一种基于帧间差分和光流技术结合的运动车辆检测和跟踪新算法

摘要

This paper intends to improve the detection and tracking method of moving vehicle in intelligent transportation. A new method for moving vehicle detection and tracking baaed on the improved combination of frame difference and optical flow technique is pre&ented. In this method, the motion area of moving objects is detected by the difference of two consecutive frames subtraction of moving objects, then the optical flow at non-zero place of difference chart is calculated, and its optical flow field is used to achieve tracking the moving target. In order to reduce the amount of calculation, the paper presents an optical flow field calculation method based on point matching technique with optimal estimation and optical flow uniform sampling method. By the processing of adaptively segmenting the threshold of optical flow after greying and morphological filtering, the real-time moving target detection and tracking have been realised.%为了改进智能交通中的运动车辆检测和跟踪方法,提出一种基于改进的帧间差分和光流技术结合的运动车辆检测和跟踪的新方法.先用帧间差分法检测出运动物体的运动区域,再计算差值图中不为零处的光流,然后利用其光流场来实现运动目标的跟踪.为了减少计算量,提出一种基于最优估计的点匹配技术和光流均匀采样策略的光流场计算方法,并通过对灰度化后的光流场进行自适应阈值分割、形态学滤波等处理,实现了实时的运动目标检测和跟踪.