犯罪工具图片快速溯源技术研究  

Technology of Fast Traceability for Pictures of Crime Tools

作　　者：潘楠[1] 阚立峰 刘益 

机构地区：[1]昆明理工大学航空学院,云南昆明650500 [2]昆明理工大学机电工程学院,云南昆明650500 [3]昆明信诺莱伯科技有限公司,云南昆明650228

出　　处：《昆明理工大学学报（自然科学版）》2017年第3期52-59,共8页Journal of Kunming University of Science and Technology(Natural Science)

基　　金：云南省科技计划重点项目(2016RA042);昆明市科技计划重点项目(2015-1-S-00284);云南省科技型中小企业技术创新项目(2017EH028);公安部技术研究计划重点项目(2014JSYJA020;2016JSYJA03);国家留学基金委创新型人才国际合作培养项目(201608740005)

摘　　要：为克服传统人眼判断方式识别和溯源犯罪工具的低效率和低精度问题,提出了一种犯罪工具图片快速溯源技术.首先通过GVF(梯度向量流)来平滑图像的梯度场,然后使用Harris角点检测算法检测出工具角点,紧接着使用SIFT算法描述特征点对应的特征向量,最后以特征值为样本基础,利用KNN算法进行投票决策.使用Eclipse+Java+Open CV对该技术进行编程实现,并付诸实验.实验中,对每个待检测样本均进行10次相似度匹配,将匹配度最高的样本作为最为相似的样本,10组数据总计100次测试.样本识别匹配耗时近2 s,选取相似度程度最高的样本作为识别匹配结果,只有2次错误匹配,匹配率约为98%.本技术经铁路公安局刑侦局和黑龙江省公安局刑侦总队痕迹检验专家的测试,证明其能够有效地应用于作案工具图片的批量对比匹配之中,为犯罪工具快速溯源提供可靠依据.In order to overcome the problem of low efficiency and low accuracy when using traditional way to judge crime tool by human eyes, a fast image tracing technology for crime tool is proposed. Firstly, the GVF （Gradient Vector Flow） is adopted to smooth the gradient field of image, the Harris comer detection algorithm to detect the comer tool, and then SIFT algorithm to describe the eigenvectors corresponding feature points. Finally, KNN algorithm is employed to vote the samples based on characteristic value. Eclipse ＋ Java ＋ OpenCV is used to implement the technology, and experiment with it. In the experiment, each sample to be tested is matched similarly for 10 times. The highest degree of matching is then taken as the most similar one. There are a total of 100 tests for the 10 sets of data. Sample ID match takes nearly two sec- onds, and the sample with the highest degree of similarity is selected as a recognition match. As the result shows, there are only 2 errors, and the matching is about 98%. This technology is tested actually by a number of experts from Railway Public Security Bureau of Criminal Investigation and the Heilongjiang Provincial Public Security Bureau of Criminal Inves- tigation Corps. It is proved that it can be effectively used for the comparison match for a variety of pictures of crime tools, which provides a reliable basis for rapid traceability of crime tools.

关 键 词：SIFT算法 梯度向量流 HARRIS角点检测 KNN算法 特征提取 

分 类 号：TN911.72[电子电信—通信与信息系统] TH39[电子电信—信息与通信工程]