扎龙湿地龙泡子水深遥感反演

发布时间：2018-01-24 12:52

本文关键词： 泡沼水深遥感反演扎龙湿地高光谱　出处：《吉林大学》2013年硕士论文　论文类型：学位论文

【摘要】：海洋、河流、湖泊的水深遥感研究已有丰富的成果，但对于湿地泡沼的水深遥感研究并不是很多，湿地在维持生态平衡等多方面发挥着重要作用。随着湿地受人类活动影响加剧，生态功能被破坏。水对湿地的形成、发展起到关键作用。为了保护湿地资源，准确掌握水深有重要意义。本文以扎龙湿地龙泡子为主要研究区，利用多光谱影像数据与实测水深建立多光谱模型，利用野外实测高光谱数据与实测水深数据建立高光谱模型，两类不同模型对比分析，尝试利用遥感手段快速反演扎龙湿地湖泡水深，建立扎龙湿地湖泡水深反演模型，实现湿地水位的动态监测。主要工作及成果包括： 1．通过对扎龙湿地龙泡子研究区野外实地观测，GPS定点获取了共84个观测点的水深数据，并同步测得26个点的高光谱数据。其中58个观测点的水深数据用于与遥感影像分析，其他26个观测点的水深数据与实测高光谱数据进行分析。 2．利用高分辨率的QuickBird遥感影像与野外实测水深数据建立多光谱模型，根据实测水深值和对应影像各波段辐射亮度值以及波段组合的相关性，选取水深反演因子，建立多光谱水深反演模型，，经过验证和精度分析，将较好的4个模型反演龙泡子的水深。其中线性模型反演效果相对较好，能够体现出龙泡子中香蒲周边水深渐变趋势。 3．以扎龙湿地内10个湖泡建立感兴趣区，利用多光谱水深遥感模型对Landsat卫星的1988年TM影像、1999年TM影像、2007年ETM影像共三期影像数据反演湖泡水深，单波段模型反演结果差，单一波段建立的模型很难适用于区域水深反演。由四个波段共同建立的多元线性模型效果较好，水深变化趋势较明显。 4．利用野外实测水深与同步测量的高光谱数据建立高光谱水深反演模型。与多光谱遥感数据相比，高光谱遥感数据具有波段多、光谱分辨率高等特点，有效地捕捉到复杂多变的水体光谱特征的细微变化。经过一阶微分处理后，水体光谱反射率与水深的相关性明显提高，建立单波段模型和多波段模型，模型拟合度R2最高为0.860，最佳波段为4个。
[Abstract]:There have been rich achievements in remote sensing of water depth in ocean, river and lake, but there are not many researches on water depth of marsh. Wetland plays an important role in maintaining ecological balance. As wetland is affected by human activities, ecological function is destroyed. Water plays a key role in the formation and development of wetland, in order to protect wetland resources. It is very important to grasp the water depth accurately. In this paper, the multi-spectral model is established by using the multi-spectral image data and the measured water depth, taking Zhalong Wetland Dragon Paizi as the main research area. The hyperspectral model was established by using the field measured hyperspectral data and the measured water depth data. The two different models were compared and analyzed, and the rapid inversion of the bubble depth of Zhalong Wetland Lake was attempted by remote sensing. The inversion model of bubble depth in Zhalong Wetland Lake is established to realize the dynamic monitoring of wetland water level. The main work and results are as follows: 1. The data of water depth of 84 observation points were obtained by GPS in Zhalong Wetland Longpaozi Research area. The hyperspectral data of 26 points were measured synchronously, 58 of which were used for remote sensing image analysis, and the other 26 points were used to analyze the water depth data and the measured hyperspectral data. 2. Using high-resolution QuickBird remote sensing images and field measured water depth data to establish a multi-spectral model. According to the correlation between the measured water depth value, the radiance value of the corresponding image and the combination of the bands, the inversion factor of water depth is selected, and a multi-spectral inversion model of water depth is established, which is verified and analyzed accurately. The better four models are used to invert the water depth of the dragon bubble, and the inversion effect of the linear model is relatively good, which can reflect the trend of the gradual change of the water depth around the cattail in the dragon bubble. 3. Ten lake bubbles in Zhalong Wetland were used to establish the region of interest. The TM image of Landsat satellite in 1988 and 1999 was obtained by using multispectral water depth remote sensing model. In 2007, the depth of lake bubble was inversed by three periods of ETM image data, but the inversion result of single band model was poor. The model established in single band is difficult to be applied to regional water depth inversion. The multivariate linear model established by the four bands is effective and the variation trend of water depth is obvious. 4. The hyperspectral water depth inversion model is established by using the hyperspectral data measured in the field and synchronously measured. Compared with the multispectral remote sensing data, the hyperspectral remote sensing data have many bands and high spectral resolution. After the first order differential treatment, the correlation between the spectral reflectivity and the water depth is improved obviously, and the single-band model and the multi-band model are established. The model fitting R2 is 0.860 and the best band is 4.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P237

【参考文献】