TOMATO PLANT DISEASE CLASSIFICATION USING LOCAL PATTERNS.

Agricultural sector has significant impact on the people health and on the economy of the world. Climate variation is important reason in causing plant diseases hence, affecting the estimated crop production. Prior detection of plant diseases is utmost important for improving the quality and quantity of production within the due course of time. In this paper, this challenge is addressed by automatically detecting tomato diseases from the hand-crafted features extracted from the plant leaves and machine learning classifiers. Different frequency bands are extracted using Gaussian filters and local statistics of leaves are captured using patterns to design frequency decomposed local ternary pattern (FDLTP). It provides a fast and accurate solution to avoid uncertainty in the farm production. Benchmarked dataset of Taiwan tomato leaves is used to verify the results. Performance of machine learning classifiers as well as deep learning solutions are compared, and 95.6% accuracy is obtained using proposed feature along with k-nearest neighbor classifier. It is a quick and easy to deploy method for real time application. [ABSTRACT FROM AUTHOR]