有约束条件的监控影像解析方法

Analysis method of constrained sequential monitoring images

根据监控影像可分解为内、外方位元素相同且完全重叠的序列影像的特点,给出一种基于自检校的单目监控录像序列影像解析方法,并还原多目标空间几何关系。使用车载移动三维检校场,自动提取图像中特征点并检校相机;以直接线性变换反变换的数学模型为基础,以路面方程为约束,求得监控范围内事故现场中各时刻各目标的特征点在物方三维空间坐标系下的坐标;辅以目标的几何特征,测算出车辆与车辆、车辆与行人、车辆/行人与周围标志物在三维空间坐标系下各自的大小、位置、方向及其互相的空间几何关系,为交通事故勘查鉴定提供可靠依据。

Considering that the images of road monitoring system can be decomposed into sequential images with the same inner and exterior orientation elements and completely overlapped,a self-calibration-based image analysis method for sequential images from monocular monitoring video is constructed,and the spatial geometric relationship of multiple targets is restored.The mobile 3D calibration field is used to extract feature points automatically and calibrate the camera.Under the constraint by the road surface equation,the mathematical model of the inverse transformation of DLT(Direct Linear Transformation)is used to obtain the coordinates of the characteristic points of the targets at each time in the accident scene under the object three-dimensional space coordinate system in the monitoring area.With the given geometric characteristics of the target,the individual size,position,direction and spatial geometry relationship between vehicles and vehicles,vehicles and pedestrians,vehicles/pedestrians and surrounding markers are calculated in the three-dimensional space coordinate system,which provides a reliable basis for traffic accident identification.