Presentations

We propose the method of estimating the 3D shape of trunk of the body by fitting the silhouettes of homologous human body model to input silhouette images. We have many homologous models of varying ages and analyze them by Princi- pal Component Analysis. Then, using the proposal system which deform the average model of each age to various body shape by changing principal component values, we deform the model’s silhouettes to input silhouette images, so that we estimate the 3D body shape of the person who has such input silhouettes. When the model’s silhouettes are well fitted to the input silhouettes, the model’s 3D shape have to correspond approximately to the body shape of the person who is the owner of input silhouettes. Therefore, we have to optimize the principal component values to fit model’s silhouettes to input silhouettes, so we calculate optimal values by using the linear independency of principal components. In this paper, we analyze 52 models of 50’s women’s trunks of the bodies by Principal Component Analysis, and generate an average model to be initial model to deform. We make an experiment using a couple of silhouette images generated from one of those 52 models as input silhouette images. The experimental result shows that we can approximately reconstruct the model’s 3D shape from only their silhouette images.

We have a lot of products designed for use by hand, like a mobile phone, mouse, and camera, etc. These products can be used by many kinds of people whose hands come in various sizes. Designing a handheld product, therefore, it is required to consider the individual difference of user’s hands. So there are the high expectations for simulating the interaction between user’s hands in various sizes and the product in order to evaluate the product design without prototyping. To achieve it, the technology that generates various Digital Hands automatically from data of hand measurements and finds how a user grasps the product is needed. In this paper, we propose the method that measures 23 lengths and 9 widths included in all fingers automatically from a single image which is got by scanning palm side of a hand using any paper scanner. And we use these resultant measurements for estimating 9 thicknesses and 5 boundary lengths included in fingers or a palm that are perpendicular to the scanned surface. Finally, we try to deform the resultant Digital Hand model into a grasping posture.