Study Artificial Neural Networks Algorithms for Solving the Nonlinear Pattern Classification Problems

The main object of this thesis is Ato study artificial neural networks algorithms for solving the nonlinear pattern classification problems. This thesis consists of an introduction, four chapters, and three appendices . It is organizedas follows:In chapter 1, we introduced the basic concepts of the pattern classification and recognition problems. We alsoA introduced the traditional and advanced pattern recognition methods. Finally, we introduced the invariant feature extraction methods by using tIn Chapter 2, we have been used a supervised backprobagation learning process of neural network to recognize a set of patterns throughout determines their invariant features. Invariant features of the pattern can be determined by the seven invariant momenSo the feature extractor of windowed areas of images is a very good than traditional method in case of huge size of images to reduce the time of calculations. We also used the invariant Fourier descriptor FD and its Laplace transform of the edge detection