基于高分一号卫星遥感数据的青藏高原高寒湿地信息提取方法研究——以玛多县为例

Study on Information Extraction Method of Alpine Wetland in Qinghai-Xizang Plateau based on Remote Sensing Data of GF-1 Satellite——Taking Maduo County for Example

青藏高原蕴含着最典型、最多样的高寒湿地类型。为了探究高寒湿地的退化状况,首先需要对高寒湿地信息进行精细化提取。选用玛多县GF1-WFV生长季影像,利用先进的分层分类法,选择分割尺度为50对影像进行分割,然后利用单波段阈值法及坡度阈值对湿地进行判识,结合非生长季影像,综合利用波谱关系法、混合水体指数法、归一化差异水体指数法及单波段阈值法对玛多县的高寒湿地分类系统的Ⅲ级类别进行逐一提取,最后得到玛多县高寒湿地类型的地物信息及分布状况。研究表明:(1)随机选取检验样点,采用混淆矩阵的方法,对提取的高寒湿地分类影像进行精度评价,分类精度达到88.59%,Kappa系数为0.8637,精度通过检验。分类结果显示,结合了影像的纹理特征和光谱特征的分层分类技术在高寒湿地信息的提取方面更加优于其他传统方法,能够实现高寒湿地的精细化信息提取与分析。(2)玛多县高寒湿地面积呈东多西少,北多南少分布,其中永久性淡水湖的面积最大为1685.58 km^2,占玛多县总高寒湿地面积的69.05%;其次是草本沼泽和永久性河/溪,面积分别为495.56 km^2和94.81 km^2,占比分别为20.34%和3.88%;季节性咸水湖、季节性淡水湖、间歇性河/溪、洪泛湿地、泥潭沼泽、灌丛沼泽、内陆盐沼及冰川积雪的湿地类型面积在1.25~73.23 km^2,所占比例不足1%,其中面积最小的是季节性咸水湖和季节性淡水湖。(3)高分数据在青藏高原高寒湿地信息提取中具有不受高海拔影响分辨率的特点,因此具有更强的的可行性和应用性,尤其选择时间分辨率不同的生长季及非生长季的影像,在获取季节性高寒湿地类型信息方面具有更大的优势。(4)玛多县高寒湿地类型中,湖泊面积有增加的趋势,但河流和洪泛湿地面积有减少的趋势,如果不采取及时、有效的保护措施,玛多县的湿地有可能继续退化。

The Qinghai-Xizang Plateau contains the most typical and diverse types of alpine wetlands.In order to explore the degradation of alpine wetlands,it is necessary to extract accurate information of alpine wetlands.Research methods Selects growing season images of GF1-WFV in Maduo County and uses hierarchical classification method.Chose the 50-segmentation scale to segment the image,and then the wetlands are identified by means of single band threshold and slope threshold.Comprehensive uses of these methods,including spectral relation method,mixed water index method,normalized differential water index method and single band threshold method.Finally,the information and distribution status of the alpine wetland types in Maduo County,Qinghai Province were obtained,combines with non-growing season images,according to the class III classification standard of alpine wetland remote sensing classification system.Research results Firstly,the sample points were randomly selected and the accuracy of the image was evaluated by the method of confusion matrix.The classification accuracy of this study reaches 88.59%,and the Kappa coefficient is 0.8637,the classification accuracy passes the test.The classification results show that the hierarchical classification technique combining image texture features and spectral features is superior to other traditional methods in the extraction of alpine wetland information,and it is able to achieve Information extraction and analysis of alpine wetland refinement.Secondly,the area of alpine wetland in Maduo County is more in the east than in the west,more in the north and less in the south,and the main types of wetlands are concentrated in the central and northern regions.The largest area of the permanent freshwater lake is 1685.58 km^2,accounting for 69.05%of the total alpine wetland area of Maduo County,followed by herbaceous swamps and permanent rivers/streams,with an area of 495.56 km^2 and 94.81 km^2,accounting for 20.34%and 3.88%,respectively.The wetland area of seasonal saltwater lakes,seasonal freshwater lakes,intermittent rivers/streams,flooded wetlands,mud marshes,scrub swamps,inland salt marshes and glacial snow cover wetlands is within 1.25~73.23 km^2,accounting for less than 1%,of which seasonal saltwater lakes and seasonal freshwater lakes are the smallest.Thirdly,the high-score data is not affected by the high-altitude resolution in the information extraction of alpine wetland in Qinghai-Xizang Plateau,so it has more feasibility and application,especially choosing the image of growth season and non-growth season with different time resolution,which has more advantages in obtaining seasonal alpine wetland type information.Lastly,the classification result shows that among the types of alpine wetlands in Maduo County,the area of lakes tends to increase,but the area of rivers and flooded wetlands tends to decrease.If protective measures are not taken timely and effectively,the wetlands in Maduo County may continue to degrade.