抄録
We have proposed an effective pattern recognition method for contour image by extracting general and local features. In this method, firstly object is seen perspectively to identify class and secondly seen in detail to identify individual object. The images in the database must be classified to carry out flexible pattern recognition. This paper describes a method to classify images in the database by using Self-Organizing Maps (SOM) and to determine a prototype each class. And we show the result of applying our system to illustrated insects reference book.
