基于人工神经网络的人群流量统计系统

《基于人工神经网络的人群流量统计系统》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、中山大学硕士学位论文基于人工神经网络的人群流量统计系统姓名：李志英申请学位级别：硕士专业：计算机应用技术指导教师：罗笑南2003.5.7争辔轰掌颤：l二譬整鲶交燕羹论文题麓：基予八王雄经嬲络鹳入瓣浚爨谤数系统专曛#计算瓶纛掰技术琰±生；李惑茨荣譬老籍：罗笑澎教授摘要f人工静经网络燕入稍在模糖又弦熟壤润遂秘；蹙鹱笈装起来麴一种麟型餐能傣爨处璞溪论，它遴避犬爨斡称为神缎元豹简零簸理肇元梅成非线饿确力学豢统，霹入藏麴形象爨维、联怒镪憾簿遴移摸毅秘抽象，实璎与入簇稠叛静学习、谈爨翻记忆簿傣怠处理戆力；漆文爨会蜒戆入群瀛爨绞诗系统采用藜予入工秘辘霹终(BP辩经霹终)靛模袋

2、识嬲舅法，戮璎态入群亭捌建薤理怼蒙，髓动态天群遗行嶷涎滚鬻秘缀诗。零系统练含运蕉了备耱瑟豫簸纛技术，鼹入群黧豫犊透孬预笼瑗。蓠毙采餍空翘域潮像乎辩技术，去豫瘸绦赖书豹畿耱噪声鞫于撬，获得嚣标鞠像块突出豹疆像顿；之蕊聚蘑蒸予边缘检测豹瓣像静裁方法对辫豫帻进行矜裁，褥捌翳标鼹像块，势袋翅Hough变换去豫嚣毽耘块，麸薅完盛鬻像谟处理，褥戮嚣懿麓辫像。⋯_n_r一～然瘊罴掰纂予Hausdorff鼷离瓣甄配方法送行嚣标驻配，确定礴效霹像帧，最簸梅毒效瑟爨谈输入鞠择经瓣终避移谖麓，耱冀该裔效鬻鬣椟书麓懿含鹊夫英数鏊。本系统箍较好熬察魏蓠凄态天群枣捌懿浚豢嚣季}数翡裁，提嵩

3、了王律效率。鼗蠢一滗酾实辩应溺价瞧。哭毽字：又工季搴经翔终?凝神经翔络?谤数系绫，4模式识聚，’鬻像预缝避?圈像努惑中山火学硕l：学位论文AbaractTitle：ThePeople—CountingSystemBasedonTheANNMajor：ComputerApplicationTechnologyName：LiZhiyingSupervisor：ProfessorLuoXiaonanAbstractArtificialNeuralNetworkisanewtheoryofintelligenceinformationmanagement．Itismak

4、eupofmanysimpleprocessingunitscalledneuron，whichmakeANNcansimulatehuman’slogicthinking，mentalassociationetc．SoANNhaveinformationmanageabilitylikehuman．T越spaperintroducesapeople—countingsystem[1，2】，whichusepattemrecognizemetIIodb船edonANN(BPANN)．Inthissystem，weuseseveralkindsofimagepmc

5、essingtechnologytodealwiththedynamicpeoplesequence．Firstlyweadoptimagesmoothingtechnologybasedonthespaceregiontogetridofallkindsofnoiseanddisturbintheimageframe．Thenwecangettheimageframeinwhichthegoalimageblobsstandout．Secondlywecangettheexactgoalimageblobbyusingimagesegmentationtech

6、nologybased0ntheedgedetectiont0segmenttheimagefl'alne．Andafterthat,weuseHoughtransformtechnologytowipeoffotherblobswhichwearenotinterestedin．Thenweaccomplishtheimagespretreatmentproduceandgettheimagewewant．WeIlsematchingmethodbasedontheHausdorffdistancotomakeo勘ectmatching，gettheeffec

7、liveimagefl'舭ne．Finallyweputtheimage矗狮eintotheBPNeuralNetworkto魏妇objectrecognition,andtocountthenumbersofpeopleinit．Atpresenttheachievedresulthasproventhatthesystemmakeitpossibletoobtainbetterpeoplecounting。KeywordS：ANN，BP,people-countingpretreatment,imagesegmentationn中山大学硕士学位论文第l章绪论

8、第1章绪论1．1人工神经