Place recognition is an essential component to address the problem of visual navigation and SLAM. The long-term place recognition is challenging as the environment exhibits significant variations across different times of the days, months, and seasons. In this paper, we view appearance changes as multiple domains and propose a Feature Disentanglement Network (FDNet) based on a convolutional auto-encoder and adversarial learning to extract two independent deep features– content and appearance. In our network, the content feature is learned which only retains the content information of images through the competition with the discriminators and content encoder. Besides, we utilize the triplets loss to make the appearance feature encode the appearance information. The generated content features are directly used to measure the similarity of images without dimensionality reduction operations. We use datasets that contain extreme appearance changes to carry out experiments, which show how meaningful recall at 100% precision can be achieved by our proposed method where existing state-of-art approaches often get worse performance.
Bibliographical noteFunding Information:
Supported by National Natural Science Foundation of China (No. 61973066, 61471110), Fundamental Research Funds for the Central Universities (N172608005, N182608004), Equipment Pre-research Foundation (61403120111), Foundation of Key Laboratory of Aerospace System Simulation (6142002301) and the Natural Science Foundation of Liaoning (No.20180520040).
Sonya Coleman received the B.Sc. degree (Hons.) in mathematics, statistics, and computing, and the Ph.D. degree in mathematics from Ulster University, Londonderry, U.K., in 1999 and 2003, respectively. She is currently a Professor with the School of Computing and Intelligent System, Ulster University, and also a Cognitive Robotics Team Leader with the Intelligent Systems Research Centre. Her research has been supported by funding from various sources such as EPSRC, The Nuffield Foundation, The Leverhulme Trust, and the EU. She was involved in the EU FP7 funded projects RUBICON, VISUALISE, and SLANDAIL. She has authored or coauthored over 150 publications in robotics, image processing, and computational neuroscience. Dr. Coleman was awarded the Distinguished Research Fellowship by Ulster University in recognition of her contribution research in 2009.
Yunzhou Zhang received B.S. and M.S. degree in Mechanical and Electronic engineering from National University of Defense Technology, Changsha, China in 1997 and 2000, respectively. He received Ph.D. degree in pattern recognition and intelligent system from Northeastern University, Shenyang, China, in 2009. He is currently a professor with the Faculty of Robot Science and Engineering, Northeastern University, China. Now he leads the Cloud Robotics and Visual Perception Research Group. His research has been supported by funding from various sources such as National Natural Science Foundation of China, Ministry of science and technology of China, Ministry of Education of China and some famous high-tech companies. He has published many journal papers and conference papers in intelligent robots, computer vision and wireless sensor networks. His research interests include intelligent robot, computer vision, and sensor networks.
© 2020 Elsevier B.V.
Copyright 2020 Elsevier B.V., All rights reserved.
- Visual place recognition
- Changing environment
- Adversarial learning
- Representation disentanglement