|
|
|
Imaging, Robotics, and Intelligent Systems Laboratory The University of Tennessee |
|
|
|
All methods, expectedly, have difficulties with highly symmetric objects or parts of objects, and most methods depend on expensive equipments and require much time in generating 3D models. Furthermore, the registration results are unreliable when the depth data have large error bound. Establishing point correspondences between different 3D views of the same object is a well known problem, instrumental to various applications, among which 3D registration, object recognition and indexing of 3D models. Automatic pair-wise 3D registration procedures are two steps procedures with the first step performing an automatic crude estimate of the rigid motion parameters and the second step is refining them by the ICP algorithm or some of its variations. The first task has been studied in many researches and still is an open area to develop new algorithms. The goal of this project is to reconstruct the surfaces of small mechanical objects and terrains using primitive relationships and functions of 3D points acquired with laser scanner and to analyze 3D images. Current range cameras typically deliver range data associated with some kind of photometric information, either in terms of gray-level intensity or of RGB data. This makes it natural to devise 3D point matching techniques capable of exploiting texture and shape information. The main range data acquisition tool that is used currently is Genex 3D FaceCam. This camera uses three high resolution CCD sensors and an encoded pattern projection system. The RGB color information from each pixel is used to compute the range data and generate an accurate 3D surface map. An automatic registration toll Rapidform04 based on ICP algorithm is used to register view pairs, and to reconstruct full object. Also another framework for registration process ,which is based an tensor voting local descriptor extraction and a matching criterion Gaussian force fields, is currently studied. Results: Currently in process No publications currently available for this project. This research is being conducted at the IRIS Lab by Muharrem Mercimek under the supervision of Dr. Mongi A. Abidi. |