利用光流法实现运动目标提取的研究

为了满足对于道路交通车流量的预测评估需求,需要实现对监控视频中运动目标的准确提取。论文提出了一种方案:采用LK高斯金字塔光流法和帧差法相结合的方法实现流动车辆的提取。LK光流法在相对理想化的条件下,无需提前了解场景信息,能够自己独自检测运动对象。但是LK光流法在实际跟踪环境中并不太理想,针对该问题,提出结合高斯金字塔和帧差法,提高光流法的跟踪性能。二者的巧妙结合恰好可以扬长避短,获取了较好的实验效果。该方法是在VS平台上用C++语言进行程序编写,并调用了OpenCV里面的光流法有关函数。实验结果表明,采用LK高斯金字塔光流法和帧差法相结合的方法可以准确统计当前待检测视频中流动车辆的数目,实用性较高,对道路交通的疏导和公共安全领域具有一定的参考价值。

Research on Moving Object Extraction by Optical Flow Method

In order to meet the demand for forecasting and evaluating traffic volume of road traffic,it is necessary to achieve accurate extraction of moving targets in the surveillance video.In this paper,a scheme is proposed.The method of combining LK Gauss pyramidal optical flow method and frame difference method is used to realize the extraction of flow vehicles.Under the rela tively idealized condition,the LK optical flow method can detect the moving object by itself without knowing the scene information in advance.However,the LK optical flow method is not ideal in the actual tracking environment.To solve this problem,the Gauss ian pyramid and frame difference method are proposed to improve the tracking performance of the optical flow method.The ingenious combination of the two can just make up for the weaknesses and gain good experimental results.This method is written in C++lan guage on the VS platform,and calls the relevant functions of the optical flow method in OpenCV.The experimental results show that the method of combining the LK Gaussian pyramidal optical flow method and the frame difference method can accurately count the number of mobile vehicles in the current video to be detected,h as high practicability,and has a certain reference value in the field of road traffic grooming and public safety value.