Detection and discrimination of disease and insect stress of tea plants using hyperspectral imaging combined with wavelet analysis.

Authors: ZHAO, XIAOHU - Hangzhou Dianzi University; Huang, Yanbo; ZHANG, JINGCHENG - Hangzhou Dianzi University; YUAN, LIN - Hangzhou Dianzi University; XU, JUNFENG - Hangzhou Dianzi University

Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/13/2022
Publication Date: 1/20/2022
Citation: Zhao, X., Huang, Y., Zhang, J., Yuan, L., Xu, J. 2022. Detection and discrimination of disease and insect stress of tea plants using hyperspectral imaging combined with wavelet analysis.. Computers and Electronics in Agriculture. https://doi.org/10.1016/j.compag.2022.106717.
DOI: https://doi.org/10.1016/j.compag.2022.106717

Interpretive Summary: Plant monitoring for disease and insect stresses are important for crop management. Scientists from Hangzhou Danzi University and USDA-ARS Crop Production Systems Research Unit at Stoneville, MS have collaboratively developed the hyperspectral imaging method to rapidly detect and discriminate disease and insect stresses of tea plants. The results indicated that the developed method can effectively detect the stresses and discriminate each stress from plant disease or insect. This study provides useful information for the hyperspectral imaging method to be applied in tea plant monitoring for effective pest management.

Technical Abstract: Compared with the traditional visual method, hyperspectral imaging can monitor plants efficiently and nondestructively, and has great potential in plant phenotyping in response to disease and insect infections. At present, most researches based on hyperspectral imaging are focused on the detection of single disease, and the methods developed can rarely discriminate multiple co-occurring diseases and insects. In this study, three stresses of tea plants including tea green leafhopper (Empoasca vitis Göthe), anthracnose (Gloeosporium theae-sinesis Miyake) and sunburn were considered as a major threat to tea plantation in yield and quality, and a multi-step approach to discrimination of tea plant stresses was proposed based on the hyperspectral imaging and continuous wavelet analysis. The methods include: (1) Extract features for detection and discrimination of tea plant stresses based on continuous wavelet analysis; (2) Detect tea leaf abnormal area based on the k-means clustering and support vector machine algorithms; (3) Construct a tea plant stress discrimination model using the random forest algorithm, to identify and discriminate the three tea stresses. The results showed that the continuous wavelet analysis can effectively identify spectral features for distinguishing the three stresses. The overall discrimination accuracy of the proposed approach reached 90.69%, among which the identification accuracy of tea anthracnose was the highest (94.12%), followed by tea green leafhopper (94.03%) and sunburn damage (83.91%). Therefore, hyperspectral imaging can provide an effective way for plant phenotyping from diseases and insect infections on tea plants.