论文部分内容阅读
针对地物光谱的不确定性及参与每一像元光谱混合的端元种类与数目的不确定性,提出一种准确快速的高光谱数据光谱分析方法。该方法利用图像端元与标准端元库端元构建分组端元库,为每种地物引入多种端元光谱,解决了地物光谱不确定性问题;通过基于多端元交叉相关光谱匹配的确定搜索与基于二进制粒子群优化的随机搜索相结合的方式为每一像元搜寻一个最优端元子集,解决了参与混合的端元种类与数目的不确定性问题,并且提高了搜索效率。通过模拟数据与真实高光谱数据的试验验证,该方法解混精度与多端元光谱混合分析法相当、远高于全限制/无限制最小二乘法与交叉相关光谱匹配法,而计算时间远小于多端元光谱混合分析,实现了快速、准确的高光谱数据解混。
Aiming at the uncertainty of the object spectrum and the uncertainty of the types and numbers of the end elements involved in the spectrum mixing of each pixel, an accurate and fast hyperspectral data spectral analysis method is proposed. In this method, a grouping endmember library is constructed by using the image endmember and the standard endmember library, and a variety of endmember spectra are introduced for each feature, which solves the problem of the spectral uncertainty. By using multi-terminal cross correlation correlation Determining the search combined with random search based on binary particle swarm optimization to search for an optimal endmember subset for each pixel to solve the uncertainty of the types and the number of endpoints involved in hybrid and improve the search efficiency . Through the simulation of simulated data and real hyperspectral data, the accuracy of this method is comparable to that of multi-elemental spectrum hybrid analysis, which is much higher than that of full-limit / unrestricted least square method and cross-correlation spectroscopy. Meta-spectrum hybrid analysis enables fast, accurate hyperspectral data unmixing.