Research On The Method Of Posture Conversion Of 3D Human Point-cloud

The posture conversion method of 3D models which has been widely used in the traffic, medical, entertainment and other fields is important for perfecting the expression ability of 3D models. This thesis conduct a research on the posture conversion method of 3D human point-cloud which was obtained by the 3D human body scanner, through this method which is consisted of several technique processes such as body part segmentation, driving skeleton extraction, standard posture description, motion data generation and joints binding, the 3D human point cloud could be convert to a standard standing posture model and a standard siting posture model. This research laid a foundation for some future studies such as critical dimension measurement and 3D animation.The major work and innovation points of the paper are as follows:1. Focused on the segmentation of human point-cloud, a Multi-constrained segmentation algorithm based on classified skeleton, geodesic distance, feature points and posture analysis was proposed. The segmentation results of this algorithm in different posture and different size human point-cloud were consistent with the human visual perception.2. By reference the 3D motion data files, a drive skeleton structure of 3D human point cloud was confirmed. Some specific location algorithms for the joints of the drive skeleton were designed and a complete drive skeleton which was consistent with human topology could be extracted from 3D human point-cloud with different posture and different size.3. The skeleton of standard standing/siting posture was constructed based on the information of drive skeleton and motion data was generated from the difference between drive skeleton and standard posture skeleton.With different joint binding method, the posture conversion could be implemented basically.