Water Body Detection Using Sentinel-2 Imagery Through Particle Swarm Intelligence: A Novel Framework for Optimizing Spectral Multi-Band Index

Water body detection from satellite imagery is still challenging due to spectral confusion and the limitation of traditional water indices. This paper proposes a new approach by incorporating Particle Swarm Optimization with a Spectral Multi-Band Water Index for the enhanced detection of water bodies using Sentinel-2 imagery. The proposed approach optimizes the coefficients of seven Sentinel-2 bands (Blue, Green, NIR, NIR-Narrow, Water Vapor, SWIR1, and SWIR2) using an intelligent PSO with adaptive inertia weight and early stopping mechanisms. This work strategy proposes a new fitness function that applies dynamic thresholding and target-based optimization, allowing it to calibrate precisely to the local characteristics of the water body. The performance of the PSO-SMBWI was evaluated against traditional water indices, including the NDWI, MNDWI, and AWEI. The results indicate that the PSO-SMBWI has the highest accuracy, which exactly coincides with the ground truth of water coverage (12.12%), while the NDWI, MNDWI, and AWEI have deviations of +1.24%, +0.53%, and +12.15%, respectively. The proposed method automatically handles multi-resolution band integration in 10 m, 20 m, and 60 m and eliminates manual threshold tuning. Furthermore, our consensus-based validation approach ensures robust performance verification. Its effectiveness is due to its adaptive optimization framework and comprehensive spectral analysis. Hence, it is most suitable for any geographical context on the ground for highly accurate water body mapping. This research contributes a lot to the area of remote sensing by introducing an automated, highly accurate, and very computationally efficient approach to water body detection.