LOCATING RANDOM ERRORS IN DIGITAL TERRAIN MODELS
Carlos López - Environmental and Natural Resources Information Systems
Royal Institute of Technology,
The increasing use of Digital Terrain Models (DTM) in Geographic Information Systems (GIS) application has put strong interest in the assessment of their quality. The present work of locating randomly distributed errors is part of an ongoing project which also involves the sensitivity analysis of the results of calculations, with respect to the noise in a DTM.
A DTM of Stockholm, with heights varying from 0 to 59 m., was seeded with randomly distributed errors within 3mt and +3mt, and afterwards they were located with a newly devised algorithm, based on Principal Component Analysis, which is described in this paper.
The preliminary results show that roughly half of the larger errors have been located with a probability over 90%, in a single step of the method. With one additional step, also half of the +-2mt ones have been located with a probability over 80%.
The possibilities of using this procedure for error detection during the DTM generation and prior to usage in different applications are considered.
Published in:International Journal of Geographic Information Science V 11, N 7, 677-698 (1997)
Here you have a draft version(191KB)