摘要： 鉴于网络上广泛使用的验证码类型存在安全性不高的问题，从验证码的破解角度，提出一种新的针对倾斜数字的自动矫正算法，可自动找出数字字符底边和倾斜方位角。算法包括3个方面，即验证码数字坐标化、通路搜索及通路分析。坐标化基于前期预处理过程，能定位倾斜角度。通路搜索确定字符连通情况并为后续通路分析阶段提供判定信息。经过实验证实，该算法对于标准字符能达到精度接近100%的矫正率，对于微形变的字符也能达到约95%的平均识别率。

关键词: 验证码, 数字坐标化, 通路搜索, 通路分析, 倾斜矫正

Abstract: Considering the widely used CAPTCHA on Internet is still at the risk of low security, a novel automated correction algorithm for declining numbers is proposed from the angle of recognition, through which can automatically find the bottom and azimuth tilt of the characters. The algorithm involves three aspects which are the coordinate of digital numbers, channel search and channel analysis. Among them, the first step which based on the early preprocessing process can be used to verify the tilt while the channel search process determines the connection of the characters and provides information for the channel analysis stage. Experimental results show that it can reach nearly 100% accuracy of the correction rate for standard characters and 95% of the average recognition rate for characters with micro-deformation.

Key words: CAPTCHA, coordinate of digital numbers, channel search, channel analysis, declining correction

中图分类号: 

• TP301.6