【摘要】 合成孔径雷达遥感具备全天时、全天候的观测能力,是多时相数据获取的有效保证。以福建省漳浦县为研究区,利用ALOS PALSAR双极化数据开展土地覆盖识别研究。首先基于多时相的强度数据构建时相稳定性指数,基于重复轨道干涉数据的相位信息计算相干性,以此分析和描述该地区典型地物的雷达数据时相特征。然后以典型地物的时相特征为基础,构建决策树分类器,进行土地覆盖识别。最后以实地考察数据、ALOS AVNIR-2影像和Google Earth影像为参考,进行分类结果的精度评价,总体精度达到81.43%,比利用不同时期的后向散射强度图像为输入波段的最大似然法的分类精度(总体精度为63.06%)高出很多。结果表明:在分类中有效融合时相信息,可以充分提高地物的可分性。更多还原

【Abstract】 Synthetic aperture radar can observe the earth all day and all weather,which can guarantee multitemporal acquisition of satellite data.Taking Zhangpu Country of Fujian Province as our test site,this paper used ALOS PALSAR dual-polarization data to identify land covers.To characterize the multi-temporal behaviour of the local land covers and introduced two indexes:①multi-temporal stability index from the multi-temporal backscatter intensities and②interferometric coherence from the phase information of the repeat pass data.With two temporal indexes as the primary source for classification,this study built a decision tree for land cover classification for our test site.We validated the produced land cover map of Zhangpu country with field inventory data,ALOS AVNIR-2image and Google Earth image.In addition,paper compared our map with that from the original backscatter intensity images with the maximum likelihood classifier.The comparison shows that our method produces higher overall accuracy(81.43%)than the maximum likehood classifier(63.06%).The result shows that the effective integration of the multi-temporal parameters of ALOS PALSAR data can improve the accuracy of the identification of land covers.更多还原