Research on SLAM Visual Odometry Combined with Depth Estimation Network
ZHANG Gongmingqing
Abstract：Simultaneous localization and mapping(SLAM)is the core technology in the field of robotics and AR/VR,and it plays an important role in linking the relationship between its own position and the environment.Aiming at the accuracy of the tradi-tional monocular SLAM algorithm in the environmental scale and the time-consuming system initialization,a flexible combination scheme of depth estimation network and ORB-SLAM2 system is proposed.According to ORB-SLAM2's mapping and positioning work based on key frames,this paper proposes an effective preprocessing scheme to reduce resource consumption.Based on the ROS architecture,the SLAM visual odometry module is combined with the depth estimation network,so that the monocular SLAM can obtain the environmental depth information in the initialization stage,shorten the initialization time of the SLAM system,im-prove the accuracy of the SLAM system,and it improves the instability and failure of system initialization.After testing on the TUM dataset,the results show that the method can effectively improve the accuracy and robustness of monocular ORB-SLAM2.
Keywords：simultaneous localization and mappingORB-SLAM2depth estimation networkROSvisual odometry
Publication Date：2025-06-20
Online Publishing Date：2025-09-23(First online date of this platform, not the publication date of the document)
Pages：7( 1601-1607 )