监控视频动态目标的空间定位方法

Spatial location method of dynamic targets in surveillance videos

针对监控视频动态目标的空间定位问题,本文在考虑相机畸变的前提下,对监控相机、目标像素坐标和地理场景之间的映射关系进行了研究,提出了基于数字表面模型(DSM)和基于平面约束的目标定位算法。首先,完成相机的标定,确定相机的成像模型;然后,提取畸变校正后的目标像素坐标,并通过目标定位算法计算目标的三维地理坐标;最后,进行了定位精度评定,并分析了两种定位算法的应用特点。本文的定位算法将以像素坐标定位的目标数据解析为以三维坐标定位的空间地理信息,为多摄像头的目标跟踪提供了统一的地理参考框架。

Aiming at the problem of dynamic targets’ location in surveillance videos, this paper studies the mapping relationship among surveillance cameras, target pixel coordinates and geographical scenes on the premise of considering the camera distortion, and proposes the spatial location methods based on the digital surface model(DSM) and the plane constraint. Firstly, the camera calibration is completed and the imaging model of the camera is determined. Secondly, the pixel coordinates of the targets after distortion correction are extracted, and the 3 D geographic coordinates of the targets are calculated by the two spatial location methods. Finally, the positioning accuracy is evaluated, and the application characteristics of the two spatial location methods are analyzed. The methods proposed in this paper parse the target data located by pixel coordinates into the spatial geographic information located by 3 D coordinates, which provides a unified geographic reference framework for multi-camera target tracking.