| Effective reuse of existing 3D-model resources is an important issue in computer graphics. We regard the common indoor man-made objects as the entry point of our research. This paper makes systematic research on the two problems of example-based model synthesis:single object modeling and indoor scenes modeling involved. And then, develop the prototype system named Indoor Objects Building System based on the previous researches. Generally, the main contributions of this paper are:(1) Modeling of Single Objects from RGBD Image. We have presented a fast modeling method based on the highly noisy and incomplete scanned data by Kinect in a single view. We take full advantage of the abundant models of the same semantic class on the Internet to build a heuristic model based on the 3D point clouds and a corresponding RGB image. Our method includes three major phases:Firstly, we analyze the structure of the 3D point clouds and the RGB image based on the semantic segmentation of the candidate model to label the scanned target’s components; secondly, we search for candidate parts for the labeled parts in the target to get matched component information; lastly, we assemble the candidate parts to create the final 3D model. We demonstrate the efficiency and speed of our method in building models based on incomplete point clouds data from Kinect.(2) Modeling of Indoor Scenes from RGBD Image. We present a fast modeling method of the indoor scenes. We obtain the indoor scenes RGBD scanning data by using the sensing device Kinect in a single view, and then take full advantage of the abundant models of indoor objects on the Internet to achieve the scenes modeling. Our method includes three major phases:Firstly, we implement the segmentation of the RGBD image in order to get the 3D point clouds and the RGB image of every target objects in the scene. Secondly, we search for proper models from model base for each target object. We use the 3D depths information and the 2D shape information of objects from RGBD data, and combine them with the relevance of indoor objects. Lastly, we assemble the models to create the final 3D modeling scene. We demonstrate the efficiency and speed of our method by experiment on several indoor scenes scanning data.(3) Indoor Objects Building System. We designed and implemented a prototype system which integrates the single objects modeling and indoor scenes modeling from our research above. This system can utilize the captured RGBD data of Kinect devices to achieve quick indoor objects and scenes modeling, and it can provide appropriate construction mold experience for professional or non-professionals persons. |
