Detection of Leaf Disease Using Principal Component Analysis and Linear Support Vector Machine

Agriculturalists find difficulties in classifying the diseases in leaves. In olden days, farmers detected the leaf diseases by observing the leaf by its appearance which does not provide accurate results. The simple, efficient and computationally effective plant disease identification system should be introduced to help farmers to detect the leaf diseases with high accuracy. In this paper, the diseases in the leaves are detected by extracting high quality features of the leaf by Principal Component Analysis. It is an image retrieval technique used to extract the high dimension feature of the image without losing its originality. The features such as perimeter, minor axis, area, major axis, equivalence diameter etc., are extracted from the diseased leaf. The input image is pre-processed by removing the noise using median filter, extracting the edges using Sobel filter, identifying the direction of the leaf using Gabor filter and segmenting the image by threshold segmentation. Finally, Linear Support Vector Machine classifier identifies the support vectors to identify the disease. Experiments are carried out using the datasets of tomato plant leaves where tomato is cultivated large scale in Coimbatore, India. Powdery mildew and early blight diseases are the most common disease in tomato which is detected and analysed in this work. The results obtained shows that there is an increase in accuracy when compared to other methodologies.