Discrimination of Kernel Quality Characteristics for Sunflower Seeds Based on Multispectral Imaging Approach

摘要

Multispectral imaging in the visible and near-infrared (405-970 nm) regions was tested for nondestructive discrimination of insect-infested, moldy, heterochromatic, and rancidity in sunflower seeds. An excellent classification (accuracy > 97 %) for intact sunflower seeds could be achieved using Fisher's linear discriminant function based on 10 feature wavelengths that were selected from the original 19 wavelengths by Wilks' lambda stepwise method. Intact sunflower seeds with different degree of rancidity could be precisely clustered by multispectral imaging technology combined with principal component analysis-cluster analysis (PCA-CA). Our results demonstrate the capability of multispectral imaging technology as a tool for rapid and nondestructive analysis of seed quality attributes, which enables many applications in the agriculture and food industry.