Detection of wheat powdery mildew based on hyperspectral reflectance through SPA and PLS-LDA
Imran Haider Khan,Haiyan Liu,Tao Cheng,Yongchao Tian,Qiang Cao,Yan Zhu,Weixing Cao,Xia Yao
DOI: https://doi.org/10.33440/j.ijpaa.20200301.67
2018-01-01
International Journal of Precision Agricultural Aviation
Abstract:Abstract: The accurate recognition and quantitative assessment at the early stage of wheat powdery mildew ( Blumeria graminis f. sp. tritici ) are vital for precision crop management for spraying the fungicides, reducing the cost, protecting the environment and enhancing the quality of crop. However, early disease detection remained highly difficult due to the subtle changes in the physiology and phenology of the plants at early infection stage. In this study, two wheat cultivars with different disease resistances were inoculated by the powdery mildew, hyperspectral reflectance and physiological parameters of leaves were obtained after inoculation at early stem elongation stage. The major contribution of this study is to extract sensitive wavebands and vegetation indices using sub-window permutation analysis (SPA) by fully exploiting the hyperspectral data for early disease identification. Extracted sensitive features by SPA were then used as input in partial least squares-linear discriminant analysis (PLS-LDA) recognition model to classify the healthy and diseased wheat leaves. Finally, validation was carried out with independent data to verify the accuracy of the recognition model. The results indicated that (1) the pigment contents and photosynthetic capacity were changed slightly at the early infection stage but decreased rapidly with the aggravation of the disease severity; (2) the visible and the near infrared bands were the most sensitive to the disease at early infection stage; (3) the overall accuracy of the PLS-LDA model constructed with the sensitive features extracted by SPA method performed better than features selected conventionally by correlation analysis. The calibration and validation accuracies at 5% disease severity were 85.12 and 84.43% for model based on wavelength features and were 82.14 and 85.63 for model constructed with spectral indices features extracted by SPA, respectively.  In conclusion, SPA is a new effective strategy for feature selection which has not been yet used in plant disease research, having the benefit of considering cooperative effect among different variables and demonstrated the potential of early disease detection. Such a technique can be an efficient and economical substitute to conventional methods, especially in case of high throughput hyperspectral crop sensing. Keywords: wheat powdery mildew, early detection, hyperspectral reflectance, fungicide spraying, sub-window permutation analysis (SPA), PLS-LDA DOI:  10.33440/j.ijpaa.20200301.67  Citation: Khan I H, Liu H Y, Cheng T, Tian Y C, Cao Q, Zhu Y, Cao W X, Yao X. Detection of wheat powdery mildew based on hyperspectral reflectance through SPA and PLS-LDA.  Int J Precis Agric Aviat, 2020; 3(1): 13–22.