Yakovleva E. Metods of one-dimensional normalization of affine transformations in the tasks of image recognition

The subject of the thesis is solving of image recognition problem in the systems of artificial intelligence. The methods of one-dimensional normalizations have been worked out; these methods allow image recognition under conditions of affine transformations and consist in the analysis of restrictions (profiles) of images on some straight lines. The methods of one-dimensional normalizations are based on the research of induced actions on straight lines caused by various transformations of plane and invariant properties of families of straight lines for the affine group and its subgroups. The developed methods allow finding the corresponding restrictions of image and defining parameters of twodimensional normalization. The recognition decision is based on the comparison of normalized and template images. The developed algorithms of realization of the proposed methods are used for creation of specialized software of image normalization and recognition on the basis of comparison with templates