DISCRIMINATION OF PADDY VARIETIES USING WAVELET FEATURES

This research proposes an algorithm to implement feature extraction technique using wavelet, and use the extracted coefficients to represent the image for classification of Grains. A total of 75 Wavelet features were extracted from the high-resolution images of paddy grains. The wavelet features were employed along with ANN to identify paddy varieties. This research is aimed at comparing Single-level discrete 2-D wavelet transform and Multilevel 2-D wavelet decomposition, using ANN for discriminating Indian Paddy Varieties and also evaluate variety-wise classification of individual grains. An evaluation of the classification accuracy of wavelet features and ANN was done to classify four Paddy (Rice) grains, viz. Karjat-6(K6) and Ratnagiri-2(R2), Ratnagiri-4(R4) and Ratnagiri-24(R24). All feature models were tested for their ability to classify these cereal grains and the most suitable feature was identified from the Wavelet features for accurate classification. Single-level discrete 2-DWT gave the best classification using ANN and more accuracy can be obtained by increasing the levels of decomposition.