一种稀疏光流计算方法.pdf

Transactions on Computer Science and Technology September 2013, Volume 2, Issue 3, PP.49-54 A Sparse Optical Flow Field Calculation Method † Naigong Yu, Li Xu , Xiaogang Ruan Department of electronic information and control engineering, Beijing University of Technology, 100124, P.R.China †Email: xulicindy@ Abstract Optical flow depicts the instantaneous velocity of image pixels between successive images and plays an important part in many computer vision tasks. So, the calculation of optical flow = has been the focus of research. However, most optical flow algorithms need to compute velocity at every image points, which is time-consuming and difficult to meet real-time requirements. That greatly hinders the application of optical flow. Furthermore, in some specific tasks, such as motion tracking and mobile robot obstacle avoidance, it is unnecessary to calculate full velocity field. Based on the above mentioned, a sparse optical flow calculation method was proposed in this paper. First of all, Shi-Tomasi corner detection was applied to find corner points with large eigenvalues, then LK optical flow algorithm to calculate optical flow at these points. Experimental results show that not only the running time is greatly reduced but also the accuracy of the optical flow is improved. Keywords: Sparse Optical Flow; Shi-Tomasi Corner Detection; LK Algorithm 一种稀疏光流计算方法 于乃功，徐丽，阮晓钢 北京工业大学 电子信息与控制工程学院，北京 100124 摘 要：光流是图像序列中像素点运动的瞬时速度，在许多计算机视觉任务中发挥着重要作用。因此，光流的计算一直 是人们研究的热点。但是，大多数光流算法需要计算每个像素点的光流，算法耗时，难以满足实时性的要求，这大大阻 碍了光流的应用。而且，在运动跟踪和机器人避障等具体任务中，计算全局光流场也是不必要的。基于此，本文提出一