Postal coding numberical string identifying method

Abstract

A method for identifying digit string of zip code includes inputting N number of images for zip code character series to K numbers of sorters for single word identification, identifying inputted image to be one of (C1...Cm...Cm) or refusing to idenfity, calculating out probability p(D|X) according to identification result of D+=(d1, d2..., dn) as D=(d1, d2...,dN) being a effective zip code in zip code dictionary bank and deciding identification result of input mode as per probability p(D|X).

Description

technical field [0001] The invention relates to a method for identifying postal code number strings. Background technique [0002] After decades of development, optical character recognition technology has gradually become practical. However, people still hope that the recognition system can achieve better recognition performance. In order to improve the recognition rate and confidence, people are more and more inclined to use a combination of multiple information sources, multiple feature extraction and multiple recognition methods to achieve a high-performance recognition system. [0003] A simple and easy way to combine the existing zip code digit string multi-classifiers is to vote, such as the majority rule and the completely consistent rule. However, these voting rules do not take into account the characteristics of each classifier itself, and implement the principle of "one person, one vote". In fact, because each classifier uses different features, based on differe...

AI Technical Summary

Problems solved by technology

In fact, because each classifier uses different features, based on different principles and methods, or the samples used in the training process are not the same, the recognition performance of each classifier is different, and there is a certain complementarity, that is, each classification There is a certain difference in the recognition ability of the device for each category

Images