SPADE-AI: Deep Learning Based Soil Property Estimation from Smartphone Image through Spectral Reconstruction and Multi-Source Data Fusion for Crop Suitability Assessment and Rotation Planning

SPADE-AI is an AI-driven research project focused on estimating soil properties from ordinary smartphone images using deep learning-based spectral reconstruction and multi-source data fusion. The project will collect geo-referenced soil samples along with RGB and hyperspectral images from different agro-ecological zones of Bangladesh, develop models to reconstruct hyperspectral information from RGB images, and estimate key soil properties such as nitrogen, phosphorus, potassium, pH, organic matter, and micronutrients. These results will be integrated with soil, climate, geospatial, and crop databases to provide location-specific crop suitability and rotation recommendations. The project will ultimately deliver a Bengali-language mobile application that enables farmers to capture soil images and receive accessible, real-time agricultural guidance, helping reduce dependence on costly laboratory testing while promoting sustainable agriculture, improved productivity, and better soil management in Bangladesh.