Rice is a staple food for one-third of the global population and a critical pillar for over 65% of China’s dietary needs. However, rice blast, a fungal disease, ravages rice crops in 85 countries worldwide, causing annual yield losses of 10%–30%, with severe cases leading to complete crop failure. Traditional detection methods, relying on manual field surveys and laboratory biochemical analyses, are inefficient, labor-intensive, and fail to meet modern agriculture’s demand for efficient, non-destructive monitoring. How can we rapidly and non-invasively detect subtle signs of rice blast disease and achieve precise field management?
Shuai Feng and Chunling Chen’s team from Shenyang Agricultural University, have developed a novel vegetation index—the Rice Blast Index (RBI)—using drone-based hyperspectral remote sensing technology, offering a solution to this challenge. The study has been published in Frontiers of Agricultural Science and Engineering (DOI: 10.15302/J-FASE-2024576).
Unlike conventional vegetation indices, RBI does not simply follow traditional formulas. Instead, it identifies sensitive spectral bands by analyzing the unique impact of the disease on rice leaf spectral characteristics. The researchers collected spectral data spanning 400–1000 nm from rice fields in Haicheng, Liaoning Province, using drones equipped with hyperspectral sensors. Through ANOVA and the Relief-F algorithm, they pinpointed three critical bands: 778, 722, and 664 nm. These bands correspond to chlorophyll decline, cellular structural damage, and canopy morphological changes caused by the disease, enabling RBI to accurately distinguish healthy from infected plants.
To validate RBI’s efficacy, the researchers compared it with widely used vegetation indices. Results showed that RBI’s absolute correlation coefficient with disease severity reached 0.98, far surpassing other indices. In classification models, RBI performed exceptionally: it achieved 95.0% overall accuracy with the K-Nearest Neighbors algorithm and 95.1% with the Random Forest model. Notably, RBI exhibited minimal overlap between disease severity levels, with only slight overlap between healthy and mildly infected plants. This high sensitivity and stability allow RBI not only to diagnose infection but also to quantify disease severity.
Technical breakthroughs rely on real-world applicability. Traditional hyperspectral detection, confined to lab settings, is complex and impractical for field use. This study, however, employed drones hovering at 100 meters to collect data, combined with radiometric and regional correction techniques, significantly reducing environmental light interference and ensuring data reliability. Additionally, the researchers interpolated spectral resolution to 1 nm and divided 250 regions of interest across five disease severity levels (from healthy to severely infected), providing robust data for model training. The “high-altitude collection + ground verification” approach preserves crop integrity while enabling rapid large-scale field scanning.
This breakthrough offers an efficient tool for early detection and graded management of rice blast, reducing excessive pesticide use and yield losses. The study highlights that RBI marks a critical shift in agricultural remote sensing from “health monitoring” to “disease-specific diagnosis”. In the future, this method may extend to crops like wheat and corn, advancing the development of smart agriculture.
DOI: 10.15302/J-FASE-2024576