposed. By calculating a weighted sum of absolute intensity difference between pixels in target region and the corresponding pixels in the candidate region, motion mode correlation function is constructed with integration of path deviation function to perform the task of object matching. Phantom points are introduced to compensate the lost tracking points causing by the occlusion or object detection failure in order to maintain the trajectory complete. Experimental results show that the method can realize multiple vehicles robust tracking simultaneously, and track the vehicles under occlusion successfully.

关 键 词：灰度分布差异 路径偏差函数 运动模式相关度 车辆跟踪 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP75[自动化与计算机技术—计算机科学与技术]