3d barrier-free verification for wheelchair access

This paper proposes a new methodology for identifying barriers encountered by wheelchair users in daily life spaces. Currently, barrier-free designs are required not only for newly constructed buildings, but also when renovating existing facilities and public spaces. However, the arrangement of furniture, equipment, and many other objects in a space often impose barriers, and even the simple bumps and steps on pathways can obstruct wheelchair passage. Furthermore, it is often difficult for administrators to envisage the full reality of barriers in their facilities because potential obstacles can be created inadvertently by a variety of objects that have complicated three-dimensional (3D) geometries. In such cases, their existence will normally remain unknown until someone actually tries to transit the area using a wheelchair. Our approach aims to capture the overall dimensions of target spaces by collecting and combining depth images taken using a hand-held RGB-D camera (also commonly referred to as a ranging camera), and then to navigate a virtual wheelchair through the target space in a computer simulation to check for obstacles. The practical egomotion capabilities of RGB-D camera sensors within actual environments make it possible to achieve real-time simultaneous localization and mapping (SLAM) functionality, which is necessary for creating accurate 3D location maps. The Microsoft Kinect™ sensor, which was originally designed as a user interface for home-use video games, is a good example for a low-cost, compact RGB-D camera. Since the Kinect device is sufficiently compact for use when capturing arbitrary objects in situ, we adopted it for use in our study and applied a SLAM technique to perform barrier checks. Our simulation employs 3D projections of all objects and wheelchair transit volumes onto a floor plane in order to detect potential obstacles. We implemented our proposed method on a laptop personal computer (PC) and collected data from actual classroom and common space locations in a university. The experimental results of our method showed effective functionality in terms of practicality and usability.