高空间分辨率遥感影像分类讲述.pdfVIP

11 2 V o l. 11, N o. 2 2007 3 JOURNA L OF REMOTE SENSIN G M ar. , 2007 : 1007- 4619 2007) 02- 0193-08 黄 昕, 张良培, 李平湘 , 430079) : ) , , , , , , : ; ; ; ; : TP751. 1 : A C lassification of H igh Spatial R esolu tion R em otely Sen sed Im agery B ased on the Fusion of Sp ectral and Shape F eatures HUANG X in, ZHANG L iang-pei, LI P ing-x iang (N ationalKey Laboratoryf or Inf ormation Engineering in Surveying, M app ing and Remote Sensing, Wuhan University, ubeiWuhan 430079, China) Abstract: C lassificat ion of m u ltisp ectral d ata w ith h igh spatial reso lu tion is investigated. H igh spatial resolut ion remote ly sen sed HSRRS) mi agesw ith mu ltispectral bands such as Q u ickB ird and IKONOS p rovide a large amount of inform ation, thus op en ing up avenues for n ew rem ote sen sing app licat ion s. H ow ever, th eir ava ilab ility poses challenges to mi age classificat ion. Due to the comp lex spatial arrangem ent and spectral h eterogeneity even w ith in the sam e class, conventional spectra l classification m ethods are gross ly inadequate for HSRRS mi agery. In order to overcome th is inadequacy, spectral features must evidently be comp lemented by one or the oth er m eans. Shape is an mi portant feature of h igh spatial resolut ion remotely sensed HSRRS) mi agery, and it is the m an ifestation of textu res on such m