摘要
为提高车辆跟踪的准确性,满足智能交通监控的需求,采用样本车型模板结合尺度不变特征变换(SIFT)算法匹配检测车辆位置,并将模板大小与车辆尺度变化联系起来,根据特征向量距离最终获得车辆准确区域;采用结合均值漂移的粒子滤波算法对车辆进行跟踪,根据跟踪尺度改变跟踪窗口大小,并通过独立粒子滤波建立了多运动车辆之间的数据关联。实际道路测试结果表明:该算法的车辆检测准确率达到90%以上,特征粒子在后续跟踪过程中状态稳定,漂移在跟踪窗口外的粒子数都保持在10%以下。
In order to improve the accuracy of vehicle tracking on road,and satisfy the need of intelligent traffic monitoring,a sample template model with scale invariant feature transform（SIFT） algorithm was used to match and detect the vehicle location,and the size of template was linked to vehicle scale changing.According to the distance of feature vector,the exact region of vehicle could be obtained.The particle filter algorithm combined with mean-shift was used to track the vehicles and the size of tracking window was changed according to vehicle scale changing.The data association between multiple moving vehicles was established by independent particle filter.Actual road test results show that the vehicle detection accuracy calculated by the algorithm is above 90%.Feature particles in the follow-up tracking process stay in stable condition,the number of particle which drifts outside the window is below 10%.
出处
《中国公路学报》
EI
CAS
CSCD
北大核心
2010年第3期89-94,共6页
China Journal of Highway and Transport
基金
国家自然科学基金项目(60772080)
天津市应用基础与前沿技术研究计划基金项目(09JCYBJC07700)
关键词
交通工程
智能交通
粒子滤波算法
车辆跟踪
特征匹配
traffic engineering
intelligent traffic
particle filter algorithm
vehicle tracking
feature matching
