Compact Environment-Invariant Codes for Robust Visual Place Recognition

Abstract

Robust visual place recognition (VPR) requires scene representations that are invariant to various environmental challenges such as seasonal changes and variations due to ambient lighting conditions during day and night. Moreover, a practical VPR system necessitate compact representations of environmental features. To satisfy these requirements, in this paper we suggest a modification to the existing pipeline of VPR systems to incorporate supervised hashing. The modified system learns (in a supervised setting) compact binary codes from image feature descriptors. These binary codes imbibe robustness to the visual variations exposed to it during the training phase, thereby, making the system adaptive to severe environmental changes. Also, incorporating supervised hashing makes VPR computationally more efficient and easy to implement on simple hardware. This is because binary embeddings can be learnt over simple-to-compute features and the distance computation is also in the low dimensional hamming space of binary codes. We have performed experiments on several challenging data sets covering seasonal, illumination and viewpoint variations. We also compare two widely used supervised hashing methods of CCAITQ [1] and MLH [2] and show that this new pipeline out-performs or closely matches the state-of-the-art deep learning VPR methods that are based on high-dimensional features extracted from pre-trained deep convolutional neural networks.

SEER: Unsupervised and sample-efficient environment specialization of image descriptors

“…Similar to the spatial pooler [15], the hashing function CCA-ITQ [13] learns to map an input descriptor to a binary output descriptor. Jain et al [18] showed that CCA-ITQ not only learns a hashing of descriptors but concurrently improves their performance. However, it requires additional labels about which of the given descriptors show same places.…”

A Trajectory-Based Approach to Multi-Session Underwater Visual SLAM Using Global Image Signatures

“…When used as global image descriptors, image hashes are functions providing fixed length outputs that are similar only in front of similar images [34]. The multi-session loop closing detection task compares image hashes of images of different sessions.…”

“…Since X pe cs expresses a transformation between the end of the previous session and the beginning of the current one, it is extremely simple to use it to join both sessions into a single one. Equations (33) and (34) show the mean and the covariance of the joined trajectory X J = N(X J , P J ), where P p and P c denote the covariances of X p and X c , respectively.…”

NetHALOC: A learned global image descriptor for loop closing in underwater visual SLAM

“…Applications of hashes, understood as exposed initially, include image retrieval in large databases, authentication and watermarking, among many others. However, in applications of scene recognition, localization or visual loop closing detection, it is accepted that similar or overlapping images are expected to produce similar or close hashes while distinct images produce clearly distinctive hashes (Bonin‐Font, Negre, Burguera, & Oliver, 2014; Jain, Namboodiri, & Pandey, 2017; Shahbazi & Zhang, 2011). According to this definition and in this latter context, an image hash is close or equivalent to the concept of global image descriptor, also used in several pieces of work to identify loop closings (Arandjelovic & Zisserman, 2013; Jégou, Douze, Schmid, & Pérez, 2010; Liu & Zhang, 2012; Negre Carrasco, Bonin‐Font, & Oliver‐Codina, 2016), each one improving its predecessors.…”