The project was a collaboration with Govt. of India, aimed to introduce automatic grain quality assessment from an image of spread out sample of grain to assist farmers in getting fair price of their produce. Using binary thresholding for segmentation, followed by an SVM classifier model, we were able to distinguish between grain and impurities with an accuracy of 88%.
Input image is first Pre-Processed(a) to give a binary image, which is then Segmented(b) to give different particles, which are then Classified© as grain/impurity.