一种基于交通视频车辆阴影去除算法的研究

摘要

Vehicle identification is the major research subject in intelligent transportation system,and the shadow is the main factor affecting vehicle identification.According to the optical characteristics of vehicle shadow formation,we proposed an automatic marking-based vehicle shadow removal algorithm with OSTU dual-threshold image enhancement.First,it uses the improved statistical histogram background update algorithm to update the background and to acquire the background images.Then it obtains the foreground binary image with moving shadow by background subtraction operation.Finally,it uses the improved automatic marking algorithm to mark the connected region of the derived binary images,and then segments the actual image corresponding to every marked connected region by OSTU dual-threshold image segmentation algorithm;and the gray enhancements are applied to the obtained segmented regions in different degrees.In experiment of video detection,not only the algorithm can effectively eliminate the moving shadow,but the problem of false shadow detection is also overcome to certain extent.Experimental results showed the effectiveness of the improved method proposed in the paper in shadow removal.%车辆识别是智能交通系统的主要研究课题，而阴影是影响车辆识别的最主要原因。根据车辆阴影形成的光学特征，提出一种基于自动标记的 OSTU 双阈值图像增强车辆阴影去除算法。首先使用改进的统计直方图背景更新算法进行背景更新，获取背景图像。其次使用背景差分运算求取包含移动阴影的前景二值图像。最后使用改进的自动标记算法对所得二值图像的连通区域进行标记，然后分别对每一个标记的连通区域所对应的实际图像使用 OSTU 双阈值图像分割方法进行分割，并对所得的分割区域进行不同程度的灰度增强。在视频检测实验中，该算法不仅能够有效消除移动阴影，而且在一定程度上克服了阴影误检等问题。实验结果表明，该改进方法在阴影消除方面的有效性。