1996 年 58 巻 Supplement 号 p. 501-504
The purpose of this study aimed to develop an automatic sorting system for vegetable fruits. It is well known that vegetable fruits, such as strawberries or green peppers, are with very irregular shapes. The sorting operation for this kind of vegetable fruits still remained at manual level. For this reason, the image processing and neural network were applied to judge the shape of vegetable fruits in this study in order to develop a automatic sorting system.
In this report, specially the feature extraction of green peppers in vegetable fruits was presented. The three methods on the feature extraction was proposed from a binarization picture of green peppers. These feature extraction were applied by the technologies of image processing and neural network in this test.
The experiment result showed that these feature extraction by using the neural network software was able to judge green peppers with higher accuracy comparing with manual operation. The coincident rate of a shape judgment on green peppers was 93 to 95% by the method II and III of the feature extraction. It needs the target of the coincident rate will be over 95%. The new method will be reported at the next chance because a new method is developing for attaining the target.