Advanced BLB disease assessment in paddy fields using multispectral UAV data and patch fragmentation metrics

Abstract

This study introduces an innovative method for assessing bacterial leaf blight (BLB) in paddy fields using multispectral UAV (Unmanned Aerial Vehicle) data and patch fragmentation analysis. Unlike traditional pixel-based approaches, which often lack spatial context, our method treats pixels as objects and evaluates their spatial relationships to determine BLB severity. Seven patch fragmentation metrics were derived from binarized vegetation indices to quantify BLB damage scores, carefully selected for their ability to describe the spatial arrangement and connectivity of potentially affected patches. This metric-driven approach captures the scale and intensity of BLB damage, facilitating precise assessment. The method demonstrated high accuracy, achieving an AUC of 0.938 with a 0.5-meter sampling window. This advancement enhances the precision of BLB damage assessment, particularly for applications such as crop insurance.

Keywords

• Crop disease
• Drone
• Fragmented patches
• Landscapemetrics
• Landscape-scale analysis
• Pest management