A Novel Error Detection Technique for Automatically Generated Digital Elevation Models

The generation of Digital Elevation Models (DEMs) is one photogrammetric process that has been successfully automated. However, a common criticism surrounding DEM generation software currently available commercially, is the lack of quality assessment procedures. This paper describes a new system called the Failure Warning Model (FWM), which automatically detects low accuracy areas and presents the information in a graphical format that is easy to assimilate. The technique makes use of the strategy parameters embedded within many current digital photogrammetric systems. Typically, these are a set of user definable parameters that control the search, acceptance and quality control procedures used to derive final point elevations in the DEM. Detailed tests revealed that accurate points are insensitive to changes to the strategy parameters and so the system identifies points that are sensitive to parameter change, which may therefore be considered unreliable. As a further refinement, the algorithm considers the impact of local slope angles and the internal classification of the height estimate. The Failure Warning Model was developed within the ERDAS Imagine environment for the ERDAS OrthoMAX DEM generation software. It was tested on a wide variety of imagery (scale: 1:70 to 1:45,000) representing a variety of landcover types and show that the approach is successful. More significantly, the method proved successful using other digital photogrammetric systems, indicating the generality of the approach. The Failure Warning Model is described in this paper and combined with results from extensive tests using both the OrthoMAX and Phodis TS software from Zeiss.