基于非下采樣shearlet梯度方向直方圖和稀疏表示的脫機(jī)手寫(xiě)數(shù)字識(shí)別

《基于非下采樣shearlet梯度方向直方圖和稀疏表示的脫機(jī)手寫(xiě)數(shù)字識(shí)別》由會(huì)員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。

1、基于非下采樣Shearlet梯度方向直方圖和稀疏表示的脫機(jī)手寫(xiě)數(shù)字識(shí)別摘要：為了更有效的提高脫機(jī)手寫(xiě)體數(shù)字識(shí)別的性能和識(shí)別率，提山基于非下采樣shearlet梯度方向直方圖特征和稀疏表示的脫機(jī)手寫(xiě)字符識(shí)別方法。首先對(duì)字符圖像進(jìn)行非下采樣Shearlet變換，得到低頻子帶圖像和高頻子帶圖像，然后將子圖劃分為若干矩形子塊，分別統(tǒng)計(jì)子塊區(qū)域的梯度方向直方圖分布，將分塊直方圖串接起來(lái)形成非下采樣梯度方向直方閣特征（HNSCOG）,最后用HNSCOH特征構(gòu)建超完備字典，通過(guò)稀疏表示重構(gòu)最小誤差實(shí)現(xiàn)字符圖像分類(lèi)。在MN

2、IST和USPS數(shù)據(jù)集上測(cè)試，與DDH+SVM方法、SparseLS-SVM方法、sub-sampling+SVM方法和MTC+linear-SVM方法的識(shí)別率比較，實(shí)驗(yàn)結(jié)果表明，HNSCOG和稀疏表示的方法可以較大地提高脫機(jī)手寫(xiě)數(shù)字的識(shí)別率。I詞：脫機(jī)手寫(xiě)數(shù)字識(shí)別，非下采樣Shearlet變換，梯度方向直方圖，稀疏表示Offlinehandwrittendigitcharacterrecognitionbasedonhistogramsofnonsubsampledshearletorientedgrad

3、ientsandsparserepresentationAbstract:Inordertoimproverecognitionrateofhandwrittendigitcharacterefficiently，anovelmethodofhandwrittennumeralscharacterrecognitionbasedonhistogramsofnonsubsampledshearletorientedgradientsfeatures(HNSCOG)andsparserepresentation

4、waspresentedinthispaper.Firstly，thenonsubsampledshearlettransform(NSST)wasutilizedtodecomposethecharacterimagesonvariousscalesandindifferentdirections，andthelowfrequencysub-bandandbandpasssub-bandcoefficientswereobtained.Then，sub-imagesofdigitalcharacterw

5、asdividedintogridsofblocksandcellstoextractHOGfeatures，andthehistogramofeachunitwascomputedandconcatenatedasHNSCOGfeaturesdescriptor.Finally，HNSCOGfeatureswerecombinedtoformanover-completedictionarywhichwasemployedbysparserepresentationtoclassifythehand

6、writtennumeralscharacter.Tocomparetheperformanceoftheproposedmethod，severalalternativealgorithmsforhandwrittennumberrecognitionhadbeenconsidered,forinstancedistancedistributionhistogram(DDH)plusSVMmethod，SparseLS-SVMmethod,sub-samplingplusSVMmethodandMTCp

7、luslinear-SVMmethod.Inordertoevaluatethesetechniques,acollectionofwellknownstandarddatabaseshadbeenused:MNISTdigitdatasetandUSPSdigitdataset.Theexperimentalresultsindicatethatthehandwrittennumeralcharacterrecognitionaccuracycanbeimprovedgreatly.Keywords:ha

8、ndwrittencharacterrecognition，nonsubsampledshearlettransform，histogramsoforientedgradients，sparserepresentation1引言脫機(jī)手寫(xiě)數(shù)字識(shí)別是光學(xué)字符識(shí)別領(lǐng)域一個(gè)具有挑戰(zhàn)性的難題，有廣闊的實(shí)際應(yīng)用，如郵政信件分揀、表單數(shù)據(jù)處理等。聯(lián)機(jī)手寫(xiě)數(shù)字識(shí)別已經(jīng)取得較好的效果，然而脫機(jī)手寫(xiě)數(shù)字識(shí)別的精度離工程實(shí)踐應(yīng)川還存在