基于SVM的检测图像型垃圾邮件技术研究

Research on detecting image-based spam based on SVM

垃圾邮件给人们带来了极大的困扰,而图像型垃圾邮件使得传统的反垃圾邮件技术失去了检测能力。在分析图像型垃圾邮件特点的基础上,首先针对传统的SUSAN算子,提出一种自适应阈值SUSAN算法;其次通过源于专家经验的启发性知识筛选图像中的垃圾区域;最后,引入机器学习的支持向量机分类方法。实验表明,论文设计的方法具有很好的鲁棒性与较高的精确度,能够过滤掉图像型垃圾邮件。

The spam has brought great distress to people＇s work and life,but with the emergence of image-based spam,the traditional anti-spam technology loses the ability to detect them. In this paper,based on a detailed analysis of image-based spam,firstly,an adaptive threshold SUSAN algorithm is created; followed by enlightening experience from expert knowledge of image spam to filter area; finally,the introduction of SVM classification of machine learning is necessary. The experiments show that the method has a good robustness,high accuracy to filter out image spam.