A generic workflow combining deep learning and chemometrics for processing close-range spectral images to detect drought stress in Arabidopsis thaliana to support digital phenotyping

Close-range spectral imaging (SI) of agricultural plants is widely performed for digital plant phenotyping. A key task in phenotyping the non-destructive and rapid identification drought stress so as to allow breeders select potential genotypes breeding drought-resistant varieties. Visible near-infrared SI a sensing technique that allows capture physicochemical changes occurring under stress. The main challenges are processing massive images extract information relevant support genotype selection. Hence, this study presents generic data workflow analysing generated real-world experiments meaningful decision making by breeders. combination chemometric approaches deep learning. usefulness proposed demonstrated on real-life experiment related detection quantification Arabidopsis thaliana grown semi-controlled environment. results show approach able detect presence just 3 days after its induction compared well-watered plants. Furthermore, unsupervised clustering provides detailed time-series where drought-related can be followed visually along time course. developed facilitates thus accelerate drought-tolerant