Fuzzy Classification Method for Knife Detection Problem

In this paper we propose a new approach for pattern recognition problems with non-uniform classes of images. The main idea of this classification method is to describe classes of images with their fuzzy portraits. This approach provides good generalizing ability of algorithm. The fuzzy set is calculated as a preliminary result of algorithm before crisp decision or rejecting that allows to solve a problem of uncertainly at the boundaries of classes. We use the method to solve the problem of knife detection in still images. The main idea of this study is to test fuzzy classification with features vectors in real environment. As a feature vectors we decided to use selected MPEG-7 descriptors schemes. The described method was experimentally validated on dataset with over 12 thousands images. The article contains results of five experiments which confirm good accuracy of the proposed method.