基于Gabor小波变换和矩变换的神经网络图像识别技术-信息与通信工程专业论文.docxVIP

ABSTRACTABSTRACT ABSTRACT ABSTRACT witll tlle developIIlem of comput盯tecllIlolo百韶锄d h曲an se娜ation rese删les， tlle tectlnology of in：lage rccogIlition bec锄e∞e sigtlific觚t丘eld Of image processiIl吕 It c雒be applied in almost all h啪an 1iV铭．B毗it is 1iIIlited in its sh嘶c0Inin擎of lackiIlg adaptation锄d not being uscd i11 dif衔ent fidd．伽s，how eIlh蛆ce its 却啪on，robus恤ess，eVen uIliVersality is a p】robleⅡ1． on me b勰is of enll盟cing ad印tatioIl，robl塔协eSs，even l|11iversality of image recogni矗on tec}lnology，tllis dissertation r髂e缸ched tlle image recogIlition tec：11nology． According to the phas锶of image reco班tion“res翩rched and discIlsscd iIl tllrce sections mch嬲，image preprocess协岛fbature ex仃action azld pattenl recognido玑 h the phase of image pmc铭sitl岛tllis dissenadon in虹oduced，disc船sed a11d s缸ulated t11e伽i廿onal metllods of iInage resto硎on，image alh觚cemalt趾d image segmeIltatioIL Funh黝orc，it proposcd a n鲫ad印dve蝴old mediaIl filta’to moVe multiple-implllse noise．More0Ve‘it proc∞s嚣bett盱ability of atterIuating multi—layef。d implllse noise than other menlods． h me phase of fba：tI】re％traction，this dissenation pr叩osed to use Gabor wave_I耽 仃觚sfo胁to gct global featllre of image and use mom曲t仃ansform to get local featllre． Thell，combining也e酉obal fean】re and the 10cal feahIre will觚e iIlto n洲fban|re to be proceSsed iIl t11e pa仕锄rccognition ph勰e． hl the ph勰e of patt锄recogni廿on，this dissertation adopted枷丘dal neural nct、)l，ork siInul劬g the s饥lctIl】呤and f11nction of tIu吼蛆neural ne脚ork to judge aIld 酗son．n took the fIlsmg f宅ature to be tlle inpm ofn删network to train and rccognize． k略t，me new tcchn0109y is completcd．The exp硎mem fe龇lts show mat it ow璐better adaptatioIl，robus恤ess，eVen嘶versality tllan neural net、)l，01．k inlage recogIlition b嬲ed onn坞Gabofwavelet仃ansft舢(坷themoment乜彻sfb衄． Keywords：image reco倒垃on，adap6Ve tllresh01d mediaIl fih％Gabor wavelet 仃锄fo咖，moment n孤sfo皿，n删ne帆rork Ⅱ 图目录图目录 图目录 图目录 图2—1 图像识别的主要步骤 4 图3一l 图像退化的模型 9 图3—2一个高斯噪声的概率密度函数 。10 图3—3一个均匀噪声的概率密度函数 。10 图3—4一个脉冲噪声的概率密度函数 ： ．．11 图3—5三种