Discussion on website verification code recognition program (1)

Author: Caocao, http://www.caocao.name/,http://www.caocao.mobi/
Reprinted please indicate Source: http://www.javaeye.com/topic/159922

As the name suggests, the hermit wants to discuss the design of the verification code recognition program. As for how to use such a program, it is not in the scope of technical discussion.

Let's take a look at the effect of the. NET program. Such verification code images are available on many websites, and the recognition effect looks good.

The following describes the features of the algorithm:
1. Only known fonts and font sizes can be identified. There are no requirements for letters, including English letters, numbers, and Chinese characters.
2. Only words that are not distorted, mirrored, or rotated can be identified.
3. not sensitive to noise and interference lines.

According to the above features, this algorithm is actually very fragile, as long as the verification code generation program is slightly rotated, it can make it invalid. So what does the algorithm do?

This algorithm only performs one thing, that is, scanning the verification code Image Based on the image of a known string, and scanning the matching image. Of course, there will be some minor adjustments in the algorithm to reduce the false recognition rate. This alone is enough to identify a large number of website verification codes. The verification code in is by discuz! One generated in 5.5, csdn currently uses the same verification code image to identify it. If you are interested, contact the hermit.