[Japanese | Thesis | Researches in Minoh Lab | Minoh Lab]
With increasing variety of computer use, 3D space often comes to be constructed in computers. In order to describe 3D space, 3D shape models are necessary. Especially, 3D shape models of human-body has various application in apparel and medical information systems.
To get a 3D shape model of a human body, we must measure the 3D human body shape. Many techniques, including stereo scopic viewing and laser range finding systems, are proposed for the measurement of the shapes of an object in the world. Generally, 3D data of the human body is obtained by using laser range finding systems. These techniques are used for getting 3D data of an object with various shapes.
For modeling the 3D shape of a human torso from measured data, there have been proposed to fit B-spline surface to them, or to make free-formed surface model using B\'ezier patches from them.
But, the measured 3D data of human body, include defects or errors of data to be the ones of human body shape, due to the limit of the measurement principles. These defects and errors in the data make a 3D shape model of human body inappropriate. These defects and errors in the data have to be detected and corrected in order to get a 3D model of real human bodies.
In this report, we have constraints that the human body is in its an upright position, and that the 3D shape of the human body is represented by the cross sections of the human body (We call each cross setion a ``slice''). Based on these constraints, we propose a method to correct the defects and errors in the data.
Although it is possible to use a standard shape model of human body to correct the data, we make use of the constraint that the human body is in its upright position. In other words, human body in upright position can be regarded as bilaterally symmetrical, and the paired slice on the same cross section can be regarded as similar. We make use of the constraint of the symmetry of a human body to correct the defects and errors.
Since the surface of a real human body is smooth, that of the 3D shape model that is obtained after correcting the data should also be smooth. Thus, 4th B-spline surface is fitted on the corrected human body data.
To verify the effect of the proposed method, the result of fitting 4th uniform B-spline surface on the human body data corrected by the proposed method is compared with that of fitting on the data without the correction. As the result, we can get better 3D shape of human body from the fitting to the corrected data than 3D model fitting to the raw data.