Visual place recognition: A survey from deep learning perspective
作者：
Highlights：
• We provide a whole picture about deep learning-based visual place recognition.
• The differences and similarities between VPR and image retrieval are included.
• We review different kinds of CNN-based methods, novel CNN features and datasets for VPR.
• New tools such as GANs and multi-modality feature fusion are discussed for VPR.
• We discuss challenges, open issues and future directions of visual place recognition.
摘要
•We provide a whole picture about deep learning-based visual place recognition.•The differences and similarities between VPR and image retrieval are included.•We review different kinds of CNN-based methods, novel CNN features and datasets for VPR.•New tools such as GANs and multi-modality feature fusion are discussed for VPR.•We discuss challenges, open issues and future directions of visual place recognition.
论文关键词：Visual place recognition,Deep learning,Visual SLAM,Survey
论文评审过程：Received 10 November 2019, Revised 4 September 2020, Accepted 25 November 2020, Available online 28 November 2020, Version of Record 19 February 2021.
论文官网地址：https://doi.org/10.1016/j.patcog.2020.107760
