Multivariate image analysis for defect identification of apple fruit images

External defect identification of apples involving humans suffer from disadvantages which can be greatly reduced using machine vision applications. The bruises appearing on the surface which may result due to post harvest handling may not show up in images taken in the visual range. Multispectral imaging technique works with images of object obtained in several bands in the visual and infrared regions of the electromagnetic spectrum. The images taken in narrow bands are highly correlated. For decomposing these correlated data principal component analysis is used. The score space and the image space can be used for feature extraction and further image segmentation. This paper discusses the multivariate image analysis technique applied to the defect segmentation of apple fruit and explains the procedure adopted with the results obtained.