东南大学学报（自然科学版）Issue(6)：1111-1115,5.DOI:10.3969/j.issn.1001-0505.2014.06.003
一种基于人体裸露皮肤形状的不良图像过滤系统
Pornographic image filtering system based on shape of naked skin
摘要
Abstract
Based on the shape of naked skin in a figure image,a novel pornographic image filtering system is proposed to improve the filter precision of pornographic images.This system contains adaptive skin segment module,features extraction module and classifier module.First,based on face detection and Graph Cuts segment method,an adaptive skin segmentation method is proposed to ac-curately detect skin color pixels of the images containing human body.Then,the feature vector, which is the combination of the global feature and the local feature,is constructed by Zernike mo-ments and grayscale co-occurrence matrix.Finally,the genetic algorithm is used to optimize the sup-port vector machine based on RBF(radial basis function)kernel.The experimental results show that the proposed filtering system can accurately screen out the pornographic images with more naked skin and the precision ratio of positive samples exceeds 94%.关键词
Zernike矩/灰度共生矩阵/支持向量机/遗传算法Key words
Zernike moments/grayscale co-occurrence matrix/support vector machine/genetic algo-rithm分类
信息技术与安全科学引用本文复制引用
黄杰,史啸..一种基于人体裸露皮肤形状的不良图像过滤系统[J].东南大学学报（自然科学版）,2014,(6):1111-1115,5.基金项目
国家高技术研究发展计划（863计划）资助项目（2013AA014001）． （）
