基于选择性视觉注意机制的遥感图像机场检测-电路与系统专业论文.docxVIP

优秀毕业论文 精品参考文献资料 复 旦 复 旦 大 学 硕 士 学 位 论 文 A b s tr a c t T arget d etectio n an d reco gn ition in rem ote sen sin g im ages b ecom es a h ot top ic in th e research area o f autom atic target detection recently. A m on g th ese, airp o rt d etection attracts lots o f attention b ecau se of its app lication s an d im p ortan ce in m ilitary an d civ il av iatio n field s. U n fortu n ately, th e com p licated b ack ground aroun d airports brin gs m u ch diffi culty into the detection . T radition al m eth ods on airp ort d etection alw ay s n eed to an alyze im ag es p ix el b y p ix el,w h ich is tim e con sum in g du e to th e large size o f rem ote sen sin g im ages.H ow ev er, as w e kn ow n , h u m an b ein g can an aly ze an d process th e inform ation in h is ey esh ot rapidly and selectiv ely, w h ich is relativ e to the ex isten ce of visu al attention m echan ism . S o th is p ap er aim s to in trod u ce th e v isu al atten tion m ech an ism in to th e airp ort detection an d p ro p o ses a n ew schem e on airp ort d etection an d reco gn ition w hich is based on v isual attention . D ifferent k in d s o f im ag e reso u rces are co n cern ed h ere an d th e m ain inn ov ation s in this p aper can b e d escrib ed as follow s: 1. A ccord in g to th e ch aracteristics o f airp o rts in p an ch rom atic b an d ，w e im prov ed the grap h -b ased v isu al salien cy (G B V S ) m od el b y addin g H o ugh tran sform an d p rior kn ow led ge o f inten sity, w h ich can ex tract region s of in terest (R O I) b etter. A fter th at, th e scale-in v arian t featu re tran sform (S IF T ) featu res are extracted on each R O I and h ierarchical d iscrim in an t regression (H D R ) tree is app lied as a classifier. F in ally th e airp ort is located b y th e classification resu lt o f S IF T featu res.E x p erim ental resu lts sh ow th at th e p rop osed m eth o d is faster an d m o re accurate th an ex istin g m