Study Of3D Model Reconstruct Technology Based On Range Image Data

Abstract:The3Dimension modeling technology is a key branch area of computer vision science. It can not only digitize the object appearance feature and rebuild the virtual environment, but also can simulate the physical process. The Range image data based3Dimension model reconstruct technology is using kinect to capture range image data and transform it into point cloud data. this method is faster and cheaper than the traditional depth censor system(Lidar or Tof camera system).The Kinect camera uses structured light technology to capture the range image data and it is developed by the Microsoft company. In order to register the synchronized color image data and range image data, we calibrated these two cameras and calculated their intrinsic parameter matrix and the rigid movement parameters. Then we use the pixels with same color in color image to fix holes in range image data. And finally, we transform the range image data into point cloud data.The classical point cloud registration algorithm:Iterative Closet Point (ICP) method wastes time in matching closet point sets of two point clouds. We proposed a new registration method in this paper. It extracts three registration planes which represent the gesture of the point cloud data and uses the position information of former point cloud frame as the future knowledge to register current point cloud frame data. We test this algorithm by simultaneously using the ICP algorithm and our method to register several successive point cloud data frames and form a complete surface. The result shows that our method is more suitable for successive point cloud frame registration.We use the registered point cloud frames and Delaunay triangulation method to connect point cloud and use Point Cloud Library develop tool to finish the surface reconstruction.