SDPL: Shifting-Dense Partition Learning for UAV-View Geo-Localization

TL;DR

A dense partition strategy (DPS) is proposed, dividing the image into multiple parts to explore contextual information while explicitly maintaining the global structure, and the shifting-fusion strategy, which generates multiple sets of parts in parallel based on various segmentation centers.

Abstract

Cross-view geo-localization aims to match images of the same target from different platforms, e.g., drone and satellite. It is a challenging task due to the changing appearance of targets and environmental content from different views. Most methods focus on obtaining more comprehensive information through feature map segmentation, while inevitably destroying the image structure, and are sensitive to the shifting and scale of the target in the query. To address the above issues, we introduce simple yet effective part-based representation learning, shifting-dense partition learning (SDPL). We propose a dense partition strategy (DPS), dividing the image into multiple parts to explore contextual information while explicitly maintaining the global structure. To handle scenarios with non-centered targets, we further propose the shifting-fusion strategy, which generates multiple sets of parts in parallel based on various segmentation centers, and then adaptively fuses all features to integrate their anti-offset ability. Extensive experiments show that SDPL is robust to position shifting, and performs com-petitively on two prevailing benchmarks, University-1652 and SUES-200. In addition, SDPL shows satisfactory compatibility with a variety of backbone networks (e.g., ResNet and Swin). https://github.com/C-water/SDPL release.

Figures (11)

• Figure 1: Dense partition strategy. (a) Yellow dashed boxes: base parts generated by LPN wang2021each. DPS recombines base parts to mine local context information while preserving global structure. Part number $P$ is marked in green triangle; (b) Recombination process for parts with various resolutions. The low-resolution features are padded to ensure resolution consistency of blocks.
• Figure 2: Feature shifting-fusion strategy. (a) Examples of partitions generated by DPS with various segmentation centers: centered, top-left shifting, bottom-right shifting; (b) Adaptive fusion strategy fuses three sets of parts using parameter $\beta_{D}$ learned from weight estimation module; (c) Hyper-parameter $\beta_{D}^{hand}$ is adopted to fuse three sets of parts for ablation experiments.
• Figure 3: Overview of SDPL framework. A--C (bottom): Diagrams of dense partition strategy with various segmentation centers. During testing, part-level image representation is extracted before classification layer in classifier module, and measures similarity by Euclidean distance.
• Figure 4: Image retrieval results obtained with SDPL, FSRA and LPN. (a) Top-5 retrieval results of drone localization on University-1652; (b) Top-5 retrieval results of drone navigation on University-1652.
• Figure 5: Visualization of heatmaps generated by Baseline, LPN and SDPL.