摘要
光流场的计算是移动机器人领域一个很重要的研究课题,但现有的光流场计算方法并未充分利用移动机器人的特点。本文假设移动机器人配备了双目视觉系统和码盘,在此基础上提出了一种基于立体匹配技术的光流场计算方法。该方法的基本思想是:首先利用Forstner算子对图像中易于跟踪的特征点进行提取;然后利用立体视觉方法计算这些特征点的三维坐标;在此基础上,进一步结合码盘信息对前后帧图像中的特征点进行预测和跟踪,从而最终实现了光流的准确提取。实验表明,该方法比传统方法具有速度更快、准确性更高的优点。
The computation of optical flow fields is one of the important branches in the study of mobile robots. However, the existing methods fail to use all the information of mobile robots. In this paper, a new method to compute optical flow fields based on stereo matching is presented. This method can be used if mobile robots are equipped with stereo vision and odometry. First, the Forstner feature points for tracking will be extracted. Second, the 3-D coordinates of these points will be obtained with the stereo vision method. Third, the information from odometry can be used to predict and track the feature points. Finally the optical flow can be obtained. Experiments show that this method can get more accurate results in less time than traditional methods.
出处
《计算机工程与科学》
CSCD
2006年第10期50-53,共4页
Computer Engineering & Science
基金
国家自然科学基金重点资助项目(60234030)
