Multidimensional UAV-Based Assessment of Rice Bacterial Leaf Blight: Integrating Spectral, Textural, Thermal, and Spatial Features with Machine Learning Arif K Wijayanto (a*)(b)(c), Lilik B Prasetyo (b), Chiharu Hongo (d)
a) Graduate School of Science and Engineering, Chiba University, Chiba 263-8522, Japan
b) Department of Forest Resource Conservation and Ecotourism, Faculty of Forestry and Environment, IPB University, Bogor 16680, Indonesia
c) Environmental Research Center, IPB University, Bogor 16680, Indonesia
d) Center for Environmental Remote Sensing (CEReS), Chiba University, Chiba 263-8522, Japan
Abstract
Bacterial Leaf Blight (BLB) disease poses a significant threat to rice yields, potentially leading to losses of up to 50%. To offset these losses, the government of Indonesia through the Ministry of Agriculture has implemented agricultural insurance schemes that depend on precise damage assessments to ensure equitable compensation. However, the program relies heavily on traditional manual assessments which often lack of consistency and accuracy. This research introduces an integrated model that combines textural, thermal, and patch fragmentation metrics, all derived from UAV-based multispectral and thermal imagery, to improve the detection of BLB damage. The study was carried out in the Cihea irrigation area in Cianjur, Indonesia, utilizing multisensory UAV data from specified irrigation blocks. The model^s performance was assessed using machine learning approaches and compared with evaluations from pest observers. Findings reveal that integrating multiple features significantly enhances the accuracy of disease classification, achieving an overall accuracy of 0.998.
