Fuzzy minimum-maximum neural network clustering method based on content image

Abstract

The method provides a fuzzy minimum-maximum neural network clustering method based on a content image. All samples are read at one time during initialization, and the iteration of an algorithm is the operation of a whole sample set at each time. The method is independent of the input sequence of samples. During operation, the method is a process of combining low-level hyperboxes into high-level hyperboxes, and the number of high-level hyperboxes continuously decreases in an operation process of the method, i.e., the successive decreasing of the cluster number, thereby improving the efficiency of the method. According to the invention, the layering idea of a layering algorithm is introduced into a fuzzy minimum-maximum clustering network learning algorithm, and the layering fuzzy minimum-maximum clustering algorithm which meets the image clustering requirements better is proposed, thereby enabling the method to better meet the requirements of searching based on the content image. A simulation result indicates that the method provided by the invention is feasible, and is better than a conventional neural network.

Description

technical field [0001] The invention relates to a neural network clustering method, in particular to a content image-based fuzzy minimum-maximum neural network clustering method HFMM, which belongs to the technical field of clustering algorithms. Background technique [0002] Since the 1990s, Content-based Image Retrieval (CBIR) has become a research hotspot in the field of image retrieval. Its main research content is based on image processing, according to the visual feature information such as color, texture, shape and object spatial relationship contained in the image, to establish the multi-dimensional feature vector of the image, and then perform similarity query on the image according to these feature vectors . Different from traditional text-based retrieval methods, CBIR also integrates image understanding, pattern recognition, computer vision and other technologies, and has the following characteristics: (1) directly analyze image content, extract features and sema...

AI Technical Summary

Problems solved by technology

However, FMM also has the following disadvantages: (1) FMM has the problem of sample input order dependence (odependent), that is, the same sample set is input into the network in a different order during training, which often leads to different clustering results; (2) FMM's The final clustering result is a collection of a series of hyperbox fuzzy sets, which is not as easy to retrieve as the tree structure formed by the hierarchical algorithm

Images