Font Size: a A A

Research Of Data Partition For Craniofacial Model

Posted on:2013-01-26Degree:MasterType:Thesis
Country:ChinaCandidate:K P LiFull Text:PDF
GTID:2248330374472134Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
Based on the statistical model of Face Appearance Reconstruction, the entire craniofacial data, which as a training sample to participate the training of the statistical model, will not only lead to small sample sized problems, but also result in some loss of local features, affecting the recovery accuracy. The main way to solve this problem is to segment craniofacial model reasonably, effectively, quickly and accurately into regions, and then implemented statistical recovery. This paper refers to putting the craniofacial model data partition into pattern classification problems, and proposed data partitioning method based on kernel methods supporting vector data description. The method can accurately partition according to the craniofacial characteristics of the model region, and can be applied to complex regional contours and shapes, being robust and flexible. The research is developing as follows.(1) Craniofacial data processing was studied. According to the physical structure of craniofacial model, the physical structure of the regional division was given.(2) Proposed automatic selection method of the model parameters based on a consistent state, which avoids the repeated and tedious process of adjusting the SVDD model parameters manually. Improve the efficiency of the craniofacial model data partition method.(3) This paper described the main method of calculating the principal curvatures and normal, and calculated the vision saliency, as well as the mean and variance of Effective Energy.(4) Proposing a multi-scale processing method, this method could enhance the using of feature information, and could effectively recover characteristics degradation due to various factors.(5) This paper had given Craniofacial Model Data Partitioning process based on kernel methods support vector data description, and designed, implemented a craniofacial model data partition system, completing the fast segmentation of the Craniofacial model. Besides, the segmentation results had been analyzed. Experiment shows that this method can be more reasonable and precise to get the characteristics of partition.This research is supported by National Natural Science Foundation of China (60736008).
Keywords/Search Tags:Craniofacial reconstruction, Craniofacial data partition, Support Vector DataDescription, Model selection, Feature processing
PDF Full Text Request
Related items