基于OpenCV的公交客流计数方法

摘要

The bus passenger flow statistics is an important research direction in computer vision field. However,the influ?ence of dynamic texture factors such as shadow and illumination variation makes the construction of a real?time accurate bus pas?senger flow statistic algorithm difficult. To solve this problem,a bus passenger counting method based on OpenCV is presented. The obtained real?time image is preprocessed. The method of combining Codebook background model with morphological filtering is adopted to build the the background model to eliminate the influence of dynamic texture on the foreground. The improved Hough transform is used to extract the head region of the passengers for tracking counting. The experiment results show that the calculation accuracy of the algorithm can reach more than 88.6%,so it can be applied to the real?time passenger counting sys?tem.%公交车客流统计是计算机视觉领域的重要研究方向，然而由于阴影、光照变化等动态纹理因素的影响使得构建一个实时精确的客流统计算法变得较为困难。针对以上难题，提出一种基于OpenCV实现单目视觉下的客流统计方法。首先对取得的实时图像进行预处理；然后运用Codebook和形态学滤波相结合的方法建立背景模型，消除动态纹理对前景的影响；最后采用改进的Hough变换提取乘客的头部区域，进行跟踪计数。实验结果表明，乘车环境中该方法的检测准确率达88.6%以上，可应用于实时的客流计数系统。