Invariant image encoding in neural network for target recognition
暂无分享,去创建一个
We present a neural network approach for invariant multi-target recognition that permits an effective training with no local minimums. A simple rule to design and to optimize the number of the hidden neurons of a multilayer neural network is proposed. As a case study, an image based invariant classification of three kinds of airplane is reported.
[1] Jacek M. Zurada,et al. Introduction to artificial neural systems , 1992 .
[2] G G Mu,et al. Real-time pattern recognition system based on a bipolar winner-take-all model with a threshold. , 1994, Optics letters.
[3] Yong Zhang,et al. Adaptive clipped model and its optical implementation , 1998 .