The price of crops depends on the grade. The grade is determined from various factors such as color, shape, size, and scratches. It takes a lot of time and effort if farmers decide the grade. This study aims to automate shape grade judgement. In the prior method, binarization and contour extraction are performed on the tomato image to define features called flatness, variant, sharpness, and irregularity, and a threshold is set for each grade. However, due to the correlation between the features, the grade judgement has not been made with sufficient accuracy. In this study, we propose several methods focusing on contour features and compare these accuracies.