Abstract
In recent years, efforts to construct three-dimensional (3D) space data have been drawing attention. Representative cases include studies using MMS or a laser range sensor. In these studies, however, their availability is limited in some areas such as the inside of buildings or underground shopping malls where GPS accuracy is low. Methods without using GPS include a proposed approach called SLAM to construct 3D space data by estimating relative positions from a characteristic object in the real space and calculating the moving path. However SLAM, which uses learning data, has a problem in that the precision depends on learning data. Then, this research proposes a method for estimating the relative positions between a range image sensor and objects, which does not need GPS or learning data. This makes it possible to construct three-dimensional space without depending on the shooting environment.
