Debin Tang | Searcharxiv

arXiv2026

GAAT: Geometry-Aware Alignment Transformer for Multimodal UAV Perception

Unmanned aerial vehicle (UAV) multimodal perception integrates visible (RGB), infrared (IR), synthetic aperture radar (SAR), and depth sensors for scene understanding under diverse conditions. However, differences in optics, resolution, and mounting often limit practical systems to global or image-center alignment. After tokenization, parallax, platform motion, and lens distortion can shift corresponding patch centers across modalities, weakening the spatial correspondence assumed by dense contrastive learning and cross-modal fusion. We propose GAAT (Geometry-Aware Alignment Transformer), an alignment-first pretrained model that estimates local correspondence reliability before cross-modal interaction. GAAT introduces syncPATC, which learns patch-center consistency under synchronized view transformations without correspondence annotations. It emits geometric priors, including token and query confidence, query centers, and sub-token offsets, that identify reliable local anchors across residual misalignment. Guided by these priors, MG-Sparse-MMA performs query-mediated sparse fusion over top-K_s reliable regions, replacing dense all-patch interaction with geometry-calibrated local updates. RA-QCGCL aligns pretraining supervision with this sparse query bottleneck through reliable patch-to-patch, patch-to-query, and query-to-query contrastive branches. We introduce UAVMeta and StateBench, which provide four acquisition-state scores derived from platform telemetry and image statistics: camera reliability, observation scale, viewpoint stability, and flight maneuver complexity. Extensive experiments across six downstream tasks demonstrate consistently superior transfer performance, establishing GAAT as a state-of-the-art multimodal foundation model for UAV perception. StateBench further enables a systematic diagnosis of real-world acquisition conditions.

cs.CV↗

arXiv2025

Edge-ANN: Storage-Efficient Edge-Based Remote Sensing Feature Retrieval

Meeting real-time constraints for high-performance Approximate Nearest Neighbor (ANN) search remains a critical challenge in remote sensing edge devices, which are essentially fusion systems like micro-satellites and UAVs, largely due to stringent limitations in primary (RAM) and secondary (disk) storage. To address this challenge, we propose Edge-ANN, an innovative ANN framework specifically engineered for storage efficiency. The core innovation of Edge-ANN lies in its departure from traditional tree-based methods that store high-dimensional hyperplanes. Instead, it leverages pairs of existing data items, termed "anchors," to implicitly define spatial partitions. To ensure these partitions are both balanced and effective, we have developed a novel Binary Anchor Optimization algorithm.This architectural shift eliminates the dimension-dependence of the space complexity. Rigorous experiments on three multi-source datasets, MillionAID, High-resolution Urban Complex Dataset, and GlobalUrbanNet Dataset, demonstrate that under simulated edge environments with dual storage constraints, Edge-ANN achieves a 30-40% reduction in secondary storage compared to the baseline, at the cost of a minor 3-5% drop in retrieval accuracy. Furthermore, its overall retrieval performance surpasses that of other mainstream methods in these constrained scenarios. Collectively, these results establish Edge-ANN as a state-of-the-art solution for enabling large-scale, high-performance, real-time remote sensing feature retrieval on edge devices with exceptionally constrained storage. The codes of Edge-ANN are available at https://github.com/huaijiao666/Edge-ANN.

cs.CG↗