Machine Learning Based Identification of Tomato Leaf Diseases at Various Stages of Development

Abstract

Mosaic, early blight, late blight, Septoria virus, leaf mold, Brown spot, and spider mite are the nine common types of tomato leaf diseases. The early and accurate analysis of tomato leaf disease can increase the productivity and quality of the tomato product. The existing research in image processing does not guarantee an accurate diagnosis of the disease. Also, existing methods are complex. In this paper, an accurate and robust method for tomato leaf disease identification as well as classification into various stages of development using machine learning is proposed. The work is carried out in two stages. Firstly the tomato leaf images will be classified into appropriate disease types. Then in the second phase, the tomato leaf disease is diagnosed at various stages of development. Identifying the stage of development of tomato leaf would help to decide the type and amount of treatment required for the plant. The diseased leaf images which are taken from the PlantVillage dataset have been classified into high, medium, low, and normal severity grading. The images are preprocessed using median filtering. For feature extraction, the system using shape, color, and texture features is evaluated. The performance evaluation is also done on various classification techniques including SVM, KNN, Naive Bayes, Decision Trees, and LDA. The research indicated that the proposed model provides a robust solution for tomato leaf disease severity grading.