Pattern recognition based on fuzzy cluster for recognizing garment style in the photo

It is an important work to extract a garment style from garment photo, which can get much valuable information to guide manufacture and design for enterprises. But this research is so difficult that no one can study it successfully so far. This paper proposes a set of computer recognition method based on cluster pattern recognition and silhouette knot points detecting. Firstly, garment photos are analyzed using ¿line by line scanning¿ and ¿8-points connecting¿ ways and built a pattern silhouette, and then the silhouette is classified to certain garment style using cluster pattern recognition. Secondly, the data structure of silhouette is established. The garment silhouette is expressed by ring data structure and line-line data structure. Thirdly, the garment silhouette is divided into different dress pieces using fuzzy cluster recognition theory, and each separated dress pieces is recognized. Finally, an example prove that this method is effective for unfold garment in the photo.