
面向对象-改进冰雪指数法消除冰湖干扰提取冰川边界的优越性分析——以各拉丹冬冰川为例
杨佳, 薛莎莎, 苏永恒, 任庆福
冰川冻土 ›› 2022, Vol. 44 ›› Issue (5) : 1665-1673.
面向对象-改进冰雪指数法消除冰湖干扰提取冰川边界的优越性分析——以各拉丹冬冰川为例
Analysis of superiority of the object-oriented-improved ice and snow index method to eliminate glacial lake interference and extract glacier boundaries: taking Geladandong Glacier as an example
冰川作为气候变化的重要指标器,其范围监测对区域生态环境以及人类社会生活具有重要意义。目前基于遥感技术的冰川范围监测应用广泛,然而传统遥感监测方法中冰雪指数阈值法的提取结果存在无法区分冰川与冰湖的现象,面向对象分类法受地物光谱纹理信息限制无法避免同谱异物现象的出现。为弥补上述不足,提出一种可区分冰湖与冰川的改进冰雪指数,并将其融入面向对象分类法中,构建了一种面向对象-改进冰雪指数法。将各拉丹冬冰川作为试验区(该地区冰川表面洁净),运用面向对象-改进冰雪指数法识别冰川边界,使用青藏高原冰川数据产品及常规遥感监测方法的提取结果作为参考数据,以验证此方法的有效性和稳健性。结果表明:面向对象-改进冰雪指数法综合了改进冰雪指数阈值法以及面向对象分类法的优点,冰川范围提取精度高达97.26%,与冰雪指数阈值提取结果相比精度提高了0.12%,与面向对象分类法提取结果相比精度提高了0.38%。此方法不仅有效地解决了地物错分的问题,还实现了冰川边界的精确识别。
Glaciers as important indicators of climate change, the melting of glaciers will not only cause sea level rise, but also cause disasters such as ice avalanches and glacial lake break. Glacier extent monitoring is important to the regional ecological environment and human social life. Glacier extent monitoring based on remote sensing technology is widely used. However, in the conventional remote sensing monitoring method, there is a phenomenon that glaciers and glacial lakes are indistinguishable in the extraction results of the ice and snow index threshold method. The object-oriented classification method is limited by the spectral texture information of ground objects, and the occurrence of different types of land cover with the same spectrum’ phenomena will appear. In order to make up for the above shortcomings, in this paper, improved ice and snow index that can distinguish glacial lakes and glaciers is proposed, and it is integrated into the object-oriented classification method, and an object-oriented-improved ice and snow index method is constructed. Taking the Geladandong Glacier as the test area (the surface of the glacier in this area is clean), the extraction results of this method are compared with the results of the Qinghai-Tibet Plateau glacier data products and the results of conventional remote sensing monitoring methods. Validation of the object-oriented-improved ice and snow index method for effectiveness and robustness. The results demonstrate that: Ice and snow have strong reflection characteristics in the blue band and strong absorption in the near-infrared long-wave band. These two bands are sensitive bands for glacier identification. They not only have a high degree of recognition for glaciers, but also there are also obvious effects in distinguishing glaciers and glacial lakes. The band characteristics of the Landsat-8 OLI sensor were analyzed, and the glacier sensitive bands were Coastal band and SWIR1 band. Three improved ice and snow indices RSI, NDSI* and DSI are proposed for the glacier-sensitive band of Landsat-8 data. Taking the Geladandong Glacier as the test area, the glacier extent was extracted based on three improved ice and snow indices, and the extraction results were compared with the 2017 glacier data on the Qinghai-Tibet Plateau. The results show that the three ice and snow indices can extract the glacier boundary well and the extraction accuracy can reach more than 95%. In the recognition of confusing ground objects, the three indices can distinguish clouds and glaciers, but in the difference between glaciers and glacial lakes, the DSI ice and snow index is significantly better than others. The object-oriented-improved ice and snow index method was used to extract the extent of Geladandong Glacier, and the overall accuracy of the extraction results was as high as 97.26%. This method combines the advantages of the ice and snow index threshold extraction method and the object-oriented classification method to make up their respective shortcomings, and identify the glacier boundary more accurately. The object-oriented-improved ice and snow index method solves the problem of the same spectrum foreign matter in the process of glacier extraction to a certain extent, especially for the distinction between glacial lakes and glaciers. However, there are still some deficiencies. The original data of this study are images at the end of glacier ablation and with a high solar elevation angle. This study did not consider in detail the indistinguishability of snow covered in glacial and non-glacial areas and the difficulty of identifying glaciers in shaded areas. In the follow-up, in view of the above shortcomings, it is proposed to use multi-source remote sensing data (high resolution images, DEM, etc.), integrate multiple glacier characteristic indicators, and further improve the object-oriented-improved ice and snow index method, and strive to fundamentally solve the difficult problem of clean glacier classification.
