Multi-Band Feature Fusion in Satellite Images for Land Cover Classification

Satellite imagery plays a crucial role in land cover classification for various applications, including urban planning, environmental monitoring, and land resource management. This paper presents a novel approach for land cover classification using satellite images by leveraging multi-band feature fusion. The proposed method aims to enhance the representation of land cover classes by performing fusion operations on the nine spectral bands beyond RGB and NIR, effectively condensing them into three fused bands. To accomplish this, we propose the EfficientN et_mb- BO architecture as the backbone, by integrating the Band-Fusion Module (BFM) into the backbone 9 spectral bands are fused into 3 feature channels and concatenated with RGB and NIR which enables the CNN network to accurately capture the class variations using multi-spectral image features. Our, results show the effectiveness of our idea in the EuroSat dataset.