Banana　is　one　of　the　main　economic　agrotypes　in　Zhangzhou,　Fujian　Province.　The　multi-temporal　ENVISAT　ASAR　data　with　different　polarization　are　used　to　classify　the　banana　fields　in　this　paper.　Principal　component　analysis　(PCA)　was　applied　for　six　pairs　of　ASAR　dual-polarization　data.　For　its　large　leaves,　banana　has　high　backscatter.　So　the　value　of　banana　fields　is　high　and　shows　very　bright　in　the　1st　component,　which　makes　it　much　easier　for　banana　fields　extraction.　Dual-polarization　data　provide　more　information,　and　the　VV　and　VH　backscatter　of　banana　show　different　characters　with　other　land　covers.　Based　on　the　analysis　of　the　radar　signature　of　banana　fields　and　other　land　covers　and　the　1st　component,　banana　fields　are　classified　using　object-oriented　classifier.　Compared　to　the　field　survey　data　and　ASTER　data,　the　accuracy　of　banana　fields　in　the　study　area　is　83.5%.　It　shows　that　the　principal　component　analysis　provides　the　useful　information　in　SAR　images　analysis　and　makes　the　extraction　of　banana　fields　easier.　©　2009　Wuhan　University　and　Springer-Verlag　GmbH.