Comparison of two robust estimations by expectation maximization algorithms with Huber’s method and outlier tests
-
Karl- Rudolf Koch
Abstract
The robust estimation by the expectation maximization (EM) algorithm is derived for the variance-inflation model in addition to the known estimation for the mean-shift model. To compare these methods with the τ - test, Huber’s robust M-estimation and the multiple outlier test, a random linear model and laser scans for fitting a plane are generated by Monte Carlo methods. It turns out that the results for detecting outliers by the EM algorithms for the mean-shift and variance-inflation model approximately agree although the numbers of convergences are different. The results are superior to the ones of the methods with which they are compared. In case of the generated laser scans, the maximum number of outliers, which can be detected, is approximately identified.
© 2013 by Walter de Gruyter GmbH & Co.
Articles in the same Issue
- Masthead
- Velocity Field across the Carmel Fault Calculated by Extended Free Network Adjustment Constraints
- Models for assessing the spatial distribution of geodetic point patterns: application to geoid prediction quality
- Automatic optimization of height network configurations for detection of surface deformations
- Comparison of two robust estimations by expectation maximization algorithms with Huber’s method and outlier tests
- Estimation of Measurement Uncertainty of kinematic TLS Observation Process by means of Monte-Carlo Methods
- A comparison of particle swarm optimization (PSO) and genetic algorithm (GA) in second order design (SOD) of GPS networks
Articles in the same Issue
- Masthead
- Velocity Field across the Carmel Fault Calculated by Extended Free Network Adjustment Constraints
- Models for assessing the spatial distribution of geodetic point patterns: application to geoid prediction quality
- Automatic optimization of height network configurations for detection of surface deformations
- Comparison of two robust estimations by expectation maximization algorithms with Huber’s method and outlier tests
- Estimation of Measurement Uncertainty of kinematic TLS Observation Process by means of Monte-Carlo Methods
- A comparison of particle swarm optimization (PSO) and genetic algorithm (GA) in second order design (SOD) of GPS networks
