The Construction Of Condensed Semantic Tree Model And Its Application In The Analysis Of Video Key Frame Clustering

With the development of Multimedia technology and Internet, the application of digital image has bee increasingly widespread and it leads to the image retrieval technology becoming more important. Since the last decade,people's need of image retrieval has been increased. In order to retrieve these huge image data effectively,Content-based image retrieval technology become popular.The most existed Content-based image retrieval systems use traditional low-level features such as color,texture and shape to describe the image content,which are usually represented by statistic data. Actually,there are big differences between these statistic data and the image content which people understand. Because people's understanding of image content isn't based on statistic and the image content is fuzzy,it couldn't be represented by simple vectors. These problems lead to deflection of the image representation and people's Understanding,which is called semantic gap. Usually,we couldn't get satisfied result if we only use the low-level feature to retrieve.So,how to describe image and to make it coincide with people's understanding become the key point of improving retrieval accuracy. In the point of cognition,people's understanding and description of image content is on semantic level. How to reduce'semantic gap',how to accurately represent content semantic of image and retrieval intention of people becomes important and critical.In this paper, we focus on the processing of the semantic information derived from the news video key frame image. After analyzing the characteristics of the meaning of the key frame images, the semantic information of images is collated and expressed. Different from the general images, the traits of news video key frames is that the categories are limited and the meaning is explicit. So in this paper, the condensed semantic tree model is chosen. The condensed semantic model not only saves the storage capacity of the database, but also makes the semantic processing of information more rapid and flexible; the efficiency of annotating the images is improved.Because of the independent existence of semantics disengaged from knowledge framework, the implicit concept model can be expressed through describing the concepts and relations of the objects with the help of ontology. in recent years, the research on ontology has been widespread concerned in the field of information science. Ontology is the explanation or description of an object, the abstract nature of objective reality is concerned. It includes not only the objects, but also the relationship between them. The accurate expression of the relations among image semantic concepts is strongly influenced by the introduction of ontology knowledge, which made semantic information is no longer isolated. There are a variety of links among the Image semantic concepts. The image information can be expressed better with the ontology information, which accords with the psychological reaction of people understanding images.According to the query complexity of the user, image content can be divided into three levels. So there are three types of expression of the image information. The object layer includes the object semantics and the spatial relationship semantics, while semantic layer includes scenes semantics, behavioral semantics and emotional semantics. In this paper, the specialties of news video information is taken into account when the semantic tree model is built, so the emotional semantics is not going to be proceeded. Object semantics, behavioral semantics and scene semantics are chosen to be organized into three modules. A semantic tree model is constructed with the introduction of ontological knowledge, which enhances the versatility; For the annotation of each image there are some reference models, which enhances the professionalism.The application of semantic tree model not only enhances the influence of the background knowledge of the images and the relationship among concepts on the description of images, but also enhances the relationship between images. The three modules can be combined into different types of image semantic concepts. The module in different types contains different semantic concepts which are organized according to the semantic information of this type. Through this method, the reusability of the modules and semantic concepts is improved ,and the capacity of the database is reduced.In this paper, semantic features of each image is expressed in the form of vector. In this way, the similarity of the semantic features of images as well as the low-level features can be measured by the Euclidean distance or be used with the related feedback algorithm to improve the retrieval result. The semantic features of the image is expressed in the form of soft vector that the evaluations of each attribute of the soft vector are different. Among the attributes, there are some concepts are the most important while some are less important, as well as in an image. In this way, the main semantic information and secondary information are able to be expressed more clearly. The evaluations for each of the three levels in the semantic tree model are the proportion that level appears in the image, so that we can determine the probability that each image belonging to a certain category. The experimental results showed that the soft vector representation method is better than the Boolean method.In the end, Visual C + +6.0 software is used in the Windows operating system to apply the semantic image clustering algorithm. The method in this paper is compared with other methods, the experimental results show that, the completeness and the accuracy of the clustering result is increased which verified the effectiveness of the method in this paper.The News Video key frame model constructed in this paper is still in its early research stage, which needs more research. For example, the semantic tree model needs improving continuously; A better method to evaluate the concepts attributes of the semantic vector needs to be discovered; The soft vector is sparse and high-dimensional, whose dimensions need to be reduced in the later research.