冰川 / 遥感 / 面向对象-改进冰雪指数法 / 各拉丹冬 {{custom_keyword}} /
glacier / remote sensing / object-oriented-improved ice and snow index method / Geladandong {{custom_keyword}} /
表1 改进的冰雪指数Table 1 The improved ice and snow indexes |
序号 | 名称 | 公式 |
---|---|---|
1 | 比值冰雪指数(ratio | |
2 | 差值冰雪指数(difference | |
3 | 归一化冰雪指数(normalized differential snow index, NDSI*) | |
表2 冰雪指数提取结果的精度Table 2 Accuracy of extraction results based on ice and snow indexes |
指数 | 阈值 | 漏分误差 | 总体精度 |
---|---|---|---|
RSI | RSI>9.172 | 3.91% | 96.09% |
DSI | DSI>2 207 | 3.14% | 96.86% |
NDSI* | NDSI*>0.860 | 4.71% | 95.29% |
表3 分割合并算法的阈值组合Table 3 Threshold combination of segmentation and merge algorithm |
分割合并算法 | 组合1 | 组合2 | 组合3 | 组合4 |
---|---|---|---|---|
Edge | 30 | 45 | 60 | 70 |
Full Lambda Schedule | 45 | 60 | 75 | 98 |
表4 属性阈值范围Table 4 Attribute threshold range |
特征影像 | 规则 | 属性 | 阈值 |
---|---|---|---|
多光谱影像数据集 | Texture | Texture Range | 871~11 035 |
DSI冰雪指数波段 | Spectral | Spectral Mean | 2 207~12 250 |
表5 冰川提取结果的精度Table 5 Accuracy of glacier extraction results |
方法 | 漏分误差 | 总体精度 |
---|---|---|
NDSI阈值法 | 2.86% | 97.14% |
面向对象分类法 | 3.12% | 96.88% |
面向对象-改进冰雪指数法 | 2.74% | 97.26% |
1 |
World Meteorological Organization. State of the global climate 2020[EB/OL]. [2021-04-19].
{{custom_citation.content}}
{{custom_citation.annotation}}
|
2 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
3 |
方金鑫, 蒲焘, 史晓宜, 等. 气候变化背景下玉龙雪山漾弓江流域径流变化及其影响因素分析[J]. 冰川冻土, 2019, 41(2): 16-22.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
4 |
张建国, 刘玉梅, 吕腾腾, 等. 各拉丹冬区域冰川变化研究[J]. 干旱区资源与环境, 2015, 29(12): 184-189.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
5 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
6 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
7 |
吴淼, 韩用顺, 张东水, 等. 表碛覆盖冰川信息提取方法: 以波密县为例[J]. 山地学报, 2017, 35(2): 238-245.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
8 |
周远刚, 赵锐锋, 张丽华, 等. 博格达峰地区冰川和积雪变化遥感监测及影响因素分析[J]. 干旱区地理, 2019, 42(6): 1395-1403.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
9 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
10 |
季漩, 陈云芳, 罗贤, 等. Landsat 8 OLI影像的高原山地阴影区冰川识别方法[J]. 光谱学与光谱分析, 2018, 38(12): 3857-3863.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
11 |
刘凯, 王宁练, 白晓华. 1993—2015年喀喇昆仑山努布拉流域冰川变化遥感监测[J]. 冰川冻土, 2017, 39(4): 710-719.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
12 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
13 |
彦立利, 高健峰, 王建, 等. 基于GF-1卫星遥感的冰川边界识别[J]. 冰川冻土, 2020, 42(4): 1400-1406.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
14 |
王忠武, 王志盼, 尤淑撑, 等. Landsat影像冰川提取的上下文感知语义分割网络法[J]. 测绘学报, 2020, 49(12): 1575-1582.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
15 |
张娟, 徐维新, 肖建设, 等. 近20年长江源头各拉丹冬冰川变化及其对气候变化的响应[J]. 干旱区资源与环境, 2014, 28(3): 142-147.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
16 |
蒙张, 胡勇, 邹洪坤, 等. 长江源各拉丹冬地区冰川变化遥感监测分析[J]. 人民长江, 2018, 49(4): 34-39.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
17 |
焦克勤, 沈永平. 唐古拉山地区第四纪冰川作用与冰川特征[J]. 冰川冻土, 2003, 25(1): 34-42.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
18 |
仲伟凡, 唐远彬, 殷腾箐. 基于纹理特征的SAR影像各拉丹冬冰川信息提取[J]. 测绘与空间地理信息, 2015, 38(10): 178-180.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
19 |
胡凡盛, 杨太保, 王晶, 等. 基于面向对象分类的马兰冰帽变化与气候响应[J]. 干旱区研究, 2017, 34(5): 1018-1026.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
20 |
李长龙. 面向对象的沙化土地GF-1遥感分类技术研究[D]. 北京: 中国林业科学研究院, 2015.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
21 |
王露, 刘庆元. 高分辨率遥感影像多尺度分割中最优尺度选取方法综述[J]. 测绘与空间地理信息, 2015, 38(3): 166-169.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
22 |
王婧, 赵天忠, 曾怡. 基于粗糙集规则提取的面向对象树种分类方法[J]. 遥感信息, 2013, 28(4): 90-97.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_ref.label}} |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
/
〈 |
|
〉 |