|
|
Intelligent Retrieval of Topographic Features from Digital Terrain Elevation Data Collaborators : University of Maine, Northop Grumman, and
Rome Laboratory Contact: Mohamad Musavi, Padma Natarajan Proposal Summary Applications like autonomous navigation in natural terrain and the automation of map making process require high level scene descriptions as well as geometrical representation of natural terrain environments. Digital terrain elevation data (DTED) or Digital Elevation Model (DEM) represent the geographical information in the form of elevation values. Ridges are among the most useful features to extract from a DTED/DEM. Three algorithms have been presented for reliable extraction of terrain features called ridges from DTED and DEM. The first algorithm is based on fuzzy systems theory. Fuzzy systems make it possible to transform linguistic processing data and inference can be applied. High classification accuracy can be achieved, as the partial membership functions of pixel allows classes of mixed pixels to be identified more accurately. The second algorithm uses Kohonen's self organizing feature map for segmenting the image. Kohonen's SOFM is an unsupervised learning network subject to a topological (neighborhood preserving) constraint. The clustering produced by the SOFM reduces the input space into representative features using a self-organizing process. The third algorithm uses Fuzzy C-Means and thresholding, for segmenting the image. Fuzzy C-Means clustering approach does an optimal partitioning of the feature space. The methodology uses the coarse fine segmentation to reduce the computational burden required for the FCM. |
|
[Overview]
- [Goals] - [Faculty] - [Students] - [Industrial
Partners] |