character recognition method

Abstract

The present invention relates to a character recognition method, which includes: step 1: read in character image; step 2: image preprocessing; step 3: character cutting; step 4: edge extraction; use edge detection operator to detect the edge of the character point; step 5: feature extraction, the feature of each edge point of each character is expressed by the distance from other edge points of the character to the edge point, represented by feature vector (P1k, P2k...PMk); step 6: feature processing , map the feature vectors (P1k, P2k...PMk) into a matrix T, so that the feature vector dimensions of all characters are the same; step 7, template matching recognition. Since the feature of each edge point of the character is expressed by the distance from other edge points of the character to the edge point, the globality of the feature is guaranteed from the whole character, and the difference of features between different characters is enlarged, thereby improving the character recognition rate.

Description

Technical field [0001] The invention belongs to the field of optical character recognition, and relates to an optical character recognition method. Background technique [0002] In the character recognition process, character feature extraction is a crucial step. Feature extraction can be regarded as the extraction of the key information of the original characters, so the features should reflect the differences between different characters as much as possible. In other words, the distance between the features of different types of characters should be as large as possible. If the features are not selected properly, no matter what kind of classifier is selected, a good recognition rate cannot be achieved. [0003] Recognizing a character is usually to extract the edge of the character or extract the skeleton of the character, and then extract the features on the edge or skeleton. Existing methods only consider the information of the current point when describing the point on the e...

AI Technical Summary

Problems solved by technology

For example, (x, y, f(x, y)) is used as the feature of a point P on the edge or skeleton, where x, y represent the coordinates of point P, and f(x, y) represents the brightness value, so the extracted features have Locality, if two characters have similar local shapes (such as Q and O), then the features extracted from the two characters have a great similarity, which is not conducive to distinguishing the two characters, resulting in a decline in the system recognition rate

Images