Automatic DEM Extraction from Stereomodels

A digital elevation model (DEM) is a digital representation of terrain heights. The most common forms are a regular grid (usually saved in raster format) or a triangular irregular network (TIN) of triangle facets (vector format). They can be created by manual collection (see preceding section) of height points, breaklines, and contours and subsequent application of interpolation algorithms or Delaunay triangulation, which convert these data to regular grids or TINs (Li, Zhu and Gold, 2005). With the advent of digital photogrammetry, however, it has become possible to extract elevation information automatically from stereomodels using stereo-correlation or image-matching techniques.

A human operator manually mapping topographic features in stereomodels basically accomplishes two tasks: placing the floating mark onto an object and point by merging the left and right image of the object into a stereoscopic view, and interpreting the nature of the object, classifying it, for example as a tree, a street, a house, or a height point. The first task only involves comparing two image subsets for their similarity and also can be performed automatically by computer processing. The second task involves recognizing, distinguishing, and interpreting individual objects, at which human beings are still much superior to computers, although pattern recognition and feature extraction techniques continue to be important research topics in all imaging sciences, including photo-grammetry and remote sensing (e.g. Bartels and Wei, 2007; Grim, 2008).

Based on the two main categories of image-matching techniques (area-based matching and feature-based matching), various hybrid methods have been developed (Wolf and Dewitt, 2000). All are concerned with statistically determining the correspondence between two or multiple image areas or features in order to identify homologous image points. A variety of parameters constrains and controls the process of searching and correlating image points. The quality and amount of height points may be affected both by the ground-cover type and by image characteristics such as noise and local contrast. Dark and shadowed regions often show low correlation coefficients owing to increased noise, and homogeneous smooth surfaces may have too little texture for successful image matching.

Figure 3-15 illustrates the results of an investigation into the role of different image types and ground-cover classes. Both analog and digital images of the same scene (taken simultaneously with a 35-mm SLR and 6.3 megapixel DSLR on a double-camera mount) were used for DEM extraction. The analog transparency slides were scanned with 2900 dpi and subsequently resampled to 1800 dpi, all images were processed with Leica Photogrammetry Suite. Results show that the digital camera is clearly superior to the film camera in terms of point density; not only are the

□ sparse low vegetation, bare soil

Digital Camera and Digital Photography

Digital Camera and Digital Photography

Compared to film cameras, digital cameras are easy to use, fun and extremely versatile. Every day there’s more features being designed. Whether you have the cheapest model or a high end model, digital cameras can do an endless number of things. Let’s look at how to get the most out of your digital camera.

Get My Free Ebook


Post a comment