Answers are grounded in this paper and cite their sources.
The integration of hyperspectral data with machine learning models represents an efficient alternative for the detection of different diseases and severity levels in soybean crops. The objectives of this study were: to construct hyperspectral signatures enabling the discrimination of target spot and Asian soybean rust, as well as their respective severity levels; to evaluate the performance of different machine learning models for disease classification using hyperspectral data; and to identify the most effective input combinations between algorithms and data types to improve diagnostic accuracy. Two field experiments were conducted during the 2022/2023 crop season for the collection of leaves infected with target spot and Asian soybean rust. Leaf sampling was performed at the R5.5 phenological stage, corresponding to the grain-filling phase. Healthy leaflets and leaflets exhibiting 25% and 50% severity were collected, with 100 leaves per severity level for each disease. The collected leaves were analyzed using a spectroradiometer. The hyperspectral data allowed the generation of spectral signatures for each disease and their respective severity levels. Spectral bands (SB) and band depth indices (RID) were also derived from the hyperspectral data, with both reductions aimed at decreasing the complexity of subsequent data processing. Principal component analysis was performed to assess the relationships among spectral bands, band depth indices, disease classes, severity levels, and healthy leaves. In addition to the full spectral dataset and its dimensionality-reduced counterparts, the data were subjected to machine learning analysis using the following algorithms: Multilayer Perceptron, REPTree and J48 decision trees, Random Forest (RF), Support Vector Machine (SMO), Random Tree, and Logistic Regression (LR) as a traditional classification benchmark. Target spot exhibited the greatest spectral distinction across all analyzed regions, whereas Asian soybean rust showed more pronounced differences in the VIS and NIR regions, with lower differentiation in the SWIR. The relevance of specific bands — particularly B11–B22 and B27 for discriminating Asian soybean rust, and the initial VIS bands (B1–B3) for healthy leaves — was highlighted. Overall, SMO combined with SB proved to be the most effective combination for classifying different disease types and severity levels in soybean leaves, while MP and LR performed best when combined with RID or SB.
No takes yet. Share an insight, caveat, or question.
Santana et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: