一种两层体系刑侦现勘图像检索方法  

A two-phase hierarchical approach for crime scene investigation image retrieval

作　　者：刘伟[1,2] 刘颖[1,3] 李大湘[1,3] 朱婷鸽[1,3] 

机构地区：[1]公安部电子信息现场勘验应用技术重点实验室,陕西西安710121 [2]西安邮电大学计算机学院,陕西西安710121 [3]西安邮电大学通信与信息工程学院,陕西西安710121

出　　处：《西安邮电大学学报》2016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);陕西省教育厅专项科研计划资助项目(14JK1664)

摘　　要：给出一种两层体系刑侦现勘图像检索的方法。使用线性核支持向量机分类器和图像的多尺度分维数特征判别查询图像的语义类别,根据类别在仅包含该语义类图像的数据集上使用按例检索方法进行检索。实验结果表明,该方法在查全率和检索时间等指标上优于按例检索方法。A two-phase hierarchical approach for crime scene investigation image retrieval is proposed.In the first phase,support vector machine with linear kernel function and multi-scale fractal dimension feature are used to classify the class label of the query image.In the second phase,according to the classified label,a smaller dataset only including images with the same label as the query image is picked out from the image database.Then query by example method is used to retrieve images from the dataset.Experimental results show that retrieval accuracy based on the proposed approach is better than that based on conventional query by example method.Particularly,the proposed approach is computationally more efficient than conventional method.

关 键 词：法庭科学 刑侦现勘图像 图像检索 特征提取 分类器 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TP751.1[自动化与计算机技术—计算机科学与技术]