摘要
针对传统视频监控中依赖于人工分析数据判定动态目标运动轨迹浪费大量资源及实时性差的问题,提出一种基于物像距离转换匹配的行人轨迹判定方法。利用检测锚框和行人身高数据求取像—物平面之间的动态转换比例系数k并与视频中对应运动像元长度结合得出行人在真实地理空间中的运动距离进而与地图API路线测距对比判定行人运动轨迹。此外,动态比例系数k的变化趋势可以检测行人运动过程中是否发生徘徊行为从而对偏移路线进行纠正。实验证明该方法能够去除行人不规则运动对于轨迹判定的影响,为跨摄像头运动目标轨迹判定提供了一种新的解决方法。
In response to the problem that traditional video surveillance relies on manual analysis of data to determine the motion trajectory of dynamic targets,which wastes a lot of resources and has poor real-time performance,a pedestrian trajectory determination method based on object-image distance conversion matching is proposed.The dynamic conversion scale factor k between the image-to-object plane was obtained using the anchor frame and pedestrian height data and combined with the length of the corresponding moving image element in the video to obtain the distance of the pedestrian in real geospatial space and then compared with the map API route distance measurement to determine the pedestrian's movement.In addition,the trend of the dynamic scaling factor k allowed the detection of wandering behavior during pedestrian movement and the correction of deviations from the route.It is demonstrated that the method can remove the influence of irregular pedestrian movements on the trajectory determination,providing a new solution for the trajectory determination of cross-camera motion targets.
作者
李景文
孟效德
姜建武
龚旭
苏志鹏
Li Jingwen;Meng Xiaode;Jiang Jianwu;Gong Xu;Su Zhipeng(College of Geomatics and Geoinformation,Guilin University of Technology,Guilin 541004,Guangxi,China;Ecological Spatiotemporal Big Data Perception Service Laboratory,Guilin 541004,Guangxi,China;Guangxi Natural Resources Vocational and Technical College,Chongzuo 532100,Guangxi,China)
出处
《计算机应用与软件》
北大核心
2025年第11期144-150,190,共8页
Computer Applications and Software
基金
国家自然科学基金项目(41961063)
广西自然科学基金创新研究团队项目(2019GXNSFGA25001)。
关键词
行人识别
物像转换
目标交接
徘徊检测
Pedestrian recognition
Image conversion
Target handover
Wandering behavior detection
