Font Size: a A A

Revolving Invariance Printing Numerals Recognition Based On Combined Classifier

Posted on:2006-03-05Degree:MasterType:Thesis
Country:ChinaCandidate:S H TengFull Text:PDF
GTID:2178360185463708Subject:Electronic Science and Technology
Abstract/Summary:PDF Full Text Request
In the technique of Optical Character Recognition, when recognizing the character with complicated space of a whole page, it is required that the character with big revolving angle should be recognized directly without rectification. Aim at this problem, revolving invariance printing numerals recognition system is set up in this paper. This system includes three parts: preprocessing part, extract feature part and recognition part. During the realizing of the system,the main studding contents and new ideas shown as follows:1. In this paper, many preprocessing ways of printing numerals are used. For example, smoothing,binaryzation, normalization and thinning. What's more, improvements on thinning arithmetic are presented. The results of experiment show that the improvements are valid.2. The paper extracted some apropriate features—different structural features and different statistical features. The structural feature, which has high accuracy, is used to classify the similar characters while the statistical feature, which is simple, speedy and stable, is used to enhance the ability of resisting noise. The results of experiment show that it's helpful to improve the performance of the whole recognition system to combine the merits of two different features.3. A combined classifier is designed. There are three recognition ways used in the classifier. They are the template matching method, fuzzy pattern recognition method and the raising one's hand method. There are three steps in the combined. Firstly, the template matching method is used to classify roughly in the first step; Secondly, the fuzzy pattern recognition method which are connected paralled is used to classify finely; At last, the third classifier, which input is the output of the second classifiers, raise one's hand as its decision rules. The results of experiment show that the classifier combination offset the disadvantage of each component classifer and retain its advantage at the same time.Involing invarence printing numerals recognition system based on the combined classifier is built up in this paper. Many apropriate features ways and many pattern recognition are used in the system. The results of experiment show that this system can get high correct rate and high reliability in recognising Printing numerals with big revolving angle. The proposed method is not only suitable for the involing invarence printing numerals recognition,but also for the other areas, such as script numerals which restrict calligraph, the method also does well.
Keywords/Search Tags:revolving invariance, feature extraction, classifier combination, fuzzy recognition
PDF Full Text Request
Related items