Modeling of shadows in radar clutter
Abstract
Investigated was the applicability of the Forgy-Jancey and minimum spanning tree clustering algorithms to the problem of locating and characterizing shadowed regions in ground clutter. Certain preprocessing techniques were also investigated for reducing the amount of data prior to clustering. The two clustering algorithms were applied to both simulated and actual radar data, the latter provided by the Signal Processing Laboratory at RADC. The algorithms were run on the IBM 360/65 at Clarkson College using programs either written at Clarkson or obtained from the pattern recognition package ARTHUR. The Forgy-Jancey algorithm was found to be flexible and capable of handling large data sets at moderate cost, but was inconsistent in its ability to detect the 'natural' clusters in unknown data. The minimum spanning tree algorithm was more reliable in this regard, but the ARTHUR implementation was found to be costly in terms of required storage. The report contains numerous figures illustrating the application of the two clustering algorithms to various configurations of data. Clustering techniques appear to hold promise for examining the structure of unknown clutter data, but the results must be interpreted with care, since a natural clustering is not always found.
- Publication:
-
Final Technical Report
- Pub Date:
- July 1980
- Bibcode:
- 1980cct..rept.....B
- Keywords:
-
- Clutter;
- Computerized Simulation;
- Signal Processing;
- Algorithms;
- Fortran;
- Pattern Recognition;
- Subroutines;
- Communications and Radar