Study On Personalized Human Model And Cooperative Motion On Its Limb

In recent years, the generation of3D personalized human body and the cooperative motion of human limb have become the important development direction for the Virtual Reality field. As a core part of3D virtual garment CAD system, it allows user to select and modify the styles and sizes of garments, and presents to the user by using the method of3D display. In addition,3D reconstruction of different human poses will lay a solid foundation for dynamic display and comfort evaluation on garments. If the3D human models obtained by the method of personalized generation and cooperative motion of limb can accurately reflect shape indicators and posture parameters depends on the generation principle and reconstruction method. Therefore, to research theories and methods of personalized human body generation and cooperative motion of human limb for speeding up the development of virtual reality and3D garment CAD technology has important theoretical significance and application value.3D personalized human body generation and cooperative motion of human limb include the theories and methods of the establishment of human models database, generation of personalized human body, extraction of human fringe skeleton and3D reconstruction of human poses. Our overall aim is to construct the platform of Internet-oriented customization and display garments, and research related algorithms on the basis of it.The generation method of different human shapes based on3D entity’s multi-direction cooperative deformation is proposed. The used template and target human models are firstly selected in the existed database of human models, and then the models are segmented into3D entities by using figure design modeling software. Various parameters are demonstrated by using the common properties of various entities. Shape characteristics of template entity and variation range of target entity are then extracted. Subsequently, direction curves of entities are generated using proposed entity curve interpolation algorithm. Various entities are obtained by the proposed method of minimum distance criterion, and accurate different human shapes are then generated.The generation method of personalized human model based on3D entities and multi-direction human contours is proposed. The principle of pixel size on human body image using the method of reference is firstly demonstrated. Based on the multi-direction human body contours, entity direction measurements are extracted by the proposed method combining part partitioning with human skeleton segment. The similar human entities are determined by searching for the human entity database using the graded matching based on height and entity direction measurements, and they can provide a basis for the deformation of human models. The3D human models with the shape characteristics of target model are generated by the proposed deformation method based on multi-direction human body contours using the located critical feature points and contours. The extraction method of human fringe skeleton based on the improved Hopfield neural network and direction features is proposed. The torso of human model is firstly located using the depth direction feature of model’s triangle meshes based on the definition of coronal plane. The feature points of torso on target human body are located using the feature points’location method combining the improved Hopfield neural network with direction feature. Subsequently, the feature points of limb are extracted accurately by the proposed limb direction feature using the algorithm of inclined section lines. Finally, the accurate fringe skeleton of target human body is extracted precisely based on human skeleton model.The cooperative motion method of human limb based on conformal transformation is proposed. The joint feature points of target human body are located by the proposed method combining color histogram with distance classification. The3D coordinates of skeleton points are obtained by the method of iterative calculating scale factor based on interconnected skeleton segment using the model of human skeleton points and segments. Then, the cooperative motion method of human limb based on conformal transformation is proposed based on calculation of direction and perspective of limb motion. Therefore, the limb cooperative motion and3D human posture reconstruction can be obtained by the above methods.The two modules of the generation of3D personalized human models and cooperative motion of human limb have been developed. Using the above modules, a lot of examples including the generation of entity direction curve, the extraction of human body contours based on human body images, the generation of3D personalized human model, the extraction of3D human fringe skeleton and the reconstruction of3D human posture are given. Furthermore, it also proves the correctness and efficiency of the proposed methods.