Back to search

South Dakota State University

A Comparison of Prediction Error Estimators as Applied to Positional Accuracy of Geometrically Corrected Satellite Images via Monte Carlo Simulations

Abstract

dc:description.abstract

A Monte Carlo simulation was used to compare the performance of ten estimators of prediction error as they relate to the positional accuracy of geometrically corrected satellite images. The estimators were evaluated in terms of bias and variability. The simulation design was based on the Landsat TM satellite/sensor system under ideal conditions. Prediction error was defined to be the expected squared error between the geo-corrected (predicted) location of a given point and its known location in a geographically referenced coordinate system. The simulation was designed so that specific values of prediction error could be computed. These values served as the standard against which the ten estimators were compared. The square of RMSE was examined for both the control points (MSE) and an independent set of validation points (VMSE). Three estimators derived from MSE were examined (AASE, CP, BIC}, along with two cross validation estimators (CV, CV2) and three bootstrap estimators (BT, RBT, CST). All estimators, except VMSE, depend only on the original control points. The square of RMSE (MSE), computed for the control points, appeared to be the worst performing estimator in terms of bias. VMSE, the independent estimator, performed the best overall. However, the cost of choosing validation points may be prohibitive to its universal application. The CP estimator derived from MSE, the leave-one-out cross validation estimator (CV), and the most complex bootstrap estimator (CBT) performed quite well overall. However, the bootstrap estimator was extremely variable when used with only 1 O control points. For this reason, and because of its relative complexity, the CBT estimator appears to be slightly inferior to the CP and CV estimators.

Degree

thesis:*
Name thesis:degree_name
Master of Science (MS)
Level thesis:degree_level
Thesis - University Access Only
Discipline thesis:degree_discipline
Mathematics and Statistics

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • VanderZee, Darrel
Contributors dc:contributor
  • Robert Lacher

Subjects

dc:subject × 2

Rights

dc:rights
Language dc:language
English

Identifiers

dc:identifier.*
Repository record dc:identifier
https://openprairie.sdstate.edu/etd2/161

Chain of custody

source
Harvested from
South Dakota State University
Base URL
openprairie.sdstate.edu/do/oai/
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

VanderZee, Darrel. A Comparison of Prediction Error Estimators as Applied to Positional Accuracy of Geometrically Corrected Satellite Images via Monte Carlo Simulations. Thesis - University Access Only thesis, https://openprairie.sdstate.edu/etd2/161