UAV vehicle detection improves in low light using RGB and infrared differences
ReDiffNet: Differential RGB-Infrared Learning for Low-Light UAV Oriented Vehicle Detection
Machine Learning
Summary
Detecting small vehicles at night using drones is hard because lights, shadows, and heat signals can be unreliable and confusing. The authors developed ReDiffNet, a method that figures out which parts of an image from RGB and infrared cameras are trustworthy. It then combines the most reliable parts and recovers missing details from the other camera type to get a clearer picture. This approach helps detect vehicles more accurately in dark conditions.
What this means in practice
- •For traffic monitoring teams: Detect small vehicles at night using drones combining RGB and infrared data for more reliable urban traffic tracking.
- •For emergency response units: Identify vehicles during nighttime emergencies by improving UAV-based vehicle detection in challenging lighting conditions.
Authors
Qifan Zhang, Ziran Zhou, Ruijie Li, Jincheng Tang, Hao Wang, Qihao Qiao, Chunliu Wang
Abstract
Low-light UAV-based RGB-infrared oriented small-vehicle detection is important for nighttime traffic monitoring, emergency response, and urban inspection. Illumination variations, headlight glare, local shadows, and thermal-response degradation cause spatially varying modality reliability, while the small visual extent of vehicles further weakens boundaries, orientation cues, and thermal responses. Accordingly, selecting trustworthy observations based on local modality reliability while further exploiting complementary discriminative information in regions with ambiguous modality preference is key to constructing effective multimodal representations. Based on this insight, we propose ReDiffNet, a reliability-conditioned differential representation network in which modality reliability guides both evidence selection and complementary recovery. Specifically, degradation-aware reliability learning estimates relative spatial reliability, uncertainty-guided differential recovery exploits cross-modal differences to recover complementary cues in ambiguous regions, and reliability-conditioned reconstruction integrates retained and recovered evidence into a unified representation. ReDiffNet achieves 85.3% and 73.9% mAP50 on DroneVehicle and VEDAI, respectively, supporting its effectiveness.