Automatic visual inspection of corn kernels using principal component analysis

Abstract

During the processing of corn, corn kernels are inspected for the purpose of determining the quality and stock value. There are many kinds of defects which can reduce the stock value whereas the difference between defects or an acceptable corn kernel can be sometimes minor. Food inspection uses computer vision more and more nowadays, but above forms a challenge when automatically inspecting corn kernels. This paper presents an algorithm based principle component analysis (PCA) to decide whether a corn kernel has an acceptable or rejectable quality. Experiments using 400 images show that the method is promising (89% of success using HSV color space and 99 histogram value), but extensions are recommended to further improve results.