摘要
该文针对互联网视频聊天中可能存在的色情内容,设计并实现了一个基于普通二进制分类树的判别系统,构成视频聊天预警系统的核心模块。该判别系统结合了相似性,肤色,人脸等特征,通过对视频截图的敏感性判别,将结果反馈给管理者,达到安全预警的目的。实验表明,系统在截图敏感性判别时可以达到90%以上的召回率和85%以上的正检率,具有较高的实用性和应用前景。
To prevent pornography content in the process of video chatting online,the paper designs a recognition system based on ordinary binary classification tree(OBCT),which is a core module of the video chat warning system.Some features such as similarity,skin and face are used to detect the sensitivity of the screenshots.For the purpose of safety,the results of recognition are feedback to the webmasters.Experiments show that the recall of the sensitive screenshots can achieve more than 90%,and the error rate can achieve less than 15%.
出处
《杭州电子科技大学学报(自然科学版)》
2012年第1期56-59,共4页
Journal of Hangzhou Dianzi University:Natural Sciences
基金
浙江省科技厅重大专项资助项目(C11049)
教育部人文社会科学研究基金资助项目(08JC740011)