Oriented Web Proxy Authentication Code Picture Identification

Validation code is an important branch of Optical Character Recognition. Itplays an important role neither in accelerating the development of network securitynor in improving intelligent people use network.Along with the further development of the network, it provides people withabundant resources. In order to ensure the security of network, now many websitesuse validation code technology to restrict users' operation. Therefore, the purpose ofthis text is to take advantage of OCR technology to realize the character recognitionof validation code. The intention is to achieve a smart download by which you canskip the process of theft proof chain's validation code to realize network's intelligentdownload function.This text will concisely describe system providing download based onvalidation code and the system for downloading based on validation code, and thenput the focus on the character recognition of validation code. The validation code ofthis text using divides into two part, one is used in the intelligent download system fora certain website. There are just regular numbers in it. The other is more difficult torecognize, because there has a lot of interference pixels, and between characters thereexists adhesion. Therefore, for a higher accuracy rate of the recognition rate, thepreprocessing must resolve the above two problems. Against noise pixels, the textused median filters algorithm to achieve the desired results. As between characters dohave a certain tilt and adhesion after initial processing, the average segmentationalgorithm can be difficult to segment projection crossed but actually does not crossedcharacters as well as adhesive ones. Through analyzing the advantages of severalreference algorithms which have limitations, this text integrated some of them tobetter deal with the segment of adhesive characters which are mentioned in this text.Recognition is the final step to achieve this function. As characters in this textwere found to be regular on the basis of many of the extracted samples, therefore, abetter result has been achieved based on the use of Hausdorff's distance templatematching.