Uncertainty-Aware Multimodal Anti-UAV Detection via Evidential Fusion and Conflict-Discounted Belief Aggregation
Computer Vision and Pattern Recognition
Summary
The authors studied ways to improve systems that detect drones using both regular (RGB) and thermal cameras. They focused on combining data from these two types of sensors while also measuring how uncertain the system is about its predictions. Their new method tries to handle conflicts between the two sensors by adding uncertainty before making a decision. Testing on a drone detection benchmark showed their combined approach works better than using just one type of sensor, but the special conflict-handling step didn’t make a difference because the test data rarely had conflicting information. They explain this limitation as being due to the test setup rather than their method.
Authors
Sharanda Suttorp, Seyed Sahand Mohammadi Ziabari, Ali Mohammed Mansour Alsahag
Abstract
Anti-UAV perception systems must remain reliable when sensor streams degrade under occlusion, fast motion, or modality-specific failure. Existing multimodal anti-UAV systems fuse RGB and thermal streams deterministically, without modeling predictive uncertainty, and cannot express doubt when streams disagree. Evidential Deep Learning (EDL) produces calibrated per-class uncertainty in a single forward pass. EDTC already exploits this for thermal-only perception, yet cross-modal evidential fusion remains unaddressed. This paper extends EDTC to multimodal RGB-Thermal perception via Discounted Belief Fusion (DBF), which converts inter-modal conflict into uncertainty mass before aggregating stream opinions. Bounding boxes are resolved by selecting the lower-uncertainty modality. On the Anti-UAV benchmark, multimodal fusion consistently outperforms either single stream (test Acc 0.670 vs. 0.604 IR, 0.598 RGB) at real-time speed (at least 38 FPS). However, DBF is empirically indistinguishable from undiscounted averaging: near-zero inter-modal conflict on this presence-dominated benchmark leaves the discounting step inert. The fused uncertainty is well-calibrated (ECE 0.057) yet expectedly a weaker localization failure detector than spatial variance (AUROC 0.626 vs. 0.739). The null result is structural: the benchmark's near-universal presence and vacuous miss-encoding jointly suppress inter-modal conflict, a diagnosis that delimits where conflict-aware fusion provides measurable benefit.