一种基于运动目标检测的视觉车辆跟踪方法

针对复杂交通场景中动态光照变化、阴影和遮挡等因素带来的影响，提出了一种基于运动目标检测的高效、鲁棒的车辆跟踪方法. 采用自适应背景建模获取动态场景中的运动信息，通过阴影去除获得准确的运动区域，并针对场景中的遮挡问题提出了相应的遮挡检测与处理策略，最后通过区域匹配获得跟踪结果，同时使用Kalman滤波器建立车辆的运动模型，对跟踪结果进行了约束和优化. 实验结果表明，提出的视觉车辆跟踪方法可以在复杂多变的室外场景下有效地解决场景中的阴影和遮挡问题，得到鲁棒的车辆跟踪结果. 

Video-Based Vehicle Tracking Based on Moving Object Detection

Presented a robust and efficient video-based vehicle tracking method based on moving object detection to handle the dynamic illumination changes,shadow and occlusion in complex traffic scene. Firstly, an adaptive background modeling was adopted to detect the moving regions. Secondly, shadow reduction was used to obtain exact moving area. Then, an occlusion resolving strategy was proposed to handle the occlusion. Finally, vehicle tracking was achieved by region correspondence, combined with Kalman filters to model the motion of vehicles and optimize the vehicle parameters. The experimental results demonstrate that the proposed vehicle tracking method can effectively resolve shadow and occlusion problems in complex and variable outdoor scenarios,and robust vehicle tracking can be achieved.