Research on verification code target recognition algorithm for multi-type interference
Verification code is a technology employed to differentiate computers and humans so as to avoid security issues such as automated program registration,login,malicious voting,and malicious password spraying.We aim to find and correct the shortcomings of verification code in time,which has important significance for improving network security.Based on the research on the verification code recognition algorithm,this paper designs a verification code recognition system,employing template matching,Hu invariant moment and deep learning algorithms to process and recognize the verification code image,and finally obtains recognition results for the verification code with multiple types of interference.Our results show the recognition accuracy of the first two verification code algorithms is 60.3％and 66.0％respectively,while that of the two forms of the deep learning algorithm is 87.7％and 95.58％.In the complex interference such as online interference and mixed interference,the recognition accuracy is markedly improved,demonstrating the feasibility of deep learning algorithm on multi-type interference verification code recognition.
国家科技期刊平台
NSTL
万方数据
