Leveraging Deep Learning for the Reconstruction of Plant Hyperspectral Data from RGB Images

Hyperspectral imaging is an important tool used in plant health assessment. It allows for early detection of plant stress which allows for timely intervention and thus improved conservation efforts. However, the high cost and complexity of hyperspectral cameras has limited their usage. To mitigate this issue, the problem of reconstructing plant hyperspectral data from RGB images is investigated. The proposed model reconstructs the visual and near-infrared range (400 - 1000 nm) while being trained solely on images of vegetation, in contrast with existing “generic” models. It is hypothesized that training a less complex model on a specific material will achieve good accuracy even with a relatively small training dataset. The HSCNN-D model is adopted with a simplified architecture. Despite being much smaller than the original model, it achieves comparable performance to state-of-the-art models on images of vegetation.