Abstract

Three validation techniques for performing positional accuracy assessments in georeferencing were critically reviewed in this study, with an aim of estimating the possibilities of identifying and removing the geometric deformations of raster images. A validation technique based on the residuals of tie points, Hold-Out and the Cross-Validation technique was implemented within PlanTra software. Accuracy indicators, obtained through these techniques, using different functional transformation models, were thoroughly tested on a typical area. It has been proven that the Cross-Validation technique produces a much more reliable estimation of accuracy in comparison to the standard approach based on the residuals of tie points.

Key words: positional accuracy assessment, georeferencing, Cross-Validation, RMSE