Selection of Leaf Orientation Insensitive Bands for Yellow Rust Detection - Computer and Computing Technologies in Agriculture VI - Part I
Conference Papers Year : 2013

Selection of Leaf Orientation Insensitive Bands for Yellow Rust Detection

Abstract

The disease detection by means of hyperspectral reflectance is influenced by the spectral differences between frontside (adaxial surface) and backside (abaxial surface) of a leaf inevitably. Taking yellow rust as an example, this study investigated the spectral differences between frontside and backside of healthy and diseased wheat leaves at grain filling stage using large size samples. We attempted to detect yellow rust with reflectance that was sensitive to the disease and insensitive to the orientation of leaves. The spectral difference between frontside and backside of leaves was analyzed by band ratioing and a pairwise t-test. The bands that were insensitive to the orientation of leaves were identified with a thresholding method. Then, with the aid of an independent t-test analysis, we recognized the bands that were sensitive to the disease. The overlapped bands were applied for developing models that quantifying disease severity by fisher linear discrimination analysis (FLDA). The results suggested that the bands within 606-697nm and 740-1000nm were suitable for disease detection yet insensitive to the orientation of leaves. Based on these bands, the model accuracies reached 71% for FLDA. These bands can be used as a basis for further selection of appropriate bands to detect yellow rust at canopy level.
Fichier principal
Vignette du fichier
978-3-642-36124-1_10_Chapter.pdf (514.28 Ko) Télécharger le fichier
Origin Files produced by the author(s)
Loading...

Dates and versions

hal-01348084 , version 1 (22-07-2016)

Licence

Identifiers

Cite

Lin Yuan, Jingcheng Zhang, Jinling Zhao, Shuhong Cai, Jihua Wang. Selection of Leaf Orientation Insensitive Bands for Yellow Rust Detection. 6th Computer and Computing Technologies in Agriculture (CCTA), Oct 2012, Zhangjiajie, China. pp.78-84, ⟨10.1007/978-3-642-36124-1_10⟩. ⟨hal-01348084⟩
50 View
74 Download

Altmetric

Share

More