基于三维模型的交通场景视觉监控

视觉监控是计算机视觉研究的前沿方向.动态场景视觉监控就是利用计算机视觉和人工智能的理论和方法.通过对摄像机拍录的图像序列进行自动分析来对场景中的运动物体进行定位、跟踪和识别,并对物体的运动行为作出判断或者解释,达到监控的目的.本文结合交通场景监控这一特定任务,实现一个包括摄像机标定、模型可视化、运动车辆的姿态优化与定位、跟踪预测、基于轨迹分析的行为理解等功能算法的交通场景视觉监控系统.从算法和实现的角度出发，文章对系统中各个功能模块进行了较为详细的描述与讨论.

3-D Model Based Visual Traffic Surveillance

Visual surveillance in dynamic scenes is an active research topics in computer vision.The aim of visual surveillance is to make it possible that the computer can watch or monitor ascene by automatic localization, tracking and recognition of moving objects and semantic interpretationof their behaviors in the watched scene. This paper aims to realize a task-specific traffic surveillancesystem which consists of modules for camera calibration, model visualization, pose refinement,tracking, trajectory-based semantic interpretation of vehicle's behaviors, etc. In thispaper, we describe each module to give readers a comprehensive view of a visual surveillance system,and also discuss possible further work.