论文部分内容阅读
在总结混合像元分解方法的基础上,提出了一种模拟真实场景的像元分解方法,该方法首先通过真实场景的模拟获得各分量的丰度,结合遥感影像与场景模拟的丰度反演端元反射率(模拟端元),最后用带约束条件的线性模型进行混合像元分解。用浙江省安吉县毛竹林调查资料及LandsatTM对该方法进行验证和对比分析表明,基于模拟端元的混合像元分解结果比基于影像端元和参考端元的精度高且具有良好的稳健性。模拟真实场景的混合像元分解方法将样地调查数据的先验知识应用于端元提取,并将三维模拟模型引入到二维的线性光谱分解中,具有一定的优势和应用推广前景。
On the basis of summarizing the method of pixel decomposition, this paper proposes a pixel decomposition method to simulate real scene. First, the abundance of each component is obtained through the simulation of real scene. Combining the remote sensing image with the scene inversion Endmember reflectivity (simulated endmember), and finally mixed pixel decomposition using a linear model with constraints. The validation and comparative analysis of this method using Moso bamboo forest survey data and LandsatTM in Anji County, Zhejiang Province show that the mixed pixel decomposition results based on simulated endmember are more accurate and robust than those based on image endmember and reference endmember. The mixed pixel decomposition method which simulates the real scene applies the prior knowledge of the sample survey data to the endmember extraction and introduces the three-dimensional simulation model into the two-dimensional linear spectral decomposition, which has certain advantages and application prospects.