The 3DHAP project at the Robotics and Mechatronics group of the University of Twente aims at the development of a handheld device for scanning skin and scoring the skin illness like psoriasis. The project is a pilot study in which the feasibility of such a device is investigated. The outcome of this study will be used as input for a grant proposal.
A possibility of the handheld device is to use visual SLAM to enable imaging larger areas of the skin. A 3D visual SLAM algorithm uses images from a camera to detect key points in the image plane, which ultimately are used to estimate 3D positions of landmarks. These landmarks are used to reconstruct a 3D surface model of the skin. In order to do so, the position and orientation of the camera during scanning must also be estimated. In the so called EKF-SLAM, the uncertainty of estimates of these landmarks and camera pose become available in terms of covariance matrices.
As the scanner is aimed to be used in a health care application, in which accurate estimation of the geometry is of vital importance, it is crucial that the given uncertainties are correct. The goal of the project is to build an experimental setup for the validation of the uncertainties as given by the EKF-SLAM.