基于视频的车标检测识别技术研究

《基于视频的车标检测识别技术研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、lZl2597四川大学硕士学位论文题目一基王塑频的奎拯捡型迟型撞盔硒究作者赵塞亘完成日期2Q笾生!Q旦旦培养单位四刖态堂指导教师迸查胜专研究方向图形图象矬理授予学位日期生县旦四川大学硕士学位论文法。大量的实验结果证明，本文采用的技术是合理的，具有较大的理论意义和实用价值。关键词：车型识别车标定位车标识别四川大学硕士学位论文Studyonvehiclelogodetection&recognitionbasedonvideoimageSpeciality：ComputerAppliedTechnologyGradua

2、l：ZhaoDakeAdvisor：Prof．YouZhi-ShenRecognitionsystemofvehiclemainlyconsistsoftwokeytechnologies：recognitionofvehiclelicense-plate(、卫P)andrecognitionofvehicletype．ItnotonlyfindS埘deapplicationinITS．butalsoiSahotpointofresearchincomputervision,ilnagcprocessingandp

3、aRemrecognition．Soitsrelatedtechnologyisattendedprevalently．Inourpaper,wepayattentiontothelocationandrecognitionofvehiclelogo．Themethodstodealwithaboveproblemsareproposedinthepaperandtestifiedintheexperiment。11lemaincontentandinnovationofthedissertationareasfo

4、llows：(1)．Vehicle—logolocatiOilalgorithm：WhileVLPhasevidenttexturefeatureandregularshape，thetexturefeature，shapeandsizeofeveryvehicle—logo31"edifferent．Itc蛆collsumemuchtimeifweadoptthegeneralobjectlocationalgorithm,suchasrelationmatch．Sotheproblemofvehicle·log

5、olocationhasbeenabottleneckinthereal-timeapplicationofvehicle—logorecognition．Forthesakeofsuchsolutions，afastandrobustmethodofvehicle—logolocationispresentedinthispaper．Byfilllutilizeofthelocafionofvehicle．1icense-plate．anditseorrelationofvehicle-logoandvehicl

6、e-head,andbyutilizeofaxissymmetryofvehicle-head,anexperimentalroughsearchingrectangleofvehicle—logoCanbelocated．Furtheron,byanadaptivemorphologyfilterandanimprovedtemplatematchingmethod,thevehicle—logoCallbeaccuratelylocated．(2)．Vehicle-logorecognitionalgorith

7、m：Afterlocation,vehiclelogorecognitionbecomesa2-Dshaperecognitionproblem．Butbeingsubjecttoillumination,noise，shapesimilarityandpartocclusion,itishardtogetsatisfactoryresultsbycommonmethods．Toimprovetheaccurateofrecognition,avehiclelogorecognitionmethodwhichint

8、egratestemplatematching、Ⅳitllfeaturematelfingisproposed．Experimentalresultsapprovetheaccurateofthemethod．Thetopicbasedonthecomputervision,pattemrecognitionandimageprocessingtechnol