Mining of RGB Features of Leaf Blast Disease Infected Image Using Fuzzy Inference System in Rice Crop

Data mining is a process to extract useful information for decision making from available set of data. Blast is a fungal disease of rice and occurs in all the rice-growing regions. It causes complete destruction of the rice crop. Present day applications require automation of process to interprets and analyze the information by using various kinds of images and pictures as source of information for interpretation and analysis. The fuzzy set theory is incorporated to handle uncertainties and fuzzy clustering is a powerful method of data mining. FIS is an expert system to approximate input-output mapping according to defined rules. In this proposed approach, leaf blast infected image in rice crop is acquired by digital camera. Image's RGB features of pixels has been extracted and further used for clustering. This clustered data is categorized as Not Affected (NA), Medium Affected (MA) and Highly Affected (HA) by leaf blast disease according to colour information and previous acquired knowledge. For each of these three categories, minimum, medium and maximum RGB ranges has been evaluated. For that linguistic term RMIN, RMED, RMAX, GMIN, GMED, GMAX, BMIN, BMID and BMAX assumed and according to range of linguistic term, triangular membership function has been taken as an input in FIS. For easier implementation of FIS, range of NA, MA and HA is also assumed and taken in form of triangular membership function. According to evaluated data, assumed data and 14 fuzzy inference rules , FIS with three input Red, Green and Blue and any one possible outcome from NA, MA and HA has been implemented. After implementation, random 60 pixels RGB information of another image has been send as input in FIS for testing and it is successfully categorized by FIS, which help to approximate loss caused by leaf blast disease in rice crop.