Feasibility of Using Optical Power Spectrum Analysis Techniques for Automatic Feature Classification from High Resolution Thermal, Radar, and Panchromatic Imagery

摘要

The objective of this study was to determine experimentally the feasibility of using optical power spectrum analysis techniques for automatic topographic feature classification from high resolution radar, panchromatic and thermal imagery. A data base of radar, panchromatic and thermal imagery was assembled. Radar and panchromatic imagery were available over the same geographical area with the same scale and perspective. An optical power spectrum data base of 6,216 individual aperture samples from the three types of imagery was collected. Feature analysis and decision software required in addition to standard FACEL routines was developed. Using this software, features and a decison rule were developed for radar imagery that achieved 90% correct classification of four classes of terrain. A quantitative statistical analysis was performed to determine the effects of aperture and sensor type on the optical power spectrum based features. In addition, a qualitative analysis was performed in order to present examples that illustrate signature differences between radar and panchromatic imagery. No single sensor performs best for all classes. The results lead to the following choices for the detection of particular terrain types: Urban-panchromatic, Water - radar or panchromatic, Agriculture - radar, and Forest -thermal or radar.