基于改进LK光流法的车流量检测

Traffic Flow Detection Based on Improved Lk Optical Flow Method

为了实现基于交通监控视频的车辆检测,提出了一种基于LK光流法与分水岭算法相结合的车辆检测方法,利用LK光流法生成车辆光流数据,对光流数据进行处理生成车辆前景图像,对前景中产生的车辆粘连的情况采用改进分水岭算法完成分割.实验结果表明,该方法一定程度上避免了LK光流法因车辆粘连而造成的误判断情况,较现有方法有所提升,有较好的识别效果.

A traffic flow detection method based on LK optical flow and watershed method has been proposed in order to realize vehicle detection based on traffic monitoring video.LK optical flow method was used to generate vehicle optical flow data which were processed to generate vehicle foreground image.The generated contour was processed to eliminate the void of the fault and divide merged vehicles by watershed algorithm.The experimental results show that the method can avoid false detection and improves the successful recognition rate of LK optical flow method and obtains better results.