基于纹理特征的地表覆盖分类算法研究-模式识别与智能系统专业论文.docxVIP

优秀毕业论文 精品参考文献资料 研究生优秀毕业论文 AbstractⅥsion-baSed Abstract Ⅵsion-baSed enviromental perCeption system is one of research hotspots in the field of pattem recogmtion aIld矾ficialintelligence，and it is also one of the essential content of intelligent vehicle research．Surf．aCe classification is all imponant pan of the coment．The studV of tlle smlctural road is relatively mature which is due to the color，texture，road signs and other fea_tures are吼ified．Tlle current study of the u11stmctured road in o昏road environment is not enou曲，and the road surf．ace classification result is unsatis匆ing in bad enviro啪ent such as under direCt guIllight，rain and Snow renection of the sunlight．Intelli寥nt vehicle、)~rorks in outdoor environment wllich light，scenery，weather，seaSon and location are complex，uIllike in indoor enviromnent、vhich light，scene巧are relatiVely constant．Theref．ore， this anicle focuses on tlle unstmctured road surface claSsification in o昏road enVironment． The main coments of thjs p印er are diVided缸o f．our pans： (1)We inVestigate surface texture feacure e姗action methods，focusing on color fIeature， LBP f．eature，SIFT features and f．eature extraction process．We experiment on t11e texture d撕baSe谢m tlle combination of di虢rent mainstre锄f．eatures aIld classifiers． (2)We apply SVM to road sur：face classification．BaSed on the idea of raJldom forest ensemble leanling，l心烈．SVM is introduced in road sur：faCe recogIlition．In addition，due to the impommce of SVM kernel如nction，SVM based on hybrid kernel如nction is also in仃oduced．To classi囟road s州沁e in o昏road enViro姗ent，one of铆。一phaSe metllods waS adopted，inⅥ雠ch t11e first stage used l心N-SVM joiniIlg而xed kemel鼬1ction to claSsi母 伊ass or non．孕aSs．The experiments pr0Ved this way can get mgher accuraCy rate in this stage． (3)W色apply random forest t0 road su“犯e classification．BaSed on the idea of maximizing ma玛iIl SVM，ensemble claSsification ma玛in and raIldom forest baSed on claLssification margiIl are ir咖oduCed in road surf．ace recogllition．111 tlle