Authors: Chin-Hung Teng and Chen-Yuan Hsieh, and Kai-Yuan Chuo
How to efficiently and effectively reconstruct 3D models from real objects is an important research topic in computer vision. In this research, we employ a low-cost RGB-D camera such as Kinect sensor for object 3D model reconstruction. A Kinect sensor can acquire accurate 3D information of the scene, but it cannot distinguish the target object from the background. Thus, we also incorporate an interactive 3D segmentation algorithm in our system to effectively separate an object from the background. To improve the interactivity of proposed system, an Arduino with BlueTooth XBee module is also employed in this system to replace mouse operations. Hence, by this system a user can freely move the Kinect sensor to acquire the required 3D information and meanwhile distinguish the foreground from background. The 3D point cloud of the object can be extracted with the background being effectively removed. Finally, a complete polygonal mesh model for the object can be created from the extracted point cloud.
Keywords: 3D model, 3D reconstruction, Kinect, RGB-D sensor, Lazy snapping, KinectFusion