On Model-Free Re-ranking for Visual Place Recognition with Deep Learned Local Features

TL;DR

Three new model-free re-ranking methods that were designed primarily for deep-learned local visual features evince high robustness to various appearance changes, which stands as a crucial property for use with long-term autonomy systems.

Abstract

Re-ranking is the second stage of a visual place recognition task, in which the system chooses the best-matching images from a pre-selected subset of candidates. Model-free approaches compute the image pair similarity based on a spatial comparison of corresponding local visual features, eliminating the need for computationally expensive estimation of a model describing transformation between images. The article focuses on model-free re-ranking based on standard local visual features and their applicability in long-term autonomy systems. It introduces three new model-free re-ranking methods that were designed primarily for deep-learned local visual features. These features evince high robustness to various appearance changes, which stands as a crucial property for use with long-term autonomy systems. All the introduced methods were employed in a new visual place recognition system together with the D2-net feature detector (Dusmanu, 2019) and experimentally tested with diverse, challenging public datasets. The obtained results are on par with current state-of-the-art methods, affirming that model-free approaches are a viable and worthwhile path for long-term visual place recognition.

Figures (9)

• Figure 1: Example distributions of standard local visual features detected in an image structure (left) and features extracted at fixed grid positions (right)
• Figure 2: Matched local visual features between images and the corresponding histogram of shifts (the number of matches was limited to 30)
• Figure 3: Anchor points principle. Individual cells represent local visual features extracted at fixed grid positions. Green cells belong to neighboring matches consistent with the selected anchor points pair (red).
• Figure 4: The new VPR system with the SSM-VPR or MixVPR filtering stage and re-ranking stage using the introduced model-free methods and D2-Net features.
• Figure 5: Example images from databases introduced in SSM-VPR and Sect. \ref{['Datasets']}. The upper row contains images from reference sequences and the bottom row depicts corresponding images from query sequences.