Hybrid Neural Model for Target Recognition.

摘要

The invariance principle is one of the important design consideration in target recognition. Some theoretical aspects of this principle were investigated. A new set of affine invariant features were developed. Geometrical examples are given, and features generated using this feature extraction technique are demonstrated. A novel artificial neural network model was developed to analyze these features and perform classification of the targets. This network acts as a dynamic model to establish classes of targets in a nonlinear fashion. Recognition is based on the combination of a unique set of features and the newly developed neural network model. This target recognition approach demonstrates that recognition can be obtained despite target orientation, size, or aspect angle.