The modeling of virtual humans is a challenge that requires the solution of several problems such as the reconstruction of a scanned model, the association of the geometric mesh with the bones of a skeleton (rigging) and texturing. In order to generate population diversity, it is also important to allow the use of techniques to deform the generated or captured models. Therefore, a literature review of these problems was carried out, searching for deformation processes based on physiological aspects of the body (e.g., physical vigor, age, body mass), in order to produce credible and varied virtual humans. We find work that performs the tasks of capturing, texting, rigging a model or deforming the mesh of the models as isolated processes, but they generally do not contemplate them all. In particular, the geometric deformations of the mesh of virtual human models have different levels of manipulation (direct or by controls), as well as distinct focuses (animation and crowd generation). Thus, this work proposes a method, called sKanner3D, that generates a geometric model of a person with texture and rigging through an automatic capture process using only an RGBD camera and then generates believable variants of the captured geometric model. Key points are identified to position the joints and perform the rigging process, and the texture mapping is generated by color information. A technique of mesh deformation by control points was used to generate variations in the geometric models, being directed by anthropometric information associated with obesity levels, applied differently to the sides (front, sides and back) of each segment of the body. In order to validate the generated model, the volume of the geometric models associated with each level of obesity was calculated and compared with the measurements collected from 7 to 10 year old children. To evaluate the generation of models, a questionnaire was also applied to professionals in the area of Body Perception, in addition to an objective analysis of the validation of the models generated. The results of these analyzes demonstrate that sKanner3D can generate credible models, despite some limitations of the capture step that can influence the deformation process. The results of these analyzes demonstrate that sKanner3D manages to generate credible models, despite some limitations of the capture step that can induce the deformation process. In addition, a difficulty in subjective assessment is noted because of the lack of consensus on the ideal body shape for each BMI group..
Download this paper in: PPGCA - CCT - UDESC
How to cite this paper:
@mastersthesis{gabrielcaixeta2018mastersthesis,
author = {Silva, G. C.},
title = {SKANNER3D: CAPTURA CORPORAL 3D COM MODELAGEM AUTOMÁTICA ENRIQUECIDA},
school={UNIVERSIDADE DO ESTADO DE SANTA CATARINA},
address={Joinville, SC},
year = {2018},
pages = {133},
}