Abstract:
In this paper, we present a simultaneous tracking and dense structured mapping method aimed at robotic tasks including obstacle avoidance, path planning and manipulation....Show MoreMetadata
Abstract:
In this paper, we present a simultaneous tracking and dense structured mapping method aimed at robotic tasks including obstacle avoidance, path planning and manipulation. This is achieved through (1) sparse-to-dense pose estimation with both keypoints and dense vertex images for keeping robust at scenes lack of geometric features, (2) dense local map algorithm, which clears map outside the neighborhood of the robot for the consistency of the system at large-scale scenes, (3) structured analysis to obtain elevation and norms of surfaces, (4) the proposed dynamic mode for ghosting-free mapping in dynamic scenes. Experiments are conducted to evaluate each part of the system. The effectiveness and efficiency are proved by quantitative experiments.
Published in: 2021 27th International Conference on Mechatronics and Machine Vision in Practice (M2VIP)
Date of Conference: 26-28 November 2021
Date Added to IEEE Xplore: 07 January 2022
ISBN Information: