Kerala 2024 Flood Detection Using WCSO-DCNN with Real-Time Drone Swarms and Neuromorphic Edge Inference

DOI: https://doie.org/10.10399/JBSE.2025688655

Ms M B Mulik, Dr P N Kulkarani, Dr V Jayashree

Keywords:

Satellite images, Flood detection, Optimization

Abstract:

Flood disasters continue to pose significant threats to life and infrastructure, particularly in vulnerable regions such as Wayanad, Kerala, which experienced severe flooding in 2024. In this study, we propose a novel flood detection framework that integrates Whale-Crow Search Optimization (WCSO) with Deep Convolutional Neural Networks (DCNN), enhanced through real-time multi-modal data captured by autonomous drone swarms and processed on neuromorphic edge computing units. The model was trained and validated using high-resolution RGB and infrared aerial imagery collected during the 2024 Wayanad flood event, combined with hydrological sensor data. Experimental results demonstrate that our WCSO-DCNN model outperforms conventional approaches, achieving a classification accuracy of 97.8%, compared to 92.3% for standard CNN, 93.5% for CNN-LSTM, and 94.1% for Adam-optimized CNNs. The inclusion of drone-based multi-perspective imaging and neuromorphic edge inference significantly reduced latency and improved model responsiveness, enabling near real-time flood mapping. The proposed model also generated risk heatmaps with high spatial precision, highlighting its potential as a decision-support tool for disaster response agencies. To our knowledge, this is the first implementation of WCSO-optimized DCNN using drone swarm-acquired data and neuromorphic processing for flood detection, establishing a new benchmark in real-time flood monitoring.

PDF