Research On Weight Distribution Automatic And Visual Topic For Image Annotation

The rapid development of mobile Internet and the popularity of smart devices, will produce a large number of images in people's life. In the initial stage, based on the artificial marking method has been unable to complete the work level of hundreds of millions of number image annotation, the emergence of automatic image annotation technology solved the problem very well. The technology can replace artificial efficient image annotation work, save time and effort.In this paper, aiming at the problem of automatic image annotation techniques result is unsatisfactory, the image low-level feature of learning and artificial intelligence. Based on the research of the feature fusion and genetic algorithm, and put forward a kind of based on feature fusion weights allocation algorithm automatically, can be more efficient for image description. Further, on the basis of the algorithm and automatic image annotation is proposed based on image visual theme, improve the performance of the annotation.In this paper, the research of the automatic image semantic annotation are as follows:(1) Study the image feature and the basic features of the low-level feature extraction method, during the fusion of image features, how to use genetic algorithm to make the weight adjustment and automatic distribution.Put forward a kind of based on feature fusion weights allocation algorithm automatically, can better describe the fused images;(2) In view of the image semantic annotation problem, will be established between image basic features and high-level semantic association.In order to reduce the gap between the two, introduce the concept of visual themes.The image as the combination of multiple image region, and the corresponding multiple annotation words, and find there are images of the visual theme. On the basis of image visual theme and combining the weight allocation algorithm automatically, puts forward a new method of image semantic annotation.By analyzing the experimental results and comparison, based on the weight distribution automatic and visual theme image semantic annotation method is feasible, and the method to improve the effect of the image annotation.