KINECT-BASED REAL-TIME RGB-D IMAGE FUSION METHOD

Abstract. 3D reconstruction of indoor environments based on vision has been developed vigorously. However, the algorithm's complexity and requirements of professional knowledge make it restricted in practical application. With the proposition of the concept of Volunteered Geographic Information (VGI), the traditional method is no longer suitable for VGI. So in this work we utilize consumer depth cameras – Kinect to enable non-expert users to reconstruct 3D model of indoor environment with RGB-D data. Considering the possibility of camera tracking failure we propose a method to perform automatic relocalization.