一种彩色旋转变形验证码的识别方法

摘要

Aiming at the color rotation deformation CAPTCHA, a character segmentation method which combines the method of connected regions and color space was proposed, and a multi-class support vector machine was used to recognize. Firstly, the CAPTCHA image was grayscale, binarization and removing noise points. Then, the character was segmented by the connected region method, and the color space method with HSI model was used to further segment the overlapping char-acters, normalized processing after segmentation. At last, the recognition experiment was conducted by support vector ma-chine. The experimental results show that the support vector machine is fast and does not require a large number of training samples, has a high recognition rate for the rotation and deformation characteristics CAPTCHA, and also show this character segmentation method is more effective than other methods.%针对彩色旋转变形验证码提出了一种连通区域法和颜色空间法相结合的字符分割方法，并采用多分类支持向量机进行识别。首先，对验证码图像进行灰度化、二值化以及去除噪声点；接着，使用连通区域法分割字符，而对于重叠字符采用HSI模型的颜色空间法进一步分割，分割后进行归一化处理；最后，通过支持向量机进行识别实验。实验结果表明，支持向量机训练速度快，不需要大量的训练样本，对旋转和变形特征的验证码具有很高的识别率，同时也说明了该字符分割方法比其他方法更有效。