Bojan Kverh and Aleš Leonardis (2002) A new refinement method for registration of range images based on segmented data. Computing, 68 . pp. 81-96.
We present a new method for registration of range images, which is based on the results we obtain from the segmentation process. We need two range images segmented into regions, each of them modeled by a paramteric model and the approximation of the transformation between the two range images. Then two sets of corresponding points, one from each range image, are chosen and the transformation between them is computed to further refine the initial approximation of the transformation. The novelty is how we obtain the a corresponding points for the original set of points from the range image. Namely, to obtain them we project set of points from the first range image onto geometric parametric models that were recovered in the second range image and viceversa. This way we obtain two sets of corresponding points. Then we compute the transformation between the two sets. Few iterations are required to improve the initial approximation of the transformation. The results have shown a significant improvement in precision of the registration in comparison with traditional approaches.
|Item Type: ||Article|
|Keywords: ||registration, segmentation, range images, reverse engineering, CAD models|
|Language of Content: ||English|
|Institution: ||University of Ljubljana|
|Department: ||Faculty of Computer and Information Science|
|Divisions: ||Faculty of Computer and Information Science > Computer Vision Laboratory|
|Item ID: ||9|
|Date Deposited: ||28 Mar 2003|
|Last Modified: ||13 Aug 2011 00:31|
Actions (login required)