[Japanese | Thesis | Researches in Minoh Lab | Minoh Lab]
Measuring an object's shape with multi-cameras and describing the shape are useful in constructing digital museums. Description of the 3D shape should be determined by considering both required details and a resolution which cameras can provide.
The required details, which we call required resolution, differ part by part all over the object and they also differ according to purposes. We suppose they can be specified by users.
In this research, we aim at a realization of shape description which can satisfy the required resolution. \par In the existing measurements and shape descriptions, cameras are often arranged equally surrounding the object and voxel-description is widely used. However, the required resolution specified by users cannot be satisfied under equally arranged cameras and voxel-description.
Our ideal cell will solve this problem. An ideal cell is a kind of 3D region that shows the resolution provided by the cameras. We use ideal cells to describe shapes. By this shape description, we can estimate the resolution part by part with a given camera arrangement. We use a real numerical value genetic algorithm to search for a camera arrangement that satisfies the required resolution.
We conducted three experiments with three different insect-models to evaluate our method. In all three experiments, we verified that the required resolutions cannot be satisfied under equally arranged cameras and voxel-description. Moreover, we succeeded to get the camera arrangements that satisfied the required resolution in all three experiments.