Wolfram Library Archive

All Collections Articles Books Conference Proceedings
Courseware Demos MathSource Technical Notes
Title Downloads

Algebraic robust estimation employing Expectation Maximization

Béla Paláncz
Organization: Budapest University of Technology and Economics
Department: Photogrammetry and Geoinformatics
Revision date


In this presentation a new robust technique employing expectation maximization to separate iteratively, outliers (corrupted data points) from inliers (true data points) represented by different Gaussian distributions is introduced. Once the parameters of these two Gaussian distributions are computed, the joint likelihood function of these two distributions can be maximized by algebraic method, employing numerical Groebner basis. Illustrative computations for plane fitting with synthetic test data as well as with data of real laser scanning experiment are presented. The results are compared with that of standard robust estimation techniques as The Danish and RANSAC methods

*Applied Mathematics > Numerical Methods

Robust parameter estimation, expectation maximization, maximum likelihood method, numerical Groebner basis

Algebraic_Robust_with_EM_BPalancz.zip (7.2 MB) - ZIP archive
output_41_392.dat (872.9 KB) - Unknown MIME type

Translate this page: