Project description: In the three dimensional (3D) data processing, registration is the process aligning multiple 3D data sets in a common coordinate system. Previous registration methods rely on accurate mechanical positioning devices, or on manual processing to estimate the viewpoints. In addition, most algorithms require many pre-processes: feature extraction, matching, and surface segmentation. This research focuses an iterative method for automatically registering multiple 3D data sets by using covariance matrix without a prior knowledge about 3D transformation between views. To achieve accurate registration, our method uses both the 3D transformations giving a relative pose between the 3D data sets, and the projective matrix representing projection of 3D space to 2D image. By minimizing the difference of two covariance matrixes of the overlapping regions in two 3D data sets, we can make a precise registration of multiple 3D sets with no complex procedures that are prone to errors and any mechanical positioning device or manual assistance.
 

Publications:

 

2004 Jung-Kak Seo, Hyun-Ki Hong, Cheung-Woon Jho, Min-Hyung, Choi, “Two Quantitative Measures of Inlier Distributions for Precise Fundamental Matrix Estimation”, Pattern Recognition Letter, Vol. 25, Issue 6, 2004, P733-741 PDF BiBTeX
 

Sang-Hoon Kim, Jung-Kak Seo, Hyun-Ki Hong, Min-Hyung Choi, “Iterative Registration of Multiple 3D Data Sets Using Covariance Matrix” Proceedings of International Conference on Virtual Systems and MultiMedia, Sept. 2002  PDF BiBTex