基于视频的夜间车辆检测与跟踪

Video-based Nighttime Vehicle Detection and Tracking

在道路交通管理中基于视频的车辆检测技术发挥了越来越重要的作用。针对夜间交通视频图像中由于照明度低和光线反射干扰导致运动目标提取困难等问题,提出一种建立矩形框来标志车辆的夜间车辆检测与跟踪的方法。通过对图像进行预处理,提取可能为车灯的亮点,建立连通区域。利用两车灯之间的水平位置,两车灯的面积应该是相近或几乎相等以及两者之间的距离应该小于设定的阈值来进行车灯配对。车灯配对成功之后,适当放大配对车灯的连线长度,得到车头宽度。进而根据车头长宽比关系得到车头区域,再通过规则集来定义多种情况下矩形框保存车辆信息的基本原则。车辆的统计跟踪通过基于邻域的方法来实现。经过实验表明,该方法能很好地适用于夜间车辆的检测,并且能满足夜间检测的要求,具备一定的稳定性和准确率。

Video-based vehicle detection technology plays an increasingly important role in road traffic management. Aiming at the difficulty of moving target extraction in video images of night traffic due to low illumination and light reflection interference,a method of establishing rectangular frame to mark the vehicle flow detection and tracking at night is proposed. By preprocessing the image,the bright spots of the possible headlights can be extracted and the connected areas can be established. Using the horizontal position between the two lights,the area of the two lights should be close or almost equal and the distance between the two lights should be less than the set threshold value to carry out the lamp pairing. After the pairing of headlights is successful,the connecting length of paired headlights should be appropriately enlarged to obtain the width of the head. Then,according to the relation of the aspect ratio of the front end,the front end area is obtained. The statistical tracking of vehicles is realized by neighborhood based method. Experiment shows that this method is suitable for night vehicle detection,and can meet the requirements of night detection,with a certain stability and accuracy.