欢迎来到天天文库
浏览记录
ID:53029570
大小:2.86 MB
页数:13页
时间:2020-04-14
《基于时空纹理的实时群体行为检测-论文.pdf》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、第2O卷第2期西安邮电大学学报Vo1.2ONO.22015年3月J0URNALOFXI’ANUNIVERSITYOFPOSTSANDTELECOMMUNICATIONSMat,2O15doi:10.13682/j.issn.2095—6533.2015.02.013Real-timecrowdanomalydetectionbasedonspatio--temporaltexturemodelWANGJing,XUZhijie(Visualisation,InteractionandVisionResearchGroup,UniversityofHuddersf
2、ield,HuddersfieldHD13DH,UK)Abstract:Aninnovativespatio—temporaltexturebasedoncrowdmodellingtechniqueanditscorrespondingpatternanalysismethodsareintroduced.Thealgorithmisdesignedforrealtimeapplicationsbydeployinglowlevelstatisticalfeaturesandavoidingcomplicatedmachinelearningandrecogn
3、itionprocesses.Throughextractingandintegratingthosecrowdtexturesfromliveorrecordedvideos,theSOcalledhomogeneousrandomfeaturesaredeployedintheresearchforbehaviouraltemplatematching.Experimentresultsshowthattheabnormalityappearingincrowdscenescanbeeffectivelyandefficientlyidentifiedbyu
4、singthedevisedmethods.ThisnewapproachisexpectedtofacilitateawidespectrumofcrowdanalysisapplicationsinthefuturethroughlayingasolidtheoreticalfoundationandimplementationstrategyforautomatingexistingClosedCircuitTelevision(CCTV)basedsurveillancesystems.Keywords:crowdanomaly,spatio—tempo
5、ralvolume,spatio-temporaltextureCLCNubmer:TP391.41DocumentCode:AArticleID:1007—3264(2015)02—0064—13基于时空纹理的实时群体行为检测王晶,许志杰(哈特斯菲尔德大学可视化交互及视觉研究组,英国哈特斯菲尔德HD13DH)摘要:描述一套全新的基于时空纹理特征的群体行为模型并对其特征进行分析。该模型针对实时环境中的安检应用,使用丰富的底层图像特征及简洁的行为识别算法,另外定义并提取视频图像中群体行为的纹理特征,使用同性随机场加以描述。所建模型可以方便地定位出异常群体行为的发生
6、位置及时间。通过实验可证明,该方法能快速有效地提取及分析视频异常行为,并能在基于闭路电视的智能安检系统中发挥其优势。关键词:群体异常行为;时空体数据;时空纹理detectionareas.Thesetechniques,suchascrowd1Introductionanomalydetection,canbeusedinmonitoringandTheincreasingdemandsofintelligenttrackingemergencysituationsoccurredincrowdedsurveillanceapplicationshavetrig
7、geredmanyscenes,includingbusymotorways,highstreets,innovativedevelopmentsinautomatedvideoeventsportingevents,andopenairconcerts.Receivedon:2015——02—‘10Supportedby:ScienceFoundationofChina(61202183);ShaanxiProvinceEducationOfficeproject(12JK0504,12JK0374)Contributedby:WangJ(1983一),mal
8、e,HaD,resear
此文档下载收益归作者所有