A Robust Method for Registration of Partially-Overlapped Range Images Using Genetic Algorithms
Registration is a fundamental stage in the 3–D reconstruction process. We consider the problem of Euclidean alignment of two arbitrarily-oriented, partially-overlapped surfaces represented by measured point sets contaminated by noise and outliers. Given two approximately aligned range images of a real object, it is possible to carry out the registration of those images using numerous algorithms such as ICP. Basically the task is to match two or more images taken at different times, from different sensors, or from different viewpoints. In this paper, we discuss a number of possible approaches to the registration problem and propose a new method based on the manual pre-alignment of the range images of arbitrarily-oriented surfaces followed by an automatic registration process using a novel genetic optimization algorithm in 3–D data registration. Results for real range data are presented with precision and robustness, combined with the generality of genetic algorithms. This procedure focuses on the problem of obtaining the best correspondence between points through a robust search method between partially overlapped images.
KeywordsGenetic Algorithm Differential Evolution Range Image Registration Method Registration Process
Unable to display preview. Download preview PDF.
- A. Myers, “Introductory literature review surface reconstruction from three dimensional range data”. Technical report, The University of Adelaide, Department of Computer Science, 1999.Google Scholar
- Y. Chen, Object modeling by registration of multiple range images. Image and Vision Computing, 10, 1992.Google Scholar
- K. Brunnstrom, Genetic algorithms for freeform surface matching. Technical report, 1996.Google Scholar
- C. Robertson and R. Fisher, Parallel evolutionary registration of range data. Computer Vision and Image Understanding, pages 39-50, 2002.Google Scholar
- M. Salomon, G. Perrin and F. Heitz, Differential evolution for medical image registration. pages 201-207, 2001.Google Scholar
- S. Rusinkiewiczs, Real-time acquisition and rendering of large 3–D models. PhD thesis, Stanford University, 2001.Google Scholar
- B. Horn, Closed-form solution of orientation using unit quarternions. Journal of Optical Society of America, 4, 1987.Google Scholar