Citrus Greening Disease (Huanglongbing) Detection Using Spectral Signatures and Machine Learning Models
Authors:
- Jiraporn Sripinyowanich JongyingcharoenJJJiraporn Sripinyowanich Jongyingcharoen
- Vishal Dashrathrao KalbandeVKVishal Dashrathrao Kalbande
- Thitima PhanomsophonTPThitima Phanomsophon
- Jeerayut HongwiangjanJHJeerayut Hongwiangjan
- Panmanas SirisomboonPSPanmanas Sirisomboon
- Bim Prasad ShresthaBSBim Prasad Shrestha
- Indra ManiIMIndra Mani
- Gopal U. ShindeGSGopal U. Shinde
- B. V. AsewarBAB. V. Asewar
- G. P. JagtapGJG. P. Jagtap
- Poom KonghuayrobPKPoom Konghuayrob
- Jayendra MishraJMJayendra Mishra
Preprint
- 27 July 2026
- preprint
- Published by Elsevier in SSRN Electronic Journal
Abstract
Artificial Intelligence (AI), Agricultural Machine Learning (AML) and Deep Learning in Agriculture (ADL) are increasingly applied for crop health monitoring and precision disease management. Huanglongbing (HLB) or citrus greening disease, is among the most destructive citrus diseases worldwide, causing substantial yield and quality losses. This study developed a non-destructive UAV-based multispectral sensing framework for real-time HLB detection and severity classification using spectral signatures and machine learning algorithms. Multispectral imagery was acquired using a Parrot Sequoia sensor mounted on a hexacopter UAV during the peak vegetative stage at Vasantrao Naik Marathwada Krishi Vidyapeeth, Parbhani, India. Spectral information from Green (550 nm), Red (660 nm), Red Edge (735 nm), and Near Infrared (790 nm) bands was used to characterize Healthy, 25% HLB, 50% HLB, 100% HLB, and Background classes. A total of 10,000 multispectral canopy patches (2,000 samples per class; 128 × 128 pixels) were extracted for analysis. Vegetation indices including NDVI, NDRE, and GNDVI were computed to evaluate canopy stress conditions. Random Forest (RF), Support Vector Machine (SVM), and K-Nearest Neighbour (KNN) classifiers were trained and evaluated using multispectral reflectance features. RF achieved the highest overall accuracy (92.45%), followed by KNN (92.05%) and SVM (91.10%), with macro-average precision, recall, and F1-score values of 92%. Multiclass ROC analysis produced micro- and macro-average AUC values of 0.99. Spectral analysis showed higher Red Edge and NIR reflectance in healthy canopies, whereas HLB-infected canopies exhibited increased Red reflectance and reduced NIR response due to chlorophyll degradation and cellular damage. ANOVA confirmed significant differences among classifiers (p < 0.05). UAV-based multispectral sensing enabled rapid and non-destructive detection of Huanglongbing severity in citrus orchards.Keywords
This publication has 29 references indexed in Scilit:
- Airborne Remote Sensing in Precision Agriculture2025
- Integrating advanced deep learning techniques for enhanced detection and classification of citrus leaf and fruit diseasesScientific Reports, 2025
- Robust fault detection and classification in power transmission lines via ensemble machine learning modelsScientific Reports, 2025
- Revolutionizing UAV: Experimental Evaluation of IoT-Enabled Unmanned Aerial Vehicle-Based Agricultural Field Monitoring Using Remote Sensing StrategyRemote Sensing in Earth Systems Sciences, 2024
- Multiclass classifier based on deep learning for detection of citrus disease using fluorescence imaging spectroscopyLaser Physics, 2023
- Variations on a theme: factors regulating interaction between Diaphorina citri and “Candidatus Liberibacter asiaticus” vector and pathogen of citrus huanglongbingCurrent Opinion in Insect Science, 2023
- A Huanglongbing Detection Method for Orange Trees Based on Deep Neural Networks and Transfer LearningIEEE Access, 2022
- Application of multiple spectral systems for the tree disease detection: A reviewApplied Spectroscopy Reviews, 2021
- Comparison of the performance of the main real-time and conventional PCR detection tests for ‘Candidatus Liberibacter’ spp., plant pathogenic bacteria causing the Huanglongbing disease in Citrus spp.European Journal of Plant Pathology, 2020
- Scaling method for fast Monte Carlo simulation of diffuse reflectance spectra from multilayered turbid mediaJournal of the Optical Society of America A, 2007