摘要
随着监控摄像头的普及,智能监控设备迎来发展热潮。如在人流量大的地点人脸识别监控等。但是,由于人脸识别要求视频的清晰度高且在特定角度才能够识别,所以在很多场景下的效果并不理想。行人再识别技术通过行人的外观特征等对目标人物进行匹配,提高准确性。基于行人再识别的轨迹重现是利用多个相机间目标人物的匹配结果及预先定义的相机坐标,输入至百度轨迹上传API得到目标人物的运动轨迹。
With the popularity of surveillance cameras,intelligent surveillance equipment has ushered in a development boom.Such as face recogni⁃tion monitoring in places with high traffic.However,since face recognition requires high definition of video and only can be recognized at certain angles,the effect in many scenes is not satisfactory.Re-Identification improves the accuracy on matching the target characters by the appearance characteristics of the pedestrians.The trajectory reproduction based on Re-Identification is to use the matching result of the target person between multiple cameras and the predefined camera coordinates,and input to the Baidu map trajectory upload API to obtain the motion trajectory of the target person.
作者
闫秋芳
YAN Qiu-fang(College of Computer Science,Sichuan University,Chengdu 610065)
出处
《现代计算机》
2020年第4期51-54,共4页
Modern Computer
关键词
行人再识别
行人检测
步态识别
轨迹重现
Person Re-id
Pedestrian Detection
Gait Recognition
Trajectory Recurrence
