基于有偏彩色纹理字典的X光安检图像检测  

Security X-ray images detection based on biased color texture dictionary

作　　者：刘颖 王伊琳 王倩 侯祥玮 

机构地区：[1]西安邮电大学通信与信息工程学院,陕西西安710121 [2]电子信息现场勘验应用技术公安部重点实验室,陕西西安710121

出　　处：《西安邮电大学学报》2017年第6期35-39,共5页Journal of Xi’an University of Posts and Telecommunications

基　　金：国家自然科学基金资助项目(41504115);公安部科技强警基础工作专项基金资助项目(2015GABJC50;2016GABJC51);陕西省自然科学基础研究计划资助项目(2015JQ6223)

摘　　要：为提高X光安检图像的检测效率和正确率,通过改进局部二值模式算子,给出一种基于有偏彩色纹理字典的X光安检图像自动检测算法。先根据被检测物体的颜色分布构建色彩权值矩阵,提取多通道图像的局部纹理特征,结合词袋模型生成多通道优化的图像字典,再通过色彩加权矩阵映射到有偏彩色纹理字典,并通过训练支持向量机对输入的测试图像进行检测。对于相同训练集与测试集,改进算法能准确对X光安检图像进行目标检测,且检测正确率和查全率均有提升。In order to improve the efficiency and accuracy of X-ray image detection,a kind of automatic detection algorithm for X-ray image based on biased color texture dictionary is proposed by revising the local binary pattern operator.Firstly,a color weight matrix is constructed according to the color distribution of objects to be detected.Then,a multi channel optimized image dictionary is generated by the bag of words model and the local texture features which are extracted from multi channel images.and is mapped to a biased color texture dictionary based on the color weight matrix.Finally,The input image can be detected by training support vector machine.For the same training and testing images,the revised algorithm can accurately detect the target of the X-ray images,and its detection accuracy and recall ratio are improved.

关 键 词：颜色分布 纹理特征 词袋模型 图像检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]