一种视频检测车辆位置违章的几何方法

Geometrical Method for Video-based Detecting Position Violations of Vehicles

针对交通视觉监控系统中普遍存在的遮挡与三维信息缺失问题,用一种三维矩形近似描述模型对车辆建模,基于此模型推导出在自遮挡和交互遮挡情况下,车辆整体轮廓、底部轮廓和单一底边的恢复规则和方法.提出将车辆运动视为其车辆底部在道路平面上的二维共面运动,通过分析车辆底部与道路标线的相对位置关系,得到位置违章的两个检测判据与检测算法.实验结果表明,该几何模型能很好的逼近车辆底边,有效的解决了遮挡的影响;该位置违章检测算法避免了以往质心检测判据与区域判据的缺陷,检测效果很好.

Occlusion and 3-D information lost widely exists in visual traffic surveillance system. By using a novel approximate 3- D rectangular vehicle model, this paper analyzes the situation of serf occlusion and mutual occlusion, proposes the rules of occlusion resolvability. Assuming that the vehicle＇s bottom and the road marks were coplanar on the road plane, by analyzing the positional relativity of vehicle＇s bottom edges and the road marks, this paper gains two detecting criterions, and proposes an algorithm for detecting position violations. Experimental results show that, by preferably approximating the bottom of the vehide, the geometrical model can resolve the influence of occlusions ; the presented detecting method is better than the centroid method and the area method.