基于神经网络的商场客流量统计系统研究

《基于神经网络的商场客流量统计系统研究》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、河北工业大学硕士学位论文基于神经网络的商场客流量统计系统研究姓名：周瑞英申请学位级别：硕士专业：模式识别与智能系统指导教师：顾军华20061101基于神经网络的商场客流量统计系统研究STUDYOFPEOPLE-COUNTINGSYSTEMINMARKETPLACEBASED-ONNEURALNETWORKABSTRACTAstheentryofmanyforeignretailenterprises,thecompetitionisfiercetoretaildaybyday.Customer-countingisveryimportant,becauseitisafoundationo

2、fretailinganditisproportionaltosalesamount.Thetraditionalcountingmethodartificiallycan’tofferthereal-timeflowofcustomerdata,whiletherateofaccuracy,whichofsimplecountingwithphotoelectrictransducerofinfraredray,islow.Soitisanimportantroutetoimprovetherateofaccuracyofinfrared-countingwithpatternreco

3、gnitiontechnologyofartificialintelligence.Thispaperdesignsanewpeople-countingsystembased-onneuralnetwork,AccordingtotheresearchapproachofpatternrecognitionsystemandtheCharacteristicofthedata,whichphotoelectrictransducerofinfraredraygets.Thissystemincludesfourphotoelectrictransducersofinfraredray,

4、whicharefixedintheentryofMarket,andtheheighttotheankle.Whenthecustomeristhroughcountingthearea,itwillproduceapattern.Andthepatternisdealtwithbytheneuralnetworkofintelligence.Itcountsthenumberofcustomersandsavesit.Thegroundworkisasfollows:Thispaperchoosesfourphotoelectrictransducersofinfraredray,i

5、nordertodistinguishcustomerswhichenteratthesametime.Andtheinitialdataispreprocessedforstrengtheningthevalidityofdata.Dataofcustomersiscontinuousspace-timesequence.Accordingtothischaracteristic,thispaperhasproposedanadaptedmethodofdatasegmentation.Andtheexperimentprovesthiskindofmethodiseffective.

6、Thispaperhasproposedanextractionmethodoffeatureparameterbasedonpulse-pulsesequence.Andtheexperimentindicatesthiskindofmethodiseffective.ThispaperchoosesresilientBlack-Propagations.Anditlearnsalotofsamplesandusedtocountthenumberofcustoms.Theexperimentindicatesthissystemisvalid.Itcandistinguishcust

7、omersbetweenwhomtherearesomespaces,andtherateofaccuracyattachesto100.Andforcustomerswhoentryatthesametime,italsohasahighrateofaccuracy.Besides,theexperimentprovesthemodelofcustomer-countingiscorrect.KEYWORDS:customerco