Robust Multiresolution Integrated Target Sensing and Recognition.
Final rept. Oct 93-Sep 96,
MINNESOTA UNIV MINNEAPOLIS DEPT OF ELECTRICAL ENGINEERING
Pagination or Media Count:
The main goal of this research effort was to develop an integrated target sensing and recognition strategy. Secondary goals of this work were to construct novel image representation and analysis algorithms to facilitate content based image retrieval. We derived a new adaptive waveform selection algorithm for radar range-Doppler target recognition. The algorithm selects the waveforms that provide maximum discrimination information at any given time by maximizing the Kullback-Leibler information number corresponding to the most likely hypothesis. As a result, it minimizes decision time for a given desired classification performance level and maximizes classification performance for a fixed data acquisition time. We have developed a novel theory of multiresolution representation of binary data that supports fast binary image matching algorithms. Only Boolean operations are needed to compute these representations. Finally, we have constructed a novel image coding technique that supports pictorial queries. The procedure minimizes a weighted sum of the expected compressed image file size in bits and the expected number of bits that need to be read to answer a pictorial query query response time.
- Target Direction, Range and Position Finding
- Active and Passive Radar Detection and Equipment