Characterization of Disease-Specific Spectra in Tomato Early Blight Using Hyperspectral Image Analysis

Abstract

The study is part of a larger project aimed at developing an integrated disease risk warning and early detection system for greenhouse-grown tomatoes, particularly for medium- and small-scale growers. Tomato plants (Solanum lycopersicum L. ‘Encore’ F1) were cultivated under controlled greenhouse conditions and artificially inoculated with Alternaria sp., the causal agent of early blight. Disease progression was monitored using a hyperspectral (HS) camera operating at 400–1000 nm, alongside visual disease assessments. Five experimental replicates were conducted, differing slightly in plant developmental stage at inoculation and post-inoculation temperature and relative humidity. A total of 1000 HS images were acquired, comprising 155 control and 845 inoculated leaves, although visible symptoms did not develop on all inoculated leaves. Mean visually assessed disease severity across replicates ranged from 0.1% to 17.7%. Principal component analysis (PCA) effectively discriminated between diseased and non-diseased tissues. The most informative spectral regions were associated with pigment absorption (beyond 550 nm), red-edge, and near-infrared domains, supporting their relevance for early detection of Alternaria-induced stress. As proof of concept, logistic regression-based leaf image segmentation achieved a mean cross-validation accuracy equal to 93.8%.