The Extraction Of Semantic Structure And Gesture Recognition For 3D Human Models

Human beings as the main part of the social environment, contains rich variety of interaction information, human behavior research contains extremely profound social value and application value. Especially, with the popularity of 3 d scanning technology, the number of 3 d human body model is increasing, and more and more widely used in the field of the motion simulation, entertainment, film and television, advanced human-computer interaction and so on.Therefore, how to extract the information of three-dimensional human body structure efficient and accurate, realize the analysis and reuse the human model, has become an important research content. Especially, the development of an effective structure of human body 3 d feature extraction methods have become the basis of further in-depth study of human behavior. Human body posture matching by the extracted structure can be a better understanding of model behavior, in the field of motion analysis, the human body model retrieval, human-computer interaction, etc, have very strong practicability.In view of the present most shape analysis algorithm, it is not accurate enough to extract human body structure, and the limitations is not general applying to diverse human body posture, this paper presents a shape analysis method of feature extraction in the human body structure, effectively keep the topology of the model structure and the semantic knowledge, and stability can be used in different gestures of the human body, for subsequent human body model compression, retrieval and identification laid a foundation. The paper main work includes:1. In this paper, the fusion of the surface model geometry characteristics and the more stable space volume features, proposed a analysis method of model morphological structure, not only keep the whole topology of the model, and more highlights on the surface of the local geometric detail with strong robustness.2. Based on the model surface geodesic distance and the characteristics of internal space volume measurement, proposed a structures detection method based on skeleton tree, and based on further anthropometry prior semantic knowledge, refined model hierarchy. This method can effectively extract the different features of human body model, and implement the semantic analysis of human model.3. Proposed a gesture recognition algorithm based on the human body structure analysis. First of all, the integration of geodesic and structure characteristics to judge and detect the human body skeleton point, greatly improving the efficiency of the algorithm, and then, implement the rigid registration and recognition with test skeleton and standard posture skeleton based on ICP algorithm, effectively simplifies the model data, improve the efficiency of the posture matching.