EVALUATION OF NEURAL NETWORK PERFORMANCE IN TARGET CLASSIFICATION APPLICATIONS

PDF

Authors
  1. Thorp, W.E.
Corporate Authors
Thorp (W E) Associates Ltd, Ottawa ONT (CAN);Defence Research Establishment Ottawa, Ottawa ONT (CAN)
Abstract
There has been much interest shown by the academic, industrial and military communities in the potential of neural networks to perform various tasks. Of special interest is the task of pattern recognition since neural networks appear to be able to classify images with great speed and accuracy. Within this report various neural networks were tested to determine their ability to recognize "toy" simulations of synthetic-aperture radar ship images. These images were used to train the neural networks. In order to test the classification performance of the neural networks, these images were then corrupted with varying amounts of noise as well as being translated and scaled. The results have been tabulated and analyzed within this report. The neural networks tested include: Adaline; Hamming; Bidirectional Associated Memory; Recirculation; and Back Propagation. TRUNCATED
Date of publication
15 Dec 1989
Number of Pages
266
DSTKIM No
90-01825
CANDIS No
64092
Format(s):
Hardcopy;Originator's fiche received by DSIS;Document Image stored on Optical Disk

Permanent link

Document 1 of 1

Date modified: