Identification of Highlighted Cells in Low-Variance Raster Data Application to Digital Elevation Models

Abstract

This study describes a new algorithm developed to detect local cells of minimum or maximum heights in grid Digital Elevation Models (DEMs). DEMs have a low variance in digital levels due to the spatial continuity of the data. Traditional algorithms, such as SIFT, are based on statistical variance, which present issues to determine these highlighted cells. However, one of the main purposes of this identification is the use of these points (cells) to assess the positional accuracy of these products by comparing those extracted from the DEM with those obtained from a more accurate source. In this sense, we developed an algorithm based on a moveable window composed of variable sizes, which is displaced along the image to characterize each set of cells. The determination of highlighted cells is based on the absolute differences of digital levels in the same DEM and compared to those obtained from other DEMs. The application has been carried out using a great number of data, considering four zones, two spatial resolutions, and different definitions of height surfaces. The results have demonstrated the feasibility of the algorithm for the identification of these cells. Thus, this approach expects an improvement in traditional procedures. The algorithm can be used to contrast DEMs obtained from different sources or DEMs from the same source that have been affected by generalization procedures.