SIA OpenIR  > 光电信息技术研究室
基于组合误差的即时定位与地图构建系统
Alternative TitleSimultaneous Localization and Mapping System Based on Combined Error
孙云雷1,2
Department光电信息技术研究室
Thesis Advisor吴清潇
Keyword即时定位与地图构建系统 特征法 直接法 重投影误差 光度误差
Pages63页
Degree Discipline控制工程
Degree Name硕士
2019-05-17
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract随着移动机器人、无人车、无人机、增强现实(AR)以及虚拟现实(VR)等行业的发展,即时定位与地图构建(simultaneous localization and mapping 简称 SLAM)技术在多个领域的需求越来越大。SLAM技术中要解决的两个问题是,在未知环境下,机器人能够根据自身搭载的传感器,在运动过程中增量式的获取周围环境的特征和地图信息,并估计机器人的运动状态。虽然近几年SLAM技术发展迅速,但提高位姿估计精度和地图重建精度依然是首要任务。针对基于特征法的ORB-SLAM2系统中相机位姿求解精度不高,对纹理特征的强依赖性,只能生成稀疏地图的问题,本文提出了一种在ORB-SLAM2系统框架上将稀疏特征法和稠密的直接法结合在一起求解相机位姿,并生成稠密模型的方法。该方法主要对ORB-SLAM2系统做了三处改进。第一处改进:在原系统使用的第三方图优化库g2o中创建一条新的稠密约束一元边,将稠密直接法的光度误差约束加入到图优化库g2o中;第二处改进:跟踪相机时先采用稠密直接法计算相邻两帧图像之间相机的旋转变换,再利用改进后的图优化库g2o同时最小化特征法的重投影误差和直接法的光度误差优化求解6 DOF相机位姿;第三处改进:在ORB-SLAM2系统框架上添加稠密重建线程,将周围场景的重建结果实时地反馈给用户。本文提出的方法在TUM RGB-D和ICL NUIM数据集上的测试结果表明,本文提出的方法在一定程度上提高了ORB-SLAM2系统中相机位姿的求解精度,不仅生成稀疏地图,还重建更高精度的稠密地图。
Other AbstractWith the development of mobile robots, unmanned vehicles, drones, augmented reality(AR) and virtual reality(VR), the demand for simultaneous localization and mapping(SLAM) technology in multiple fields is increasing. Two problems to be solved by the Simultaneous Location and Map System (SLAM for short) are that, in an unknown environment, the robot can incrementally acquire the characteristics of the surrounding environment and estimate the pose of the robot during the motion according to the data acquired by the sensors carried by the robot. Although SLAM technology has developed rapidly in recent years, improving the accuracy of pose estimation and map reconstruction accuracy is still the top priority. Aiming at solving problems that the camera pose estimation accuracy is not high and only generating sparse map in the ORB-SLAM2 system, a method of combining the dense direct method with the sparse feature method adopted by the original ORB-SLAM2 system framework to compute camera pose and generate dense map is proposed. This method mainly made three improvements to the ORB-SLAM2 system. The first improvement: a new dense constraint unary edge is created in the third-party general graph optimization library g2o (General Graph Optimization) used in the original system, and the photometric error constraint of the dense direct method is added to the general graph optimization library g2o. The second improvement: the rotation transformation between two executive frames is calculated with the dense direct method, and then the improved general graph optimization library g2o is adopted to minimize the re-projection error of the feature-based method and the photometric error of the direct method at the same time to compute the 6 DOF (Degree of Freedom) camera pose. The third improvement: adding a dense reconstruction thread on the ORB-SLAM2 system framework, and reporting the reconstruction result of the surrounding scene to the user in real time. The experiment results on TUM RGB-D and ICL NUIM datasets show that the proposed method improves the accuracy of the camera pose in the ORB-SLAM2 system to a certain degree, not only produces sparse maps, but also reconstructs higher precision dense map.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/25200
Collection光电信息技术研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院大学
Recommended Citation
GB/T 7714
孙云雷. 基于组合误差的即时定位与地图构建系统[D]. 沈阳. 中国科学院沈阳自动化研究所,2019.
Files in This Item:
File Name/Size DocType Version Access License
基于组合误差的即时定位与地图构建系统.p(2295KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[孙云雷]'s Articles
Baidu academic
Similar articles in Baidu academic
[孙云雷]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[孙云雷]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.