資源描述:
《基于邊緣幾何特征的高性能模板匹配算法_吳曉軍》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在行業(yè)資料-天天文庫(kù)。
1、第34卷第7期儀器儀表學(xué)報(bào)Vol.34No.72013年7月ChineseJournalofScientif;%96%96icInstrumentJul.2013*基于邊緣幾何特征的高性能模板匹配算法1,21吳曉軍,鄒廣華(1.哈爾濱工業(yè)大學(xué)深圳研究生院深圳518055;2.深圳先進(jìn)運(yùn)動(dòng)控制技術(shù)與現(xiàn)代自動(dòng)化裝備重點(diǎn)實(shí)驗(yàn)室深圳518055)摘要:基于灰度相關(guān)的模板匹配算法在很多情況下難以得到準(zhǔn)確的匹配結(jié)果,提出一種基于邊緣幾何特征的高精度模板匹配算法。利用曲面擬合方法獲得邊緣的梯度方向和亞像素坐標(biāo)作為匹配信息,采用圖像金字塔
2、的搜索策略對(duì)算法加速,最后利用最小二乘平差理論得到亞像素級(jí)的定位精度和精確的旋轉(zhuǎn)角度信息。實(shí)驗(yàn)表明,對(duì)于目標(biāo)旋轉(zhuǎn)、均勻或非均勻變化的光照、部分遮擋的情況下可以得到良好的匹配結(jié)果,而且在保證高精度的同時(shí)算法可以滿足實(shí)時(shí)性要求,重復(fù)定位精度優(yōu)于商業(yè)化軟化包MIL8.0的GMF算法。關(guān)鍵詞:模板匹配;邊緣特征;曲面擬合;最小二乘平差中圖分類號(hào):TP391.41文獻(xiàn)標(biāo)識(shí)碼:A國(guó)家標(biāo)準(zhǔn)學(xué)科分類代碼:510.40Highperformancetemplatematchingalgorithmbasedonedgegeometricfe
3、atures1,21WuXiaojun,ZouGuanghua(1.ShenzhenGraduateSchool,HarbinInstituteofTechnology,Shenzhen518055,China;2.ShenzhenKeyLaboratoryforAdvancedMotionControlandModernAutomationEquipment,Shenzhen518055,China)Abstract:Gray-scalecorrelationbasedtemplatematchingalgorithmc
4、anhardlyobtainaccuratematchingresultsinsomeconditions,afastandhighprecisiontemplatematchingmethodbasedonedgegeometricfeaturesisproposed.Surfacefittingmethodisusedtoobtainthegradientdirectionsandsub-pixelcoordinatesoftheedgepoints,whichareusedasthematchinginformati
5、onincalculatingthesimilaritybetweentemplateandtarget.Inordertosatisfythereal-timerequirement,imagepyramidsearchingstrategyisemployedtoacceleratethealgorithm.Furthermore,theleastsquareadjustmenttheoryisadoptedtocalculatethesub-pixelpositioningprecisionandpreciserot
6、ationangleinformation.Experimentresultsdemonstratethatthealgorithmintroducedinthispapercanobtaingoodmatchingre-sultsinthecaseoftargetrotation,uniformornoon-uniformilluminationdisturbance,partialocclusion,andetc.Moreover,besidesthestability,reliabilityandhighprecis
7、ion,theproposedalgorithmalsocanmeetthereal-timerequirements.TherepeatpositioningprecisionofthealgorithmisbetterthanthatoftheGMFalgorithmofthecom-mercializedmachinevisionpackageofMIL8.0fromMatrox.Keywords:templatematching;edgefeature;surfacefitting;least-squareadju
8、stment[1-3]蹤等領(lǐng)域中得到廣泛應(yīng)用?;诨叶认嚓P(guān)(gray-scale1引言correlation)的模板匹配算法以圖像灰度信息直接進(jìn)行匹配,原理簡(jiǎn)單、實(shí)現(xiàn)容易,但其抗干擾性差,對(duì)于非均模板匹配算法是機(jī)器視覺(jué)(machinevision)技術(shù)重勻變化的光照和部分遮擋問(wèn)題束手無(wú)策?;趲缀翁卣饕?/p>