计算机与现代化

• 图像处理 • 上一篇    下一篇

一种结合区域检测和语义分割的SLAM技术

  

  1. (国网浙江省电力有限公司信息通信分公司,浙江杭州310012)
  • 收稿日期:2019-01-03 出版日期:2019-07-05 发布日期:2019-07-08
  • 作者简介:王文(1985-),男,浙江衢州人,工程师,本科,研究方向:信息化项目管理,E-mail: wangwen@zj.sgcc.com.cn; 徐亦白(1992-),男,山东枣庄人,助理工程师,硕士,研究方向:信息化项目管理,E-mail: xuyibai@zj.sgcc.com.cn; 卢杉(1992-),男,浙江丽水人,助理工程师,硕士,研究方向:信息化项目管理,E-mail: lushan@zj.sgcc.com.cn; 冯宇(1982-),男,浙江杭州人,工程师,本科,研究方向:办公自动化业务及数据交换等平台类信息系统运维,E-mail: fengyu@zj.sgcc.com.cn。
  • 基金资助:
    国网浙江省电力有限公司科技项目(5211XT17000C)

A SLAM Technology Combining Area Detection and Semantic Segmentation

  1. (Information and Communication Branch, State Grid Zhejiang Electric Power Co., Ltd., Hangzhou 310012, China)
  • Received:2019-01-03 Online:2019-07-05 Published:2019-07-08

摘要: 提出一种结合区域检测和语义分割的即时定位和建图(SLAM)技术,通过引入高精度图像描述子SIFT来实现前端视觉里程计(VO)过程中帧间像素匹配的精度。为了降低引入操作带来的计算复杂度,设计一个实时区域检测算法,在相邻帧间检测大致相似的ROI(Region of Interest)关键区域,使得SIFT描述子的提取和匹配只在ROI区域内完成,其余区域仍旧采用精度略低、效率更高的ORB算子。同时,为了提高后端BA(Bundle Adjustment)的精度,减少累积误差,结合语义图,在原有的基本投影误差函数上添加一个语义误差。该语义图采用实时语义分割算法完成,同时只针对ROI区域进行分割。通过与原SLAM方案对比实验,表明本文提出的方法,在提高一定精度的同时,仍能满足SLAM实时定位和建图的要求。最后,在电力作业场景下验证了该方案的效果。

关键词: SLAM, 区域检测, 语义分割, 视觉里程计, 计算机视觉

Abstract: This paper proposes a real-time location and mapping (SLAM) technology that combines region detection and semantic segmentation. The precision of inter-frame pixel matching in the visual odometer (VO) process is realized by introducing high-precision image descriptor SIFT. In order to reduce the computational complexity caused by the introduction operation, we design a real-time region detection algorithm to detect the region of interest (ROI) between adjacent frames, so that the SIFT descriptors are extracted and matched only in the ROI region. At the same time, in order to improve the accuracy of the bundle adjustment (BA) and reduce the cumulative error, the paper combines the semantic information. The semantic map is implemented by a real-time semantic segmentation algorithm. Compared with the original SLAM scheme, the proposed method can improve the accuracy of SLAM and meet the requirements of real-time localization and mapping. Finally, we verify the effect of the scheme in the power operation scenario.

Key words: SLAM, area detection, semantic segmentation, visual odometer, computer vision

中图分类号: