复合正则化联合稀疏贝叶斯学习的高光谱稀疏解混算法

将稀疏贝叶斯学习引入线性混合像元分解中,提出一种基于复合正则化联合稀疏贝叶斯学习的高光谱稀疏解混算法.在多观测向量的稀疏贝叶斯框架下,对各参数建立概率模型,经贝叶斯推断得到基于L2,1正则化的联合稀疏贝叶斯解混模型,并将丰度向量的非负与和为一约束加入到凸优化的目标函数中,通过变量分离法将复合正则化问题分解成多个单一正则化问题交替迭代求解,并利用参数自适应算法对正则化参数进行更新.模拟数据和真实数据的实验结果表明,该算法比贪婪算法和凸优化算法能获得更高的解混精度,并且适用于端元个数较多和信噪比较低的高光谱数据.

Compound regularized multiple sparse Bayesian learning algorithm for sparse unmixing of hyperspectral data

A compound regularized multiple sparse bayesian learning algorithm for sparse unmixing is presented, in which sparse bayesian learning model is integrated in the linear hyperspectral pixel unmixing. On the framework of sparse Bayesian Learning model based on MMV (Multiple Measurement Vectors), the parameters in the model is established with the probability, and a L2,1 norm regularization-based multiple sparse bayesian learning model for spectral unmixing is constructed by bayesian inference, taking the non-negativity and sum-to-one property of abundances into the convex objective function. The compound regularization problem is decomposed into several single regularization problems solving by a variable separation method, and the regularization parameters of the model are updated by an adaptive adjustment algorithm. Experimental results on both synthetic and real hyperspectral data demonstrate that the proposed method outperforms the greedy algorithms and the convex algorithms with a better spectral unmixing accuracy, and is suitable for complex combination of endmembers and low signal-to-noise ratio hyperspectral data.