Development of Spectral Disease Indices for Southern Corn Rust Detection and Severity Classification

Abstract

Southern Corn Rust (SCR) is one of the most destructive diseases in corn production, significantly affecting corn quality and yields globally. Field-based fast, nondestructive diagnosis of SCR is critical for smart agriculture applications to reduce pesticide use and ensure food safety. The development of spectral disease indices (SDIs), based on in situ leaf reflectance spectra, has proven to be an effective method in detecting plant diseases in the field. However, little is known about leaf spectral signatures that can assist in the accurate diagnosis of SCR, and no SDIs-based model has been reported for the field-based SCR monitoring. Here, to address those issues, we developed SDIs-based monitoring models to detect SCR-infected leaves and classify SCR damage severity. In detail, we first collected in situ leaf reflectance spectra (350–2500 nm) of healthy and infected corn plants with three severity levels (light, medium, and severe) using a portable spectrometer. Then, the RELIEF-F algorithm was performed to select the most discriminative features (wavelengths) and two band normalized differences for developing SDIs (i.e., health index and severity index) in SCR detection and severity classification, respectively. The leaf reflectance spectra, most sensitive to SCR detection and severity classification, were found in the 572 nm, 766 nm, and 1445 nm wavelength and 575 nm, 640 nm, and 1670 nm wavelength, respectively. These spectral features were associated with leaf pigment and leaf water content. Finally, by employing a support vector machine (SVM), the performances of developed SCR-SDIs were assessed and compared with 38 stress-related vegetation indices (VIs) identified in the literature. The SDIs-based models developed in this study achieved an overall accuracy of 87% and 70% in SCR detection and severity classification, 1.1% and 8.3% higher than the other best VIs-based model under study, respectively. Our results thus suggest that the SCR-SDIs is a promising tool for fast, nondestructive diagnosis of SCR in the field over large areas. To our knowledge, this study represents one of the first few efforts to provide a theoretical basis for remote sensing of SCR at field and larger scales. With the increasing use of unmanned aerial vehicles (UAVs) with hyperspectral measurement capability, more studies should be conducted to expand our developed SCR-SDIs for SCR monitoring at different study sites and growing stages in the future.

Southern corn rust caused by Puccinia polysora Underw: a review

“…Additionally, a molecular detection system for monitoring SCR has also been established (Leng et al 2012), and spectral disease indices (SDIs)-based monitoring models have been developed for evaluating infected leaves and classifying the severity of SCR damage. Such models have provided a sound theoretical basis for remote sensing of SCR in the field (Meng et al 2020). Additionally, a prediction model has been established for the prediction of SCR disease index (Li et al 2019).…”

Making Sense of Light: The Use of Optical Spectroscopy Techniques in Plant Sciences and Agriculture

“…These applications have been extended to other species and crops, namely, to discriminate between multiple cultivars of a crop species [59], to predict critical crops yield, and in the early detection of potentially dangerous diseases. These include the laurel wilt disease in avocado [60]; fire blight in pear trees [61,62]; canker, black spot, decay, and Huanglongbing (HLB) in citrus [1,10,11,32,63]; Yellow Rust in wheat [64]; Southern Corn Rust (SCR) [65]; the Potato Virus Y (PVY) disease in visually asymptomatic infected potato plants [59]; or to identify and classify grapevines inoculated with the Grapevine Vein-Clearing Virus (GVCV) at the early asymptomatic stages, under field conditions [57].…”

“…Overall, for all sensors systems described above, the ultimate breakthrough is linked with today's explosive development of advanced and powerful machine learning methods of data processing, harnessing big data to infer critical information, such as, the classic partial least squares (PLS), support vector machines, artificial neural networks, classification techniques, deep learning, and other artificial intelligence (AI) approaches [60,65,68,70,81,[86][87][88][89]. This opens a number of novel perspectives in the assessment and classification, beyond the stateof-the-art, whose current landmarks can be represented by the following examples: the automated identification and classification of Chinese medicinal plants with different sensing techniques, including Vis-NIRS [90]; the prediction of quality attributes and internal browning disorder in "Rocha" pear by Vis-NIRS reflectance and semi-transmittance spectra taken under real-life conditions met in an automated inline grading system [79,80,91]; the assessment of citrus ripening on-tree [83]; the in situ grapevine identification (down to the level of varieties) via leaf reflectance spectra [92]; the anthocyanins fingerprinting in intact grape berries [93]; the detection of mercury induced stress in tobacco plants [94].…”

“…This opens a number of novel perspectives in the assessment and classification, beyond the stateof-the-art, whose current landmarks can be represented by the following examples: the automated identification and classification of Chinese medicinal plants with different sensing techniques, including Vis-NIRS [90]; the prediction of quality attributes and internal browning disorder in "Rocha" pear by Vis-NIRS reflectance and semi-transmittance spectra taken under real-life conditions met in an automated inline grading system [79,80,91]; the assessment of citrus ripening on-tree [83]; the in situ grapevine identification (down to the level of varieties) via leaf reflectance spectra [92]; the anthocyanins fingerprinting in intact grape berries [93]; the detection of mercury induced stress in tobacco plants [94]. Additionally, it is worth mentioning the use of specific algorithms, as the hyperspectral insect damage detection algorithm (HIDDA), which allowed automatic detection of insect-damaged coffee beans using only a few bands and one hyperspectral signature [70]; or the RELIEF-F algorithm used to select the most discriminative features (wavelengths) and two band normalized differences for developing spectral disease indices for SCR detection and severity classification [65].…”

Proximal Methods for Plant Stress Detection Using Optical Sensors and Machine Learning

“…VIs are computed using ratios and combinations of reflectance measurements at a few specific wavelengths and have been used extensively for plant stress monitoring [40][41][42]. In addition to VIs, hyperspectral data can be used to develop spectral disease indices (SDIs) with the purpose of discriminating between specific plant diseases [43] (Table 2). Some examples include indices for detecting powdery mildew in wheat [44] and sugar beet [45], cercospora leaf spot in sugar beet [45], leaf rust in wheat [46], and myrtle rust [47].…”