【摘要】 以扎龙湿地为研究区,利用空间分辨率高、纹理信息丰富、定位精度高和购买价格相对低廉的ALOS遥感影像,测定了研究区水体等5种典型地物和云阴影的波谱亮度值;根据水体和云阴影的光谱特征,利用阈值法对近红外波段ALOS影像进行阈值处理,将包含云阴影信息的水体信息提取出来;基于ALOS彩色影像,再采用决策树方法自动提取湿地水体信息,同时将云阴影信息去除,并再补充提取因去除云阴影而损失的水体信息;与目视解译结果相比,用本研究方法得到的研究区湿地水体信息精度高于85%。近红外波段阈值法和彩色影像决策树法相结合的方法是一种能有效去除云阴影而高精度获取湿地水体信息的方法。更多还原

【Abstract】 Lots of applying researches indicated that cloud shadows make a serious confusion on the extraction of water bodies from different kinds of remote sensing images when macro - monitoring water bodies in wetlands,especially during the flood inundation period. Zhalong wetlands taken as a study area is located in the western part of Songnen Plain,the downstream of Wuyuer River,has the most complete,the most primitive and the most open wetland ecosys- tem in the same latitude regions in Northeast China,which provide habitats for the Grus japonensis and many other wa- ter birds. ALOS remote sensing images with high spatial resolution,positioning precision,rich texture information and lower price,are used to extract water bodies of the study area in this paper. After comparing the spectral characteris- tics of water bodies with other 5 objects of the study area,it demonstrated that most cloud shadows were very similar to water bodies in spectral characteristics (except near infrared band). So the main question of this article was to find effective method to remove the cloud shadows. Firstly,threshold method was used to do threshold treatment in near infrared ALOS images,in which the water bodies with cloud shadows were extracted. Secondly,by further exploring spectral and spatial characteristics (such as spatially symmetrical distribution) of water bodies and cloud shadows,de- cision tree method based on color image was used to automatically extract the water bodies in wetlands. During this treatment process,cloud shadows were removed and the parts of water bodies lost in the removal of cloud shadows were added. The accuracy of extracting water bodies in wetlands of the study area was higher than 85% by comparing the water body area automatically extracted with visually interpreted. The results showed that the combination of threshold method based on single band (near infrared) and decision tree method based on color image could effective- ly remove cloud shadows and precisely obtain water bodies in wetlands. It also demonstrated that it is feasible to ex- tract water bodies on a large scale with high spatial resolution ALOS images. Furthermore,decision tree method has good portability.更多还原