Web页面色情图片层次化识别算法研究

Research Of An Hierarchical Pornographic Image Recognition Algorithm On Web

随着网络技术的不断发展,互联网上色情图片的传播也越演越烈。为有效解决此问题,本文提出Web页面色情图片层次化识别算法,该算法融合了肤色判断、人脸识别和图片语义识别技术,依照"由粗到细,由表及里"的原则对色情图片进行逐层识别和过滤,并在语义识别模块中应用Bayes算法将图片的底层视觉特征映射到高层语义特征,实现了图片的语义识别过程。本地和实网测试结果表明,该算法能够在不影响网络正常运行的前提下,达到较高的准确率,具有一定的实用性。

With the incessant development of network technology, spread of pornographic image on web is more and more intense. In order to solve this problem, this paper develop a hierarchical pornographic image recognition algorithm. The algorithm synthesize tech- nique of skin recognition, face recognition and semantic image recognition. It recognize and filter pornographic image hierarchically according to principle which is ＂from wide to fine, from exterior to inner＂. We use Bayes algorithm to transform visual characters to semantic characters of image in semantic recognition module in order to realize semantic recognition. The result of local experiment and network experiment shows that, the algorithm can achieve high recognition rate on condition that network can work normally, which has more practicability.