Abstract In large-scale and long-term visual SLAM, robust place recognition is essential for building a global consistent map. However, sensor viewpoints and environmental condition changes, including lighting, weather, and seasons, bring a huge challenge to place recognition. We propose a place recognition algorithm based on CNN features and graph model. Firstly, CNN features of images are extracted though an AlexNet network with migration characteristics, and N-nearest neighbor image descriptors of the current image descriptor are found by approximate nearest neighbor searching. Then, according to the difference between descriptors, a weighted directed acyclic graph (weighted DAG) model which describes a cost of context matching between images is established. Finally, a candidate matching sequence with minimum cost on this model is achieved by using Dijkstra algorithm. Compared with SeqCNNSLAM and Fast-SeqSLAM, the experimental results demonstrate higher recognition accuracy and robustness of our algorithm.
[1]
Tomohiro Shibata,et al.
High performance loop closure detection using bag of word pairs
,
2016,
Robotics Auton. Syst..
[2]
Antonios Gasteratos,et al.
Fast loop-closure detection using visual-word-vectors from image sequences
,
2018,
Int. J. Robotics Res..
[3]
Paul Newman,et al.
Appearance-only SLAM at large scale with FAB-MAP 2.0
,
2011,
Int. J. Robotics Res..
[4]
Xuejun Yang,et al.
CNN Feature boosted SeqSLAM for Real-Time Loop Closure Detection
,
2017,
ArXiv.
[5]
Paul Newman,et al.
FAB-MAP: Probabilistic Localization and Mapping in the Space of Appearance
,
2008,
Int. J. Robotics Res..