结合双向光流约束的特征点匹配车辆跟踪方法
发布时间：2018-11-15 13:51
【摘要】：针对复杂交通场景中动态光照变化、目标尺度变化和部分遮挡等因素带来的影响,提出了一种基于特征点的稳定可靠的车辆跟踪方法.针对运动车辆高速行驶时具有较大帧间运动的特点,构造KLT算法的金字塔模型,根据前向和后向跟踪偏移量,对稳定性较差的特征点进行剔除.同时,采用SURF特征匹配算法对目标特征点集进行更新和校正.最后,利用特征点之间的位置信息,确定目标的尺度和旋转变化因子,从而实现当前帧中目标区域的定位.实验结果表明,提出的车辆跟踪方法可以有效地解决复杂场景中目标形变和部分遮挡等问题,对尺度和旋转变化也具有较强的鲁棒性.
[Abstract]:A stable and reliable vehicle tracking method based on feature points is proposed to deal with the effects of dynamic illumination variation target scale change and partial occlusion in complex traffic scenarios. In view of the large inter-frame motion of moving vehicles at high speed, the pyramid model of KLT algorithm is constructed, and the feature points with poor stability are eliminated according to the forward and backward tracking offsets. At the same time, the target feature set is updated and corrected by SURF feature matching algorithm. Finally, using the position information between the feature points, the scale of the target and the rotation change factor are determined, so that the location of the target region in the current frame can be realized. Experimental results show that the proposed vehicle tracking method can effectively solve the problems of target deformation and partial occlusion in complex scenes, and is robust to scale and rotation changes.
【作者单位】： 西安石油大学计算机学院;
【基金】：国家自然科学基金(61572083)~~
【分类号】：TP391.41
[Abstract]:A stable and reliable vehicle tracking method based on feature points is proposed to deal with the effects of dynamic illumination variation target scale change and partial occlusion in complex traffic scenarios. In view of the large inter-frame motion of moving vehicles at high speed, the pyramid model of KLT algorithm is constructed, and the feature points with poor stability are eliminated according to the forward and backward tracking offsets. At the same time, the target feature set is updated and corrected by SURF feature matching algorithm. Finally, using the position information between the feature points, the scale of the target and the rotation change factor are determined, so that the location of the target region in the current frame can be realized. Experimental results show that the proposed vehicle tracking method can effectively solve the problems of target deformation and partial occlusion in complex scenes, and is robust to scale and rotation changes.
【作者单位】： 西安石油大学计算机学院;
【基金】：国家自然科学基金(61572083)~~
【分类号】：TP391.41
【相似文献】
相关期刊论文 前10条
1 卞建勇;徐建闽;裴海龙;;基于强化学习的视频车辆跟踪[J];华南理工大学学报(自然科学版);2008年10期
2 张晓媚;陈伟海;刘敬猛;赵志文;;基于柔性曲杆的车辆跟踪算法设计与实现[J];北京航空航天大学学报;2011年07期
3 甘玲;潘小雷;;一种应用于交通环境中的运动车辆跟踪方法[J];重庆邮电大学学报(自然科学版);2013年03期
4 郭锋;王秉政;杨晨晖;;复杂背景下车辆跟踪的改进算法及逆行检测[J];图学学报;2013年04期
5 魏玉强;王成儒;;多车辆跟踪时分割粘连车辆的方法[J];电视技术;2009年11期
6 王文龙;李清泉;;基于蒙特卡罗算法的车辆跟踪方法[J];测绘学报;2011年02期
7 李t，
本文编号：2333484
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2333484.html
