Showing 1–2 of 2 results for author: Kuse, M

Loop-box: Multi-Agent Direct SLAM Triggered by Single Loop Closure for Large-Scale Mapping

Authors: M Usman Maqbool Bhutta, Manohar Kuse, Rui Fan, Yanan Liu, Ming Liu

Abstract: In this paper, we present a multi-agent framework for real-time large-scale 3D reconstruction applications. In SLAM, researchers usually build and update a 3D map after applying non-linear pose graph optimization techniques. Moreover, many multi-agent systems are prevalently using odometry information from additional sensors. These methods generally involve intensive computer vision algorithms and… ▽ More In this paper, we present a multi-agent framework for real-time large-scale 3D reconstruction applications. In SLAM, researchers usually build and update a 3D map after applying non-linear pose graph optimization techniques. Moreover, many multi-agent systems are prevalently using odometry information from additional sensors. These methods generally involve intensive computer vision algorithms and are tightly coupled with various sensors. We develop a generic method for the keychallenging scenarios in multi-agent 3D mapping based on different camera systems. The proposed framework performs actively in terms of localizing each agent after the first loop closure between them. It is shown that the proposed system only uses monocular cameras to yield real-time multi-agent large-scale localization and 3D global mapping. Based on the initial matching, our system can calculate the optimal scale difference between multiple 3D maps and then estimate an accurate relative pose transformation for large-scale global mapping. △ Less

Submitted 29 September, 2020; originally announced September 2020.

Comments: Material related to this work is available at https://usmanmaqbool.github.io/loop-box

Journal ref: IEEE Transactions on Cybernetics, 2020

Learning Whole-Image Descriptors for Real-time Loop Detection andKidnap Recovery under Large Viewpoint Difference

Authors: Manohar Kuse, Shaojie Shen

Abstract: We present a real-time stereo visual-inertial-SLAM system which is able to recover from complicatedkidnap scenarios and failures online in realtime. We propose to learn the whole-image-descriptorin a weakly supervised manner based on NetVLAD and decoupled convolutions. We analyse thetraining difficulties in using standard loss formulations and propose an allpairloss and show itseffect through exte… ▽ More We present a real-time stereo visual-inertial-SLAM system which is able to recover from complicatedkidnap scenarios and failures online in realtime. We propose to learn the whole-image-descriptorin a weakly supervised manner based on NetVLAD and decoupled convolutions. We analyse thetraining difficulties in using standard loss formulations and propose an allpairloss and show itseffect through extensive experiments. Compared to standard NetVLAD, our network takes an orderof magnitude fewer computations and model parameters, as a result runs about three times faster.We evaluate the representation power of our descriptor on standard datasets with precision-recall.Unlike previous loop detection methods which have been evaluated only on fronto-parallel revisits,we evaluate the performace of our method with competing methods on scenarios involving largeviewpoint difference. Finally, we present the fully functional system with relative computation andhandling of multiple world co-ordinate system which is able to reduce odometry drift, recover fromcomplicated kidnap scenarios and random odometry failures. We open source our fully functional system as an add-on for the popular VINS-Fusion. △ Less

Submitted 15 April, 2019; originally announced April 2019.