针对道路现场实时车流量检测问题,提出了一种改进的帧间差分法的运动车辆检测算法，并将该检测算法成功移植到了嵌入式系统上。将帧间差分法与采用长度、宽度、面积筛选轮廓及用质心距离的车辆跟踪算法结合，实现运动车辆的检测；将U-Boot引导程序、Linux内核、Yaffs2文件系统和检测算法移植到S3C6410上，通过摄像头实时采集交通视频，检测结果由触摸屏显示。复杂交通场景的实时测试结果表明,本系统的检测时间为0.298秒/帧，准确率超过88%，基本能够实现在道路现场的车流量实时检测。

Aiming at the problem of real-time traffic flow in the road, an improved algorithm for moving vehicle detection based on frame subtraction was proposed, and the algorithm was successfully transplanted into the embedded system. The Frame subtraction was combined with vehicle tracking algorithm using length, width, area selection contour and centroid distance to realize the detection of moving vehicle. The U-boot program, Linux kernel, Yaffs2 file system and detection algorithm were transplanted to the S3C6410, and the traffic video was collected through the camera, and the detection results were displayed by the touch-screen. The real-time test results of complex traffic scene show that the detection time of the system is 0.298 seconds/frame, and the accuracy rate is over 88%, which can basically realize the real-time vehicle flow detection in the road scene.