To retrieve 3D digital data from a large database of digital images, we normally use three-dimensional Content Retrieval systems that are used methods adding description text to 3D content files. Thus, related 3D files can be found through the retrieval system. Through research, the automated process creates test descriptions for 3D contents since the existing manual processing could not annotate 3D files so efficiently. The demand for 3D content has been tremendously increased to retrieve 3D information. Among the methods available in 3D content retrieval, the shape matching method that compares and contrasts similarities among the models have won appreciation from one and all. The technology has accepted different kinds of methods such as content-based and text-based method. Content-based methods perform on the description of model’s features rather than the user’s requirement. Text-based retrieval methods have been used in some search engines. But the drawback in this retrieval method is - the match is done based on the identifier of the node content and neglects the semantics of the user given keywords. In spite of good annotations, the retrieval results are not appreciable. The research analysis lay much emphasis on the study of semantic-based retrieval methods. The CYC is regarded as one of the typical semantic knowledge bases. Care must be taken while relying on the semantic of the 3D model. There lies a couple of difficulties. Firstly, many 3D models cannot describe semantics with a good number of keywords. Secondly, many semantic words have different meanings. So, the models do have sufficient knowledge base.
Basing on the semantic similarity among the retrieval terms and semantic tree nodes with WordNet tool and return the model that has a strong semantic correlation. Annotations are used to elucidate semantics of the models. Finding the similarity in the process of text mode model retrieval is translated into the computation of semantic similarity of words.
The semantic tree is designed based on the model-based level classification. The semantic tree is very flexible for adding and deleting the tree node according to the specific real-time application. Semantic tree nodes correspond to the classification of different size in the model library.
The necessary conditions to construct a semantic tree are hard and the brief semantics of all the models in manual modes are indispensable.
Comments
Post a Comment