Digital Processing and Analysis of Active Multispectral Data.
Final technical rept. 15 Jun 75-4 Feb 76,
MCDONNELL AIRCRAFT CO ST LOUIS MO
Pagination or Media Count:
Seven runs of active multispectral data from a scanning laser sensor were digitized and analyzed to examine the classification capability of the difference figure of merit DFOM algorithm and to examine the effects of sensor bandwidth and scan speed, atmospheric transmission and polarization on the classification rate. The data represented the magnitudes of the scattering cross sections from a calibrated resolution target, comprising panels of many colors, sizes, and orientations. Co-polarized and cross polarized components of the back scattered radiation were measured at four frequencies in the visible spectrum. A multispectral signature vector was derived for each sample of the digitized data. The DFOM analysis compares the direction of the unknown multispectral signature vector with standard signature vectors measured in the laboratory. Classification occurs when a directional match is found. Using a priori knowledge of target configurations, the probability of correct classification was found to be high for most target materials. Discrimination capability decreases for similar colors and small scattering cross sections. Increasing bandwidth allows greater distinction between similar signatures but does not improve discrimination between dissimilar signatures. Co-polarized data could not be classified with the DFOM technique. Results of the scan speed and atmospheric transmission studies were inconclusive. Author
- Optical Detection and Detectors