东北大学学报:自然科学版 ›› 2020, Vol. 41 ›› Issue (6): 841-846.DOI: 10.12068/j.issn.1005-3026.2020.06.014

• 机械工程 • 上一篇    下一篇

一种融合单目信息的RGB-D SLAM优化方法

马树军, 王英蕾, 金铁铮, 白昕晖   

  1. (东北大学 机械工程与自动化学院, 辽宁 沈阳110819)
  • 收稿日期:2019-05-21 修回日期:2019-05-21 出版日期:2020-06-15 发布日期:2020-06-12
  • 通讯作者: 马树军
  • 作者简介:马树军(1982-),男,河北保定人,东北大学副教授.
  • 基金资助:
    国家自然科学基金资助项目(51505076); 辽宁省自然科学基金资助项目 (2015020105); 中央高校基本科研业务费专项资金资助项目(N180304016,N140304010); 辽宁省高等学校创新团队项目(LT2014006).

A RGB-D SLAM Optimal Method for Fusing Monocular Information

MA Shu-jun, WANG Ying-lei, JIN Tie-zheng, BAI Xin-hui   

  1. School of Mechanical Engineering & Automation, Northeastern University, Shenyang 110819, China.
  • Received:2019-05-21 Revised:2019-05-21 Online:2020-06-15 Published:2020-06-12
  • Contact: MA Shu-jun
  • About author:-
  • Supported by:
    -

摘要: 由于RGB-D传感器本身存在的局限性,其在实际应用中会有深度信息不存在或者不完整的情况,这使得多数依赖于深度信息工作的RGB-D SLAM系统的精度会受到影响.提出了一种融合单目信息的RGB-D SLAM优化方法.通过对深度图像进行分析,从深度信息的存在性、准确性和离散性分别提出三个判断条件,作为选择单目或者RGB-D处理方式的依据.该系统使得用RGB-D相机工作的SLAM系统更具精确性与鲁棒性.最后通过运行数据集将系统与其他SLAM系统进行比较,实验结果验证了本系统的可行性.

关键词: RGB-D SLAM, 单目信息, 融合, 判断机制, 优化

Abstract: Owing to the limitations of RGB-D sensor, there will be some situations where the depth information does not exist or is incomplete in the practical application, which will affect the accuracy of RGB-D SLAM system that mostly depends on the depth information.An RGB-D SLAM optimization method is proposed to fuse monocular information.Through the analysis of depth image, three judging conditions are put forward respectively from the existence, accuracy and discreteness of depth information, which can be used as the basis of the selection of monocular or RGB-D processing mode.This new system makes the SLAM system with RGB-D camera more accurate and robust.Finally, the system is compared with other SLAM systems by running data sets, and the experimental results verify its feasibility.

Key words: RGB-D SLAM, monocular information, fusion, judgment mechanism, optimization

中图分类号: