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\n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\n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Landsat TM 对该方法进行验证和对比分析表明, 基\n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\n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\n有一定的优势和应用推广前景。
This paper proposes a new unmixing method based on the simulation of real scenario. Fractions of the components\nare firstly obtained through the real scenario simulation. Then reflectance values of the endmembers (simulated endmembers)\nare calculated by combining the image reflectance values and corresponding simulated fractions. A constrained linear model is\nfinally used to unmix pixels based on the simulated endmembers. Comparative analysis of the different endmember extraction\nmethods, such as simulated endmembers, image endmembers, and reference endmembers, indicates that the simulated\nendmember method has the highest estimation accuracy and robustness for the crown closure of moso bamboo. The advantage\nof the real scenario simulation is to use field data as a priori knowledge for endmember extraction and introduce a\nthree-dimensional simulation model into a two-dimensional linear spectral decomposition.