摘要
为了对道路车辆进行流量的统计与监控跟踪,提出一种联合检测与跟踪思想的方法。该方法利用初始分割时产生的目标数量的冲突集描述分割阶段产生的错误以及遮挡问题,并通过建立车辆近邻关联事件和与之对应的关联标签变量,将汽车监控跟踪建模为一个结构化预测问题,利用相应的关联标签变量建立全局目标函数,从而将车辆跟踪问题转化为一个通过求解带约束的整数规划问题,最后求解得到车辆轨迹的全局最优解。
A joint detection and tracking method for traffic statistics and monitoring tracking of vehicles on road is proposed.In this method,the errors and occlusion problems produced in the segmentation stage are described by using the conflict set of the number of targets in the initial segmentation.The vehicle monitoring and tracking is modeled as a structural prediction mode by establishing the vehicle adjacent correlation event and the associated label variable corresponding to the event,and the global objective function is established by the corresponding associated label variables,so as to transform the vehicle tracking problem into an integer programming problem with constraint.The global optimal solution of vehicle trajectory is obtained.
作者
任亚婧
张宏立
REN Yajing;ZHANG Hongli(School of Electrical Engineering,Xinjiang University,Urumchi 830002,China)
出处
《现代电子技术》
北大核心
2019年第15期29-32,共4页
Modern Electronics Technique
基金
国家自然科学基金项目(51767022)
中国新能源汽车产品检测工况研究和开发项目资助~~
关键词
交通监控
随机森林分类器
联合检测跟踪
整数规划
结构化预测
支持向量机
traffic monitoring
random forest classifier
joint detection tracking
integer programming
structural prediction
support vector machine
