Diagnosis of diseases on cotton leaves using principal component analysis classifier

This paper addresses the problem of diagnosis of diseases on cotton leaf using Principle Component Analysis (PCA), Nearest Neighbourhood Classifier (KNN). Cotton leaf data analysis aims to study the diseases pattern which are defined as any deterioration of normal physiological functions of plants, producing characteristic symptoms in terms of undesirable color changes mainly occurs upon leaves; caused by a pathogen, which may be any agent or deficiencies. The predictions of diseases on cotton leaves by human assistance may be wrong in some cases. Using machine vision techniques, it is possible to increase scope for detection of various diseases within visible as well invisible wavelength regions. After implementing PCA/KNN multi-variable techniques, it is possible to analyse the statistical data related to the Green (G) channel of RGB image. Green channel is taken into consideration for faithful feature collection since disease or deficiencies of elements are reflected well by green channel. In most of the cases diseases are seen on the leaves of the cotton plant such as Blight, Leaf Nacrosis, Gray Mildew, Alternaria, and Magnesium Deficiency. The classification accuracy of PCA/KNN based classifier observed is 95%.