基于GF-1卫星遥感的冰川边界识别

高分一号（GF-1）卫星的成功发射开启了国产高分辨率对地观测的新时代， 为探讨国产高分卫星在冰川边界识别中的有效性， 在缺少短波红外和热红外波段的情况下， 将GF-1影像、 建立的波段比值Band1/Band4， 数字高程模型和坡度相结合， 采用面向对象的分析方法， 经过反复试验， 确定影像分割和合并尺度， 进而确定冰川边界的知识规则， 最终实现冰川边界的识别。以研究区第二次冰川编目数据集作为参考数据， 采用混淆矩阵的方法对识别结果进行验证， 总体精度和Kappa系数为90.05%和0.79。同时将识别结果与人工修订冰川边界进行对比， 可以发现除少量冰舌末端冰川外， 该方法可以有效地对冰川进行识别。建立的知识规则显示仅仅利用蓝色波段和DEM就可以有效地提取裸冰区， 波段比值、 坡度和纹理特征更有助于冰舌的提取。该研究表明GF-1卫星数据可以有效识别冰川范围， 为冰川研究提供可靠的数据和研究基础。

Glacier mapping based on GF-1 satellite remote sensing

GF-1 satellite’s successful launch opens a high resolution satellite earth observation. To discuss the effectiveness based on GF-1 satellite in the absence of short-wave infrared and thermal infrared bands in mapping glacier， in this study， glaciers are mapped based on the object-oriented method using GF-1 images， band ratio of Band 1 and Band 4， digital elevation model and slope information. The segmentation and merge levels and rules of glacier mapping were defined by trial and error， then the boundary of glacier was obtained. Confusion matrix was used to test and verify the results based on the Second Glacier Inventory Dataset of China （Version 1.0）， with overall accuracy and Kappa confidence of 90.05% and 0.79. Meanwhile， comparing this result with the result by visual revision showed that this approach could distinguish glacier well， except for the glacier terminal， the rule for glacier indicated blue band and DEM could distinguish ice from its surrounding， and the band ratio， slope and texture characteristic were helpful for extracting glacier terminal. This study showed that GF-1 satellite could map the boundary of glacier effectively， and could provide reliable data and research foundation for glacier research.