A High-accuracy and Semi-dense Feature-based VSLAM System

  • Zhen Li Kyushu Institute of Technology
  • Yuren Du Yangzhou University
  • Miaomiao Zhu Changshu Institute of Technology
  • Shi Zhou Kyushu Institute of Technology
  • Seiichi Serikawa Kyushu Institute fo Technology
  • Lifeng Zhang Kyushu Institute of Technology


SLAM (Simultaneous Localization and Mapping) technology is the basis of a mobile robot, also one of the most important conditions for robot intelligence. VSLAM (Visual Simultaneous Localization and Mapping) technology uses visual sensors to build incremental environment maps and simultaneously realize the self-positioning of a moving robot. Currently, feature-based VSLAM systems tend to build sparse maps, but sparse maps are not accurate enough for the applications of navigation and obstacle avoidance. Therefore, this paper proposed a high-accurate and semi-dense VSLAM system with a monocular camera. The proposed work used polar line searching and block matching modules to help feature-based VSLAM systems measure depth information more accurately. Experimental results show that the proposed semi-dense VSLAM system can reconstruct more dense point cloud maps, and the synchronization tracking trajectory accuracy of the system is 9.13% higher than that of ORB-SLAM2 system.


J. K. Makhubela, T. Zuva and O. Y. Agunbiade, “A Review on Vision Simultaneous Localization and Mapping (VS- LAM)”, 2018 International Conference on Intelligent and Innovative Computing Applications (ICONIC), pp.1-5, 2018.

B.D. Gouveia, D. Portugal and L. Marques, “Speeding up Rao-Blackwellized particle filter SLAM with a multi threaded architecture”, Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp.1583-1588, 2014.

H. Strasdat, J. Montiel and A. J. Davison, “Scale drift-awarelarge scale monocular SLAM”, Proceedings of the Robotics: Science and Systems, pp.27-30, 2010.

B. Bescos, J. M. Fa ́cil and J. Civera, “DynaSLAM: Tracking, Mapping and Inpainting in Dynamic Scenes”, IEEE Robotics and Automation Letters. Vol.3, 2018.

J. Tang, L. Ericson and J. Folkesson, “Efficient Correspondence Prediction for Real-Time SLAM”, Computer Vision and Pattern Recognition, 2019.

R. Mur-Artal, J. D. Tards, “Orb-slam2: An open-source slam system for monocular stereo and rgb-d cameras”, IEEE Transactions on Robotics, Vol.33, no.5, pp.1255-1262, 2016.

A. Geiger, J. Ziegler and C. Stiller, “Stereoscan: dense 3d reconstruction in real-time”, Intelligent Vehic1es Symposium (IV), pp.963-968, 2014.

J. Engel, J. Sturm, D. Cremers, “Semi-dense visual odometry for a monocular camera”, International Conference on Computer Vision (lCCV), 2013.

E. Rublee, V. Rabaud and K. Konolige, “ORB: an efficient alternative to SIFT or SURF”, IEEE Intl. Conf. on Computer Vision (ICCV), Vol.13, 2011.

A. Hornung, K. M. Wurm, M. Bennewitz, “OctoMap: An efficient probabilistic 3D mapping framework based on octrees”, Autonomous Robots, Vol.34, no.3, pp.189-206, 2013.

How to Cite
Li, Z., Du, Y., Zhu, M., Zhou, S., Serikawa, S., & Zhang, L. (2021). A High-accuracy and Semi-dense Feature-based VSLAM System. Journal of the Institute of Industrial Applications Engineers, 9(4), 124. https://doi.org/10.12792/jiiae.9.124