Curing characteristics, primarily manifested as easy curing characteristic (EaCC) and endurable curing characteristic (EnCC), are key indicators for quality grading and intelligent curing of fresh tobacco leaves. However, existing detection methods rely on complex physicochemical analyses, which are destructive and unsuitable for real-time applications. Therefore, this study proposed a physicochemical indicator-driven multimodal spectral fusion method for rapid and non-destructive detection of fresh tobacco leaf curing characteristics. Key physicochemical indicators associated with EaCC and EnCC were identified through significance analysis. These indicators were used as intermediary variables to explicitly guide feature band selection from chlorophyll fluorescence (ChlF), visible-near infrared (VNIR), and short-wave infrared (SWIR) spectra. Multimodal fusion models were developed using selected bands and Support Vector Machine (SVM) optimized by Sparrow Search Algorithm (SSA). The results indicated that: (1) Chlorophyll-a, chlorophyll-b, carotenoids, leaf thickness, and SPAD value were the dominant factors influencing EaCC, with spectral responses in ChlF and VNIR modalities. Moisture content, leaf thickness, chlorophyll-a, and nicotine influenced EnCC, with responses in VNIR and SWIR modalities. (2) The proposed method screened 129 and 96 characteristic bands for EaCC and EnCC, eliminating 95.04% and 96.31% of redundant bands, respectively. (3) The SSA-SVM model developed using physicochemical indicator-driven multimodal spectral bands, achieved accuracies of 83.00% and 82.00% for EaCC and EnCC, respectively, outperforming data-driven methods such as CARS, SPA, and IRIV. This study provides a rapid, non-destructive method for assessing curing characteristics, highlighting its potential for real-time quality monitoring and intelligent evaluation of fresh and wet agricultural materials.
Huang et al. (2026) studied this question.