Fuzzy数合成运算应用于图象分类

APPLICATION OF COMPOSITE OPERATION OF FUZZY NUMBER IN PATTERN CLASSIFICATION

提出了一种图象有监分类的新方法.该方法基于图象的特征因子对于图象的类别的模糊性,因而可以用一个隶属函数来表示.根据模糊数的合成运算,考虑到全部特征后,给出图象隶属于所有的类的可能分布.若选定某一阈值λ_α进行截取,就可能确定出该图象属于某类的最大可能性.实例表明用该方法对图象进行的分类与用最小距离分类得出的结果一致.

This paper presents a new method of supervised classification of patterns. Based on the fuzziness of pattern features with respect to pattern classes, the features can be represented by membership function. According to the composition operation of fuzzy number, when all fertures are considered, the all possible contributions of pattern memberships to deferent classes are derived. Then if certain threshold is selected, the maximum possible distribution of pattern membership to certain class can be found. Experiment shows that the pattern classification results by this method agree to that by minimum distance classification method.