计算机工程与应用 ›› 2017, Vol. 53 ›› Issue (2): 236-240.DOI: 10.3778/j.issn.1002-8331.1504-0075

• 工程与应用 • 上一篇    下一篇

基于改进RANSAC的消防机器人双目障碍检测

王军华1,李  丁1,刘盛鹏2   

  1. 1.上海大学 通信与信息工程学院,上海 200072
    2.公安部上海消防研究所,上海 200438
  • 出版日期:2017-01-15 发布日期:2017-05-11

Improved RANSAC estimation based fire-fighting robot obstacle detection using binocular vision

WANG Junhua1, LI Ding1, LIU Shengpeng2   

  1. 1.School of Communication and Information Engineering, Shanghai University, Shanghai 200072, China
    2.Shanghai Fire Research Institute of Ministry Public Security, Shanghai 200438, China
  • Online:2017-01-15 Published:2017-05-11

摘要: 针对消防机器人自主作业的障碍物快速检测问题,给出了一种基于改进随机采样一致性估计的双目障碍物检测算法。该算法首先采集双目视觉左右视图,进行半全局立体匹配获取视差信息,然后采用随机采样一致性估计的平面拟合法提取地平面模型,并采用预检验法和内点阈值限定法同时对随机采样一致性估计进行改进,从而提高算法效率,实现障碍物快速检测。实验结果证明该方法能够准确、快速检测障碍物,满足消防机器人作业需求。

关键词: 立体匹配, 随机采样一致性, 预检验, 平面拟合, 障碍物检测

Abstract: Considering the fast obstacle detection of autonomous stereo vision fire-fighting robot, an improved algorithm based on RANSAC(Random Sample Consensus) is proposed in this paper. The method collects left and right view from stereo vision system firstly to obtain the disparity by using SGBM(Semi-Global Stereo Matching), then extracts ground plane directly though fitting. In order to enhance algorithm efficiency, RANSAC estimation is improved by pre-verification and threshold-constrained. Experimental result shows that the proposed method satisfies the task of autonomous fire-fighting robot adequately, and is able to detect obstacles not only accurately but in real time.

Key words: stereo matching, Random Sample Consensus(RANSAC), pre-verification, fitting ground plane, obstacle detection