基于深度学习的多摄像头协同目标跟踪方法研究
首发时间:2021-01-20
摘要:目前,一般的公共安全监控系统采取在监控区域布设大量摄像头设备的方案以实现对大范围区域的监控。通常这些摄像头是离散的,并且拍摄的监控信息缺乏快速完善的整合机制,导致视频信息的处理仅基于图像,最终未能有效结合视频监控系统的空间位置信息。该文章针对目前视频监控系统存在的不足,并基于摄影测量学、地理信息系统以及深度学习等理论方法,在目标检测、目标定位以及多摄像头协同追踪三个方面对视频监控系统提出了改进方案。进而实现在多摄像头视频监控场景下的目标检测与跟踪,最后通过实验验证了提出的改进多摄像头协同目标跟踪方法对特定目标实时跟踪的有效性。
关键词: 智能视频监控系统 多摄像头协同 坐标映射 目标检测 目标跟踪
For information in English, please click here
Research on Multi-camera Collaborative Target Tracking Method Based on Deep Learning
Abstract: At present, the general public safety monitoring system adopts a scheme of deploying a large number of camera devices in the monitoring area to achieve monitoring of a large area. These cameras are usually discrete, and the captured monitoring information lacks a fast and comprehensive integration mechanism, resulting in the processing of video information based only on images, and ultimately fails to effectively combine the spatial position information of the video surveillance system. In view of the shortcomings of current video surveillance systems, this paper proposes an improved scheme for video surveillance systems in terms of target detection, target location and multi-camera collaborative tracking based on photogrammetry,Research on Multi-camera Collaborative Target Tracking Method Based on Deep Learning GIS and deep learning. In this way, the target detection and tracking in the multi-camera video surveillance scene is achieved. Finally, the effectiveness of the improved multi-camera collaborative target tracking method for real-time tracking of specific targets is verified by experiments.
Keywords: intelligent video surveillance system multi-camera collaboration coordinate mapping target detection target trackingkey
引用
No.****
动态公开评议
共计0人参与
勘误表
基于深度学习的多摄像头协同目标跟踪方法研究
评论
全部评论0/1000