A Local Filter for Improve Digital Elevation Model
Subject Areas : Geospatial systems development
Mohammad Amin Ghannadi
1
,
Matin Shahri
2
1 - Surveying Engineering Department, Faculty of Geoscience Engineering, Arak University of Technology
2 - Surveying Engineering Department, Faculty of Geoscience Engineering, Arak University of Technology
Keywords: Blunder removal, Digital Elevation Model, DEM refinement, adaptive filter,
Abstract :
An accurate, high-quality Digital Elevation Model (DEM) from the ground is essential for many applications. Due to some data collection problems as well as weaknesses in DEM production techniques, including interpolation methods, these models are associated with some blunder errors that must be edited manually or automatically. In this study, a method for removing noise and blunders as well as improving the DEM is proposed. In this two-step method, first, with the standard deviation of the alpha-trimmed filter, points with blunders are identified and removed. Then, an adaptive inverse distance weighted filter is applied to remove blunders and refine the DEM. The proposed method and some common competitive methods have been applied and evaluated on a simulated DEM and a DEM extracted from satellite stereo images from the southwest of Mashhad city. The root means square error of the DEM extracted from satellite images using the proposed method is 1.89 meters, while this criteria for the inverse distance weighted filter method is 2.43 meters. The experimental results show that the proposed method, despite a 20% increase in time cost, can improve the accuracy of the modified DEM by at least 22% compared to the weighting method using the inverse distance. Therefore, the filter proposed in this study can be used to remove noise and improve DEM when increasing the accuracy is a priority.
