Paddy Disease Recognition using Image Processing and Radial basis Function Network

Objectives: To automate paddy disease recognition process by statistical pattern recognition model. Methods/ Statistical Analysis: This paper presents, designing of a radial basis function network model to automate paddy diseases pattern recognition. Total 5 categories of diseased infected paddy images along with non infected images had been captured by the digital camera in an uncontrolled environment in day lighting. The captured images had been cropped to select aregion of interest and resized to reduce space/time complexity of the system. These preprocessed images had been segmented using multi-level of thethreshold. The discrete wavelet features of red, blue and green components of highest inensity level segmented images had been extracted. The 80% of the extracted features had been used to train the model and remaing 20% had been used to test the performance of the model. Findings: The results of the study show the advantages of the proposed method compared to some other existing methods in terms of recognition efficiency and generalization. The average diseased/non-diseased pattern recognition using radial basis function network model is 97.9% for training data sets and 95.5% for testing data sets. The method is implemented in an uncontrolled environment instead of laboratory setup give advantages of easy generalization. Application/Improvements: The method can be used to design aplant diseases monitoring system for farmers to inspect and asses threats of diseases and make on-location decision to cure and/or control the spread of diseases. The government agencies can also use this monitoring system to generate periodical reports/warnings from acomputer database for high threat potential diseases.