摘要
利用Kalman滤波思想对运动目标的前时刻状态信息进行预测,获取重心位置与形态紧密度估计值;将估计值与当前时刻观测值进行匹配,根据匹配误差修正运动目标的速度与紧密度变化值,通过递归算法实现常态下运动目标的准确、快速跟踪.针对复杂场景下由于运动遮挡造成无法准确估计目标运动轨迹,采用灰色模型GM(1,1)保证了跟踪过程的连续、稳定.最后,通过不同交通场景的视频序列对本文算法进行了验证,结果表明本文方法具有较好的适应性、鲁棒性,可实现复杂遮挡情况下连续、稳定、实时的目标运动跟踪.
Motion targets tracking is an important part of video surveillance,which offers technical support for researching motion characteristic and traffic behaviors of moving objects in traffic flow.Firstly,we predict the moving target state information of the former observation time by Kalman filter,and obtain the geometric center and the compactness of the object;then calculate the velocity and the compactness variety with matching errors between predicted values of the former time and observed values of the current time.From these step,we can achieve accurate,fast motion tracking results through recursive algorithm.In order to ensure the continuity and the stabilization in the tracking process,occlusion handling method based on gray model（GM（1,1）） is proposed.At last,we validate the proposed algorithm under different traffic scenes.Results show that the algorithm is robust and adaptive in multiple targets motion tracking even in the case of occlusion in real-time.
出处
《北京工业大学学报》
EI
CAS
CSCD
北大核心
2010年第12期1683-1690,共8页
Journal of Beijing University of Technology
基金
国家'863'计划项目资助(2009AA11Z210)
国家自然科学基金青年科学基金项目(50808092)
吉林省科技发展计划项目(20080432)
关键词
智能交通
视频监控
运动跟踪
特征匹配
KALMAN滤波
灰色模型
intelligence transportation
video surveillance
motion tracking
feature matching
Kalman filter
gray model
