Fuzzy Knowledge Modelling for Image-based Paddy Disease Diagnosis Expert System

Abstract

One of the most crucial agricultural activities in Malaysia is the paddy plantation. There are several factors that cause paddy-rice production becomes slow and less productive and one of the factors is disease. Traditionally, paddy farmers use their traditional method to determine the type of disease that occurs in their paddy farms. Paddy farmers are being trained by Malaysia Agricultural Research and Development Institute (MARDI) experts to recognize paddy diseases in order to ensure that early prevention or treatment is taken. The main feature of paddy leaves that determine the paddy disease is the type of lesion and colour of paddy leaves. These are among the symptoms that cause the paddy diseases. Currently, the expert plays the major role in determining the type of disease through his experience in recognizing the form of abnormality of the leaves. The identification of paddy disease through its leaves may cause errors due to the vagueness in appearance of lesion and colours of the infected paddy. To develop an expert system for diagnosing the paddy diseases, the issues of vagueness need to be handled. In this paper, we present a fuzzy logic approach to handle the uncertainty and vagueness of paddy diseases. Ten linguistic variables are introduced to determine the type and level of seriousness of the diseases such as colour and type of lesion, colour of boundary and leaves, percentage of damaged lesion and paddy age level. These variables are obtained through series of interviews held with the paddy disease expert. A hybrid approach consisting of membership exemplification, direct and reverse rating methods to develop the membership function is employed in different phases. We present a fuzzy knowledge model for image-based paddy diseases diagnosis expert system which shows that this approach is able to produce a reliable membership function.