Invariant in the Pattern Recognition

Consider the problem analysis and selection of informative features, which are the luminance and geometric invariants. An analytical overview of the works, which raises the problem of constructing invariants and solve practical problems. Among the invariants to the multiplicative and additive transformations brightness discussed the correlation coefficient, the normalized histogram of some special relationship of line segments and angles. It is considered as the simplest geometric and integral invariants resistant to affine transformations, as well as the transformation strain. Among all geometric invariants allocated invariants Hu, in which interest is not waning and now. Since there is no evidence in the original sources of the invariance of the allegations, the great attention is paid to checking on some points invariance to rotation of frequently used operations. The above features can be used for solving problems of pattern recognition, distance measurement, analysis of complex scenes. It is expected that in the future on the basis of brightness and geometric invariants will be solved the problem of recognizing individual rights. This article is intended for a wide range ofpost-graduate students and engineers involved in pattern recognition tasks in various applications.

image, graphical image, recognition, brightness invariants, invariant moments, affine transformations, stretching