| As the development of the 21 th century,the comprehensive national strength of china is becoming stronger.The pace of law construction is accelerating,and the idea of handling things according to the law is widely spread.Handwriting identification,as a part of evidence technology in law,plays an important role in political and economic activities.Especially,in work of audit department,handwriting identification is usually critical.However,due to the need of strong professionalism,we have to submit the handwriting material to professional department and get an identification result.By doing this,I usually waste a lot of manpower and resources.What's more,the result we got is sometime s not highly Credible because of too much human intervention.All the reasons mentioned above make this technology far from being satisfied.Auto writing identification is used to solve this problem.In this thesis,I do a lot of research on auto handwriting identification and develop an identification system of high accuracy and high efficiency.According to the common methods of handwriting and lots of experience in daily work,I propose an auto writing identification technology by combining various features in bills.I divide the technology into four stages,including handwriting pre-processing,feature extraction,features fusion and discrimination.In the pre-processing stage,I get a picture of bill and roughly extract text content,then I do further processing to refine the result.As the same time,I do pre-processing on several contrast data.Now I get pairs of refined data.Next,in the feature extraction stage,I reference the whole-and-part principle and combine several features,such as height-weight radio,inclination,SIFT,to increase the amount of information without losing uniqueness.At last,according to the features extracted,I compute the similarity between testing sample and comparing samples.Experimental results show that,the system I developed can complete handwriting effectively. |
PDF Full Text Request