Development of multi-sensing technologies for high-throughput morphological, physiological, and biochemical phenotyping of drought-stressed watermelon plants
Mohammad Akbar Faqeerzada, Eunsoo Park, Jinsu Lim, Kihyun Kim, Ramaraj Sathasivam, Sang Un Park, Hangi Kim, Byoung–Kwan Cho
Abstract
High-throughput plant phenotyping (HTPP) technologies are rapidly transforming plant science by enabling real-time, non-invasive, and large-scale monitoring of complex morphological, physiological, and biochemical traits. However, existing platforms often lack integration across sensing modalities and analytical depth necessary for early and comprehensive phenotypic trait analysis. In this study, we developed a fully automated, multimodal HTPP system combining RGB, shortwave infrared (SWIR) hyperspectral, multispectral fluorescence imaging (MSFI), and thermal imaging to characterize drought-stressed watermelon (Citrullus lanatus) plants. RGB imaging facilitated detailed morphological analysis by extracting color-based traits, quantifying plant height and canopy area, and accurately distinguishing growth stages. SWIR hyperspectral imaging (HSI) enabled non-invasive biochemical assessment by detecting drought-responsive compounds, such as flavonoids, phenolics, and antioxidant activities, while also supporting the classification of stress severity. This spectral profiling revealed key biochemical alterations triggered by water deficit. MSFI liquid crystal tunable filter (LCTF-based) measured chlorophyll a (Chl-a), chlorophyll b (Chl-b), and total chlorophyll (t-Chl) levels, providing critical insights into photosynthetic performance under drought stress. Thermal imaging further enhanced drought assessment by capturing canopy temperature variations, which were used to derive thermal indices for indirect estimation of soil volumetric water content (SVWC). By integrating complementary imaging modalities, the proposed system captured comprehensive phenotypic responses with high predictive accuracy for early detection of drought stress and assessment of plant health. Advanced machine learning (ML) and deep learning (DL) models further enhanced trait extraction and classification, enabling robust analysis of complex, high-dimensional data. This automated, multimodal platform offers scalable, non-invasive crop monitoring, providing precise insights to support drought resilience and precision agriculture.