摘要
为了快速、有效地分类出敏感视频,提出一种基于人体肤色识别的敏感视频分类方法。首先通过帧差图像提取前景运动区域,然后构建人体肤色模型,并结合前景运动区域识别出视频中的人体,最后用多因素的级联分类器进行敏感视频分类。实验结果分析表明,文中方法有较好的分类准确率和查全率。
In order to effectively classify the obscene videos, a novel obscene video classification approach based on human skin recognition is proposed. We first extract the foreground motion area by using frame difference images. Then we make up the human skin model and recognition human body by combing with the extracted foreground motion area. Finally, the multi-feature cascade classifier is used for classify the obscene videos. The experimental results show that our proposed method has better precision and recall rate.
作者
梁鹏
林智勇
贾西平
LIANG Peng, LIN Zhi-yong, JIA Xi-ping (School of Computer Science, Guangdong Polytechnic Normal University, Guangzhou 510665, China)
出处
《电脑知识与技术》
2016年第5期181-183,200,共4页
Computer Knowledge and Technology
基金
广东省自然科学基金-博士启动(2015A030310340)
广东省高等学校优秀青年教师培养计划(Yq2013108)
关键词
敏感视频分类
肤色识别
前景运动
级联分类器
obscene video classification
skin recognition
foreground motion
cascade classifier