A Marker-Free Calibration and Registration Process for Multiple Depth Maps from Structured Light Sensors and its Application in Video Avatar Systems

摘要

Virtual and augmented reality systems for education, health and entertainment purposes may require the modeling of realistic avatars. In order to create a more complete 3D geometric representation of the user, it is often necessary a complex calibration process using markers for registering streams from multiple 3D cameras. This paper presents a proposal of a marker-free registration process for a global model creation that represents the human body surface using multiple structured light sensors. Besides that, this method allows greater scalability by the addition of new devices to the setup. The paper also covers two experiments to check the feasibility of this setup, the first one related to noise measurement using two infects simultaneously and the second one, an evaluation of ICP algorithm behavior for multiple depth maps registration.