Detection of sprout damage in Canada Western Red Spring wheat with multiple wavebands using visible/near-infrared hyperspectral imaging

• Published in: Biosystems engineering Vol. 106; no. 2; pp. 188 - 194
• Main Authors: Xing, Juan, Symons, Stephen, Shahin, Muhammad, Hatcher, David
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01.06.2010; Elsevier
• Subjects: Agricultural machinery and engineering; Agronomy. Soil science and plant productions; Biological and medical sciences; Canada Western Red Spring wheat; crop damage; detection; Fundamental and applied biological sciences. Psychology; Generalities. Biometrics, experimentation. Remote sensing; grain sprouting; hyperspectral imagery; image analysis; spring wheat; sprout damage; Triticum aestivum; visible/near-infrared hyperspectral imaging; Canada; North America; America; Bioengineering; Monocotyledones; Vegetals; Multiple; Spring crops; Spectral band; Hyperspectral imagery; Use; Electromagnetic wave; Cereal crop; Spectral data; Near infrared radiation; Engineering; Gramineae; Angiospermae; Visible radiation; Spermatophyta; Agriculture; Damage; Detection; C3-Type; Triticum aestivum
• ISSN: 1537-5110; 1537-5129
• DOI: 10.1016/j.biosystemseng.2010.03.010

A visible near-infrared (VNIR) hyperspectral imaging system (400–1000 nm) was used for detecting sprouted and severely sprouted wheat kernels. Canada Western Red Spring (CWRS) wheat kernels were individually scanned using the hyperspectral imaging system. Average spectra of sprouted and severely sprouted kernels had higher reflectance responses compared to sound kernels in the wavelength region above 720 nm. The ratio of the reflectance at 878 nm to that at 728 nm could be used to identify sprouted from non-sprouted kernels. The Principal Components Analysis (PCA) loadings plot identified four wavelengths that contributed to distinguishing the different quality of wheat kernels. Using the morphological features smoothness, size and value range of the third principal component score image, severely sprouted kernels were clearly separated from the sound kernels. Utilising a classification procedure, incorporating both spectral and spatial features, 100% of the sound kernels, about 94% of the sprouted kernels and 98% of severely sprouted kernels were correctly classified. 1 1 GRL Publication # 1028.