An artificial intelligence (AI) model developed by researchers in China has shown promise for detecting HLB in leaves before visible symptoms appear. The technology could eventually give citrus growers another tool for earlier disease detection and more targeted grove monitoring.
The study combined three types of information: standard color photographs, hyperspectral imaging and measurements of leaf chlorophyll content.
The researchers examined 2,700 citrus leaves representing healthy trees and five levels of HLB disease status, including infections without visible symptoms. They used laboratory molecular testing to help confirm whether leaves were infected.
Hyperspectral imaging measures how leaves reflect light at different wavelengths, potentially revealing changes that ordinary cameras cannot detect. Researchers also used a soil-plant analysis development meter to estimate leaf chlorophyll content, another indicator of plant health.
When the AI model combined all three information sources, it achieved 97.62% average accuracy in classifying five levels of HLB severity. By comparison, a model using only standard photographs achieved 87.64% accuracy, while one relying only on hyperspectral imaging reached 91.43%.
In a separate test distinguishing healthy leaves, infected leaves without visible symptoms and mildly infected leaves, the combined model achieved 91.25% accuracy.
Early HLB detection remains an important challenge because infected trees may not initially show obvious symptoms. A technology capable of identifying infection sooner could help growers improve surveillance and make more informed disease-management decisions.
The study was conducted under controlled conditions and evaluated individual leaves rather than entire trees. The researchers identified field testing, evaluation across more citrus varieties and methods for assessing whole tree canopies as important next steps.
For growers, the findings offer encouraging evidence that combining different types of leaf measurements with AI may improve early HLB detection. Further testing will determine whether the approach can deliver reliable results under real-world grove conditions.
An article about the study was published in Smart Agricultural Technology. It was written by Jian Tao, Shiqing Dou, Jiancheng Du, Guichuan Wu and Jichi Yan with Guilin University of Technology and Zhengmin Mei and Yaqin Song with Guangxi Academy of Specialty Crops.
Share this Post