智慧国土视频监控运动目标提取新方法研究

A New Method for Moving Objects Detection in Wisdom Territory Video Surveillance

近年来,由于国土资源违法违规现象频发,视频监控在国土行业得到了广泛应用:它是通过监控设备对耕地、矿区等重点区域进行实时监控从而自动识别、判断、预警的新执法监察手段。自动化和智能化的实现,迫切需要对运动目标进行有效提取、分析。针对在智慧国土视频监控中获得广泛应用的PTZ(Pan-Tilt-Zoom)相机,提出了一种快速有效的运动目标提取新方法。与传统建立全景背景模型的PTZ相机运动目标提取方法不同,该方法建立了匹配特征点的密度估计,减少了计算时间与内存需要,且对光线变化有很好鲁棒性,可以方便地应用到"智慧国土"项目——视频监管系统建设中去。

In recent years,due to the frequent violation phenomenon of land and resources,video monitoring as a new law enforcement supervisory method,has been widely applied in industry,which is a new law enforcement supervision means by monitoring the equipment of cultivated land,mines and other key areas to automatic identification,judgment and warning.The realization of automation and intelligence urgently needs to extract and analyze the moving target effectively.Focused on PTZ（Pan-Tilt-Zoom）camera which is widely used in smart land and resource video monitoring,a new effective method for moving target extraction is proposed.Different from the tradition method which extracted the moving target based on PTZ camera with panoramic background model,the moving target from establishing density estimation of matching feature points is extracted by this method,which greatly reduce the computation time and memory requirements.And it has good robustness to light changes,which can be conveniently applied to the construction of smart land and resources video monitoring.