Multi-resolution mobile vision system for plant leaf disease diagnosis

The process of detecting plant disease by human naked-eye is difficult and very expensive practice, particularly in developing countries like India. Designing and providing a fast-reliable automated mobile vision based solution for such tasks, is a great realistic contribution to the society. In this paper, a mobile client–server architecture for leaf disease detection and diagnosis using a novel combination of Gabor wavelet transform (GWT) and gray level co-occurrence matrix (GLCM), opens a new dimension in pattern recognition, is proposed. Mobile disease diagnosis system represents a diseased patch in multi-resolution and multi-direction feature vector. Mobile client captures and pre-processes the leaf image, segments diseased patches in it and transmits to the Pathology Server, reducing transmission cost. The Server performs the computational tasks: GWT–GLCM feature extraction and k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$k$$\end{document}-Nearest Neighbor classification. The result is sent back to the users screen via an SMS (short messaging service) with an accuracy rate of 93 %, in best condition. On the other part, paper also focus on design of Human-mobile interface (HMI), which is useful even for the illiterate farmers, to automatically monitor their field at any stage by just a mobile click. Android is currently used to run this system which can be easily extended to other mobile operating systems.