2006 Volume 18 Issue 2 Pages 145-153
Recently agricultural product grading machines based on imaging technologies have been spread widely at cooperative associations in Japan. In this paper, experiments on feature extraction of tomato color and shape conducted by use of top images of various color and shape tomato, because these are the most important information for agricultural product grading to enhance its market value. From above, following results were obtained. 1) An index based a red and green color component ratio gave similar results with human evaluation. 2) Indices using distances between center of gravity and boundary pixels were effective for shape evaluation to eliminate influence of ellipse shape by a transformation method from ellipse to circle. 3) It was possible to grade various shape fruits by the combination of three different indices, because they are complement to each other for uncertain shape evaluation. 4) Adjusting the setting values of indices depending on fruit size made possible to get practical grading results, which were similar to the human evaluation.