Abstract: A algorithm is developed for counting vehicles. The approach includes calculating and updating the background image correctly, extracting the vehicle regions with a self-adaptive threshold, removing the shadows, segmenting different vehicles, determining the vehicle number when the centers of vehicle regions are matched in several uninterrupted images. This algorithm is not affected by divided roadways, and it also can diminish the man-made mistakes. Experimental results demonstrated that the algorithm can count the vehicles accurately in an outdoor circumstance.

Key words: background subtraction, shadow removal, flow measuring, seriate images processing

摘要： 本研究提出了一种车流量的统计算法，处理步骤包括：提取与更新道路背景图像、使用自适应阈值提取运动车辆、去除车辆阴影、区分不同车辆和利用车辆中心的连续匹配来确定车辆数目，该算法不受车道的限制，人为影响小，计数准确性高。实验结果表明，本算法能够适应道路背景的多种变化，实现车辆的准确计数。

关键词: 背景提取, 阴影去除, 流量检测, 连续图像处理