Method for recognizing character based on affine invariant

Abstract

The invention belongs to the field of computer vision and relates to a method for recognizing a character based on the affine invariant. The method for recognizing the character based on the affine invariant comprises the following steps: firstly, getting a convex hull from an objective character, constructing a series of straight lines intersecting with the contour of the character, and determining the characteristic ratio value on each corresponding straight line by means of introduction of the affine invariant, namely the characteristic ratio; secondly, combining the characteristic ratios into an eigenvector in sequence, and expressing the eigenvector as a series of characteristic ratio spectrum; lastly, comparing the distances among the characteristic ratio spectrum by means of a DTW algorithm, obtaining the similarity among characters, and conducting character recognition. The experiment shows that the method for recognizing the character based on the affine invariant not only has the higher recognition rate of serious affine deformation but also has a good distinguishing effect on characters which have higher similarities.

Description

technical field [0001] The invention belongs to the field of computer vision, relates to the description and matching of symbols, in particular to a symbol recognition method based on affine invariants. Background technique [0002] Symbol recognition is a basic problem in the field of computer vision, which has important theoretical significance and extensive application value, and can be applied to a large number of image processing applications, including robot navigation, image retrieval, etc. As an important basis for human observation and description of objects, symbols have always been concerned. The purpose of symbol recognition is to find a simple and effective symbol feature description and comparison method, which can not only distinguish the shape difference of different types of objects, but also tolerate the serious deformation of symbols under the change of viewing angle. [0003] A common approach in object recognition is to find corresponding feature points...

AI Technical Summary

Problems solved by technology

However, it is not suitable for symbol recognition under affine conditions

[0005] Chinese invention patent of Huazhong University of Science and Technology, application number 201110080680, "Symbol recognition method based on sixteen-value transformation", this patent generates 16-value representation features from the positional relationship of pixels, and then performs feature matching and recognition, with feature extraction Simple, high recognition rate, fast recognition speed, etc. However, it is still unable to recognize symbols with large deformation in real scenes, especially for similar symbols.

Images