Turns Ordinary RGB Photos Into Hyperspectral Cubes to Diagnose Sick Oil

A study in Smart Agricultural Technology proposes turning ordinary RGB leaf photos into hyperspectral “data cubes” to diagnose oil-palm problems earlier and more cheaply. Researchers used a smartphone/drone-style RGB camera to reconstruct 31 spectral bands, aiming to detect issues such as basal stem rot, fungal leaf spots, and nutrient deficiencies in nitrogen, phosphorus and potassium—conditions that can damage yields before humans notice symptoms. The technical hurdle is that three RGB channels can’t uniquely determine a high-dimensional spectrum, so the team staged reconstruction: it first generates a coarse spectral prior, then extracts multi-scale texture features with a Multi-scale Atrous Convolutional Residual Network, reconstructing visible (400–700 nm) before estimating near-infrared (700–950 nm). The dataset used 7,312 images across five health classes, with potassium deficiency heavily dominating.