Automatic Fruit Recognition Based on DCNN for Commercial Source Trace System

《Automatic Fruit Recognition Based on DCNN for Commercial Source Trace System》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、硕士留学生学位论文AutomaticFruitRecognitionBasedonDCNNforCommercialSourceTraceSystem作者姓名HussainIsrar学科专业信息与通信工程指导教师贺前华所在学院电子信息与通信工程学院论文提交日期2018年05月日AutomaticFruitRecognitionBasedonDCNNforCommercialSourceTraceSystemADissertationSubmittedfortheDegreeofMasterCandidat

2、e:HussainIsrarSupervisor:Prof.HeQianHua贺前华SouthChinaUniversityofTechnologyGuangzhou,ChinaABSTRACTAutomaticfruitrecognition-basedonmachinevisionisconsideredaschallengingtaskduetosimilaritiesbetweenvarioustypesoffruitsandexternalenvironmentalchangese-glight

3、ing.Fruitandvegetableclassificationisoneofthemajorapplicationsthatcanbeutilizedinsupermarketandfruitshopstoautomaticallydetectandrecognizethekindoffruitsandvegetablespurchasedbycustomersandtodetermineitsprices.Althoughautomaticfruitrecognitionisgettingmor

4、eandmoreimportant,thistechnologyisstillfarfrombeingmature.Themainworkandinnovationsofthisthesisareasfollows:(1)AfruitrecognitionalgorithmbasedondeepconvolutionneuralNetwork(DCNN)isproposed.Mostoftheprevioustechniqueshavesomelimitationsbecausetheywereexami

5、nedandevaluatedunderlimiteddataset,furthermoretheyhavenotconsideredexternalenvironmentalchanges.Weevaluatedourmodelinmuchmorecomplicateddatasetwhichmostlymeetallreal-worldchallengestomakeourmodelrobust.(2)Weestablishedfruitimagesdatabasespanning15differen

6、tcategorieswhichcompriseof44406imagescollectedinourlabenvironmentduringaperiodof6monthsunderdifferentreal-worldconditions&keepinginviewthelimitationsofexistingdataset.(3)WealsoproposeanalgorithmforIntra-classrecognitionoffruitsusingDCNN(deepconvolutionaln

7、euralnetwork).Theresultsofcarryingouttheseexperimentsdemonstratethattheproposedapproachcouldbeautomaticallyrecognizethefruitwithaccuracyof99%.TheInputimagesweredirectlyinputtoDCNNfortrainingandrecognitionwithoutfeaturesextraction&theDCNNlearnoptimalfeatur

8、esfromimagesadaptively.Thefinaldecisionwasmadebasedonafusionofallregionsclassificationusingprobabilitymechanism.Keyword:FruitRecognition;Deeplearning;DeepConvolutionalNeuralNetworkITableofContentABSTRACT............