Abstract: 1 m-resolution satellite video has been able to achieve the real-time monitoring of the tiny ground moving target. Concerning to the features of the vehicles in satellite video date that only contain one or several pixels, this paper presents a method to estimate the traffic flow parameter from road satellite video date based on optical flow. The method firstly uses the Shi-Tomasi corner detection method to realize the vehicle detection and vehicle counts according to the vehicle target features that show as pixels in video date, and then calculates the bi-directional average traffic flow speed through the corner location information derived from optical flow in consecutive video frames. After that, a comparative analysis is made for the result. It’s an attempt to estimate the traffic flow parameter of the tiny vehicle target from satellite video data.

Key words: satellite video, traffic flow, optical flow, velocity

摘要： 抽样分辨率达1米的高清卫星视频已经能够实现对地面较小的运动目标的实时监控。针对卫星视频中运动车辆目标仅显示为一个或几个像素点的特点，提出了一种基于光流法的卫星视频交通流参数提取的思路与方法。该方法利用卫星视频中车辆目标为像素点的特点，结合Shi-Tomasi角点检测方法实现车辆检测及车辆计数；在车辆检测的基础上利用光流法得到的连续视频帧中角点的位置信息进行双向车辆平均车速的计算，并对实验结果进行了对比分析。该文是基于卫星视频中小微运动车辆目标进行交通流参数提取的一次有益尝试。

关键词: 卫星视频, 交通流, 光流法, 速度