Deep Next-Best-View Planner for Cross-Season Visual Route Classification

Deep Next-Best-View Planner for Cross-Season Visual Route Classification
复制标题

用于跨季节视觉路线分类的深度次最佳视图规划器

DOI:
10.1109/icpr48806.2021.9412043
复制
发表时间:
2020
期刊:
Proceedings of the 25th International Conference on Pattern Recognition
影响因子:
--
通讯作者:
Tanaka Kanji
Tanaka Kanji
中科院分区:
--
文献类型:
--
作者:
Kurauchi Kanya;Tanaka Kanji

文献摘要

参考文献

相似文献

本文从长期自主的新角度解决了主动视觉位置识别(VPR)的问题。在我们的方法中,次佳视图(NBV)规划器规划最佳的动作观察序列,以最大化视觉路线分类任务的预期成本性能。由于 NBV 规划人员要在不同领域(一天中的时间、天气条件和季节)接受培训和测试,因此出现了一个困难。现有的 NBV 方法可能会因领域的转移而变得混乱和恶化,并且需要付出巨大的努力才能使其适应新的领域。我们通过一种新颖的基于深度卷积神经网络(DNN)的 NBV 规划器来解决这个问题,该规划器不需要适应步骤。我们在本文中的主要贡献总结如下:(1)我们提出了一种新颖的域不变 NBV 规划器,专门为基于 DNN 的 VPR 量身定制。 (2) 我们将主动 VPR 表述为 POMDP 问题,并提出一个可行的解决方案来解决其固有的棘手问题。具体来说,可用 DNN 输出的概率分布向量 (PDV) 用作域不变观测模型,无需重新训练。 (3) 我们通过具有挑战性的跨季节 VPR 实验验证了所提出方法的有效性,证实所提出的方法在 VPR 准确性和动作观察成本方面明显优于之前基于单视图或多视图的 VPR。
This paper addresses the problem of active visual place recognition (VPR) from a novel perspective of long-term autonomy. In our approach, a next-best-view (NBV) planner plans an optimal action-observation-sequence to maximize the expected cost-performance for a visual route classification task. A difficulty arises from the fact that the NBV planner is trained and tested in different domains (times of day, weather conditions, and seasons). Existing NBV methods may be confused and deteriorated by the domain-shifts, and require significant efforts for adapting them to a new domain. We address this issue by a novel deep convolutional neural network (DNN) -based NBV planner that does not require the adaptation step. Our main contributions in this paper are summarized as follows: (1) We present a novel domain-invariant NBV planner that is specifically tailored for DNN-based VPR. (2) We formulate the active VPR as a POMDP problem and present a feasible solution to address the inherent intractability. Specifically, the probability distribution vector (PDV) output by the available DNN is used as a domain-invariant observation model without the need to retrain it. (3) We verify efficacy of the proposed approach through challenging cross-season VPR experiments, where it is confirmed that the proposed approach clearly outperforms the previous single-view-based or multi-view-based VPR in terms of VPR accuracy and action-observation-cost.
基于气泡空间在拓扑图中放置表示
DOI: 10.1109/icra.2012.6225367
发表时间: 2012
期刊: 2012 IEEE International Conference on Robotics and Automation
影响因子: --
作者:
O. Erkent;Isil H. Bozma
通讯作者: Isil H. Bozma
基于支持向量学习的室外环境中基于视图的定位
DOI: 10.1109/iros.2005.1545445
发表时间: 2005
期刊: 2005 IEEE/RSJ International Conference on Intelligent Robots and Systems
影响因子: --
作者:
Hideo Morita;M. Hild;J. Miura;Y. Shirai
通讯作者: Y. Shirai
DOI: --
发表时间: 2018-01
期刊: ArXiv
影响因子: --
作者:
Devendra Singh Chaplot;Emilio Parisotto;R. Salakhutdinov
通讯作者: Devendra Singh Chaplot;Emilio Parisotto;R. Salakhutdinov
DOI: 10.1109/lra.2019.2932575
发表时间: 2019-10-01
影响因子: 5.2
作者:
Gottipati, Sai Krishna;Seo, Keehong;Paull, Liam
通讯作者: Paull, Liam