Automatic sushi classification from images using color histograms and shape properties

Self-service sushi vendors are becoming popular with Thai people. The cashier operation is usually a bottleneck in a purchase process and can cause long queues at the cashier counter. This paper proposes an automatic sushi point-of-sale system that captures sushi images from a webcam to identify types of sushi and calculate the total cost of that purchase. The system segments each piece of sushi from the background using color information. Histograms of color in RGB and HSV color spaces and shape properties of each piece of sushi are used as features for k-Nearest Neighbor classifier. Experimental results showed the accuracy of the proposed system at 93.7%.