Terrain Skeleton

Stacy Hoppen

A terrain skeleton is the network of ridges and valleys in a piece of topography--the points of highest and lowest elevations. The terrain skeleton is identified in order to perform further analysis, such as drainage patterns or hillshading.

Extraction of terrain skeleton

  • Manual Methods

    - Georeference a DEM and a satellite image so that each point has an elevation assigned to it.

    - Use two satellite images to create a 3D stereo model, and extract the elevations.

    There are semi-automated versions of both these methods as well.

  • Automated Methods

    A brief history of the algorithmic methods to identify streams and valleys (Band, 1984):

    - Peucker and Douglas (1975) identified concave pixels to find streams and convex pixels to find ridge points.

    - Mark (1983) uses an algorithm that starts at an upstream drainage area and identifies successively lower pixels. The advantage of this method is that it yields a set of drainage lines that are guaranteed to be connected. Peucker and Douglas's method sometimes yields broken segments.

    - Toriwaki and Fukumura (1978) use a verbal verbal classifiction scheme to assign a category to each pixel: PIT, PEAK, PASS, RAVINE, and RIDGE. The pixels are connected into a network, and then thinned to a set of lines.

    There are two limitations to algorithmic methods

    The terrain skeleton that is extracted can only be as good as the original data that was collected for the DEM. It is assumed that those doing data collection correctly identified the VIP (very important points) defining the ridges and valleys in the region. Also, algorithmic methods can be led astray by noise in the data: erroneous pits and peaks. One solution for this problem is to first smooth the terrain, but of course this leads to data loss.

    The parallel development of ridge extraction in other fields

    Algorithms which extract ridges, called "crests", from a digital image calculate the relative height of a pixel based on the intensity of gray. This is used to define the 3-dimensionality of a remotely sensed image such as a MRI (magnetic resonance image) in the medical world. The algorithm uses first, second, and third derivatives to extract the curvature and rate of change of curvature at each pixel to find points with the maximum values.

    These algorithms work for images in which only one variable is described by the surface because they depend on being able to assign a meaning, or elevation to every value of gray. A remotely sensed image of contains more information than these algorithms can process: vegetation and buildings in addition to elevation.

    References


    Return to the Terrain Analysis Home Page



    URL http://everest.hunter.cuny.edu/terrain/skeleton.html
    Last Update: March 11, 1996