Diagnosis of CTV-Infected Leaves Using Hyperspectral Imaging

Abstract

Hyperspectral reflectance images of healthy and diseased leaves infected with different isolates of Citrus tristeza virus (CTV) including TRL514, CT30, CT32 and CT11A were collected in the visible and near-infrared region of 400–1000 nm. Average reflectance spectrum was generated from each hyperspectral image individually obtained from 60 healthy and 240 CTV-infected leaves. The spectra were transformed with 15-point Savitzky Golay second derivative. Then principal component analysis was performed on the transformed data in order to reduce the dimension of data. Comparative analysis was performed among supervised classification models, including back-propagation neural network (BPNN), linear discriminant analysis (LDA) and Mahalanobis distance (MD). When the second derivative spectra were analyzed, classifier models including BPNN, LDA and MD can discriminate the healthy and CTV-infected leaves with the highest classification accuracies of 100% in the spectral range of 400–1000 nm and 760–1000 nm. Nine optimal wavelengths (405, 424, 920, 947, 957, 972, 978, 980, and 998 nm) selected by stepwise regression resulted in 97.33% total classification accuracy for differentiation of healthy and CTV-infected leaves and showed great potential in CTV diagnosis. However, the overall classification accuracy of different CTV isolates infected leaves resulted in 70% based on the MD model using the selected optimal wavelengths. Further study is required to find out whether the method is suitable for CTV detection under field conditions.

Keywords::

• Hyperspectral imaging
• Citrus tristeza virus (CTV)
• Back-propagation neural network
• Linear discriminant analysis
• Mahalanobis distance

Additional information

Notes on contributors

Dongmei Guo

D. Guo received B.Sc. degree in Horticulture from Southwest University, Chongqing, China in 2012. She is a graduate in Southwest University, Chongqing, China. Her research interests include fruit tree physiology and information technology.

Rangjin Xie

R. Xie received B.Sc. degree in Pomology from Southwest Agricultural University, Chongqing, China in 2002, M.Sc. degree in Pomology from Southwest Agricultural University, Chongqing, China in 2008, and Ph.D. degree in Pomology from Zhejiang University, Hangzhou, China in 2011. His research interests include citrus cultivation physiology, and molecular biology.

Chun Qian

C. Qian received B.Sc. degree in Horticulture from Southwest University, Chongqing, China in 1996, the M.Sc. degree in Pomology from Southwest University, Chongqing, China in 2006, and Ph.D. degree in Pomology from Southwest University, Chongqing, China in 2010. His research interests include resource utilization of horticultural plants, fruit breeding, and biotechnology.

Fangyun Yang

F. Y. Yang received B.Sc. degree in Pomology from Southwest Agricultural University, Chongqing, China in 1996, M. Sc. degree in Pedology from Southwest Agricultural University, Chongqing, China in 2004, and Ph.D. degree in Pomology from Southwest University, Chongqing, China in 2012. His research interests include citrus virus disease.

Yan Zhou

Y. Zhou received B.Sc. degree in Plant Protection from Southwest Agricultural University, Chongqing, China in 2001, M. Sc. degree in Plant Disease from Southwest Agricultural University, Chongqing, China in 2004, and Ph.D. degree in Plant Disease from Southwest University, Chongqing, China in 2007. His research interests include citrus virus disease.

Lie Deng

L. Deng received B.Sc. degree in Pomology from Southwest Agricultural College, China in 1982. His research interests include intelligent systems, citrus physiology and citriculture technology, systems modeling and analysis. Prof. Deng is the Elite researcher of the Chinese Academy of Agricultural Sciences, standing committee member of the Chinese Society of Citriculture, Vice-president of the Chongqing Society of Horticulture, and Editor-in-chief of the China Fruit News.