基于梯度方向恒定性的运动车辆阴影检测

交通参数的视频检测是智能交通系统的一个研究重点，其中运动车辆的分割是视频检测过程中的一个关键环节。目前，运动车辆阴影的检测与剔除是准确、有效的分割出运动车辆所面临的一个难题。本文发现并证明了梯度方向恒定性原理，在此基础上提出了一种基于梯度方向恒定性的阴影检测与剔除方法。首先建立路面背景梯度矢量图，根据与当前帧图像的梯度矢量图的比较结果，判断是否是路面背景或是运动车辆，然后对运动车辆区域进行形态滤波，弥补内部空洞和剔除杂点，进而准确分割出车辆。试验结果表明，该方法适应性强，车辆分割效果好。

Method of Detection and Removing Shadow of Moving Vehicle Based on the Direction Constancy Principle of the Texture Gradient

Moving vehicles segmentation is the key step of detection by video which is important research in intellectual traffic system. At present , detecting and removing the shadow of moving vehicle is a difficult problem in segmentation the moving vehicle accurately and effectively. The paper discover and proof the direction constancy principle of the texture gradient, and propose a method of detection and removing shadow of moving vehicle based on it. Set up the background gradient vectograph of highway surface at first, according to the comparative result with the gradient vectograph of the present frame image, judge whether it is highway surface background or vehicle , then carry on morphological filtering which can remedy the inside cavity and reject isolated points, and the vehicle is segmented accurately last. The result of the test indicates this method has a good adaptability and effectually.