Grain Quality Classification System

Abstract

Exploring the possibility of using technology for grain quality classification is necessary to the consumer market to protect consumers who susceptible to any form of contamination that may occur in the market. This research focused on providing a better approach for identification of rice quality by using neural network and image processing concepts. In this research, a four layered back propagation neural network has been developed for quality classification. Also, it has been extracted texture and color features from rice images as the feature values to the neural network. The texture feature extraction was based on the Gray Level Co-Occurrence Matrix and the color feature extraction was done by using both RGB and HSV color space. Thirty one features were used for discriminate analysis and the classification accuracy for the training set were 94%, 93%, 71%, 68% with respect to premium, grade1, grade 2 and grade 3. When the model was tested on test data set, the classification accuracy was 91%, 90%, 72%, 69% for premium, grade1, grade 2 and grade 3 respectively.