Abstract—

In this paper, we propose a method of obstacle detection for guide dog robots based on stereo vision. Theoretically, we analyze the lens imaging principle of a binocular camera and 3D ranging method and calibrate the checkerboard grid images of different angles based on MATLAB. Through the study and analysis of the stereo-matching algorithm, we obtain accurate depth information. Using the contour extraction algorithm to solve the contour of the obstacle and the area of the convex package; then set the area threshold to filter out the false detection part, to quickly and effectively calculate the position coordinates of the obstacle, and finally obtain the 3D stereo information of the environment ahead. The algorithm proposed in this paper is verified on the experimental platform built. The experimental results show that the algorithm is effective and feasible.

中文翻译：

---

基于立体视觉的GDR障碍物检测方法研究

摘要-

在本文中，我们提出了一种基于立体视觉的导盲犬机器人障碍物检测方法。理论上分析了双目相机的镜头成像原理和3D测距方法，并基于MATLAB对不同角度的棋盘格图像进行了标定。通过对立体匹配算法的研究和分析，获得了准确的深度信息。利用轮廓提取算法求解障碍物轮廓和凸包面积；然后设置面积阈值，过滤掉误检测部分，快速有效地计算出障碍物的位置坐标，最终获得前方环境的3D立体信息。本文提出的算法在搭建的实验平台上得到了验证。实验结果表明该算法有效可行。