Invariant Target Recognition Using Self-Organizing Networks

摘要

This paper presents an overview of neural network technology and specifics of theself-organizing networks. An artificial neural network is described which recognizes objects regardless of their spatial orientation. This network is invariant to rotation, scale, and translation. Invariance is built into the network by introducing a unique set of features which were developed in-house. This overcame the two shortcomings of long training times and combinatorial explosion of terms often present in other networks. Preliminary results suggest that with further refinement and enhancement, the system described in this report will have the capability to reliably recognize targets under adverse environmental conditions.