Continual De-Raining, Defect Detection, Multimodal Fusion, and Confidence-Aware Infrastructure Analytics

Abstract

Resilient infrastructure analytics requires robust visual perception, defect detection, multimodal fusion, and reliable model interpretation. Continual rain removal improves image quality under changing rainy conditions, while domain adaptation and dual prompt learning address visual degradation in outdoor scenes. NeRF-based defect detection contributes a 3D representation approach for identifying structural anomalies. Infrared-visible image fusion and LiDAR-guided hyperspectral cross-attention provide complementary multimodal sensing strategies for target detection and image classification. Battery diagnostics and power electronics research address the energy reliability of sensing platforms. Pavement friction prediction connects visual indicators to transportation safety. LLM confidence research is relevant because infrastructure analytics systems must communicate uncertainty clearly when supporting maintenance or emergency decisions. This literature cluster supports safety-critical monitoring frameworks that integrate weather-robust vision, structural defect detection, multimodal sensing, energy diagnostics, and confidence-aware AI.