基于角点检测的双目视觉测距新方法

A New Method for Binocular Vision Ranging Based on Corner Detection

为解决双目视觉中立体匹配困难、效率低的问题,提出了一种基于角点检测的人工匹配方法实现双目视觉测距。首先介绍了双目测距基本原理,对摄像机进行了标定,并采用Bouguet立体校正算法对双目视觉系统进行立体校正;然后采用J.Shi和C.Tomasi提出的角点方法对立体校正后的左右摄像机图像中被测目标上的同一特征点进行了亚像素级角点提取,并利用提取的匹配角点坐标结合双目视觉测距公式实现距离测量。

It is difficult to make stereo matching in distance measurement based on binocular vision, and the efficiency is low. To solve the problems, we proposed a new method based on comer detecting and manual matching. The basic theory of binocular vision based distance measurement was presented. The cameras were calibrated, and stereo rectification of binocular vision system was realized d by using arithmetic of Bouguet. Then the method proposed by J. Shi and C. Tomasi was used to extract the sub-pixel comer points, which belonged to the same feature-point of the target in the left and right images. Finally, the coordinates of the matched corner points and distance-measuring formula were used to implement the distance measurement.