一种新的发光二极管自动精确计数方法-计算机应用技术专业论文.docx

复H大卞硕士学位论文摘要平均准确率为99% 以上。实验结果表明，该检测方法可实现对LED管芯的实时检测，计数速度和精度均达到企业要求。使用计算机对LED芯片图像进行处理和计数，这种方法可以大大提高工作效率，提高准确度，具有较高的实用价值和应用前景。关键词：发光二极管计数；均值滤波；过检测策略；局部区域描绘子；k-m eans聚类。乂旦人学硕士，位论文A bstractAbstractLight-emittingdiode(LED),withitsadvantagesoflonglifeandlow powerconsumption,hasbeenwidelyusedintheareasofinstructions,displayandsoon.However,artificialdiecountingisslowwithlowcalculatingaccuracyandhighlaborintensityandsensorbaseddiecountinghasthedisadvantagesofhighequipmentcostand maintenancedifficulties.How todesign a LED diecounting method inthepremiseofensuringlowcost,fastresponse,accuratecountandhighrobustness,isatoughproblem needstobesolvedimmediatelyinthisindustry.ThespecificprocessofLED diecounting bycomputervisiontechnologyreferstocompletingdiecountingwiththeuseofindustrialcamerastogettheLED chippicturesintopview andvisionalgorithmstoprocessandanalyzetheimages.Thecomm onlyusedprocessingmethodsoftheLED diecountingareasfollows:Blobanalysisandsub-pixellocationalgorithm ,thegraphicalcontourrecognitionalgorithmandsub-pixellinearedgedetectionalgorithm.Thesethreealgorithms,calculatingthenumberofthediesbythebroadcontoursofLED dies,maketheidentificationbetterandtheefficiencyhigherinthecaseofaclearoutlineandobviousdistinctionoftheLED dies.W hentakingphotosoftheLED chipsbyusingactualindustrialdigitalcameras,lightimpact,cameraexposureorotherfactorswillmaketheverticalviewoftheLED chipshavelocalfuzzy,connectedordeformedregions,leadingtobigerrorswhenrecognizingtheLED dieswiththealgorithmsmentionedaboveandmakingtheprecisecountingdifficult.Thispaperpresentsareal-timeandaccuratemethodofLED diecounting.First,noiseintheimagesisremoved bymeanfilter;then,allthepossiblelocationsoftheLED dieintheLED chippicturesintopview isextractedbyusingaover-detectionstrategy;after that,localregiondescriptorsareextractedbythelocationofthegrayvalueoflocalareatodistinguishbetweentheLED dieandim ageinterferencesignal;finally,k-meansclustering algorithm isusedtoselectthetargetgroup.Experimentalresults