University of Twente Student Theses


Assessment of the performance of CosiCorr software in the detection of horizontal displacements using artificial images

Ramos, Diego Erick Callejas (2012) Assessment of the performance of CosiCorr software in the detection of horizontal displacements using artificial images.

[img] PDF
Abstract:The surface displacements at regional scale (e.g. earthquakes) or small scale (e.g. landslides) are always subject of analysis in pursuit of its detection and monitoring. Among some methodologies for horizontal displacement detection, the CosiCorr software is an important recent implementation that automatized the process inherent to the correlation of a pair of images. Although the correlation of optical images by using CosiCorr software was extensively developed and applied in the recent years, it is still present many doubtful situations or inaccuracies in the results that make necessary the discarding and reconsideration of the horizontal displacements detected by the software by using filtration processes. For this purpose, the present study aims the assessment of the horizontal displacements detected by the software, performing a sensitivity analysis of their more important input parameters: the window size and step size. In order to count with predictable results several artificial images were artificially created, with different spatial resolution, and representing different amount of sub-pixel (less than one pixel) and multi-pixel (more than one pixel) displacements. The use of this artificial input imagery have not any precedent in a study, thus is considered at last as a useful input data for similar future studies. Using such imagery, it was determined that the resolution of the images used doesn’t affect the correlation results; moreover the correlation of different resolution images gives more noisy results. Regarding the software parameters, the step size showed no influence in the results; on the contrary, the window size had a remarkable influence in the results of a correlation. It was found that a displacement, by using artificial images, is accurately detected when the window size that was used is three times higher than the displacement (in pixels) to be detected, considering multi-pixel displacements. Similar relation for sub-pixel displacements was not possible to establish because there were many questionable results obtained.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page