摘要
正交匹配追踪算法是稀疏求解中常用的方法,用于噪声影响下的高光谱数据稀疏解混时,其解混效果不理想.针对这一问题,提出了全约束DOMP算法.通过引入广义Dice系数代替内积作为匹配度量准则,更充分地利用了光谱信息,提高了算法的抗噪能力.同时,为了满足丰度的"非负"及"和为1"的性质,对丰度系数进行了全约束,进一步改善了解混效果.模拟及真实数据仿真结果显示,改进算法明显提高了解混精确度,验证了算法的有效性.
It will lead to undesirable unmixing results when OMP algorithm is used for noisy hyperspectral sparse unmixing.In order to solve this problem,a new constraint DOMP-based sparse unmixing algorithm is proposed,where the important spectral information could be highlighted by introducing Dice coefficient as the matching measurement criteria instead of inner product.Also considering constraints of abundance non-negativity and abundance sum-to-one,imposing ＂fully constraints＂on the abundance values which improved the unmixing results.Experiments on both simulated and real hyperspectral data show that the proposed algorithms outperform the OMP sparse unmixing algorithm,the feasibility of the algorithms is proved.
出处
《沈阳大学学报（自然科学版）》
CAS
2015年第3期206-213,共8页
Journal of Shenyang University：Natural Science
基金
国家自然科学基金资助项目(61405041)
黑龙江省自然科学基金重点资助项目(ZD201216)
哈尔滨市优秀学科带头人基金资助项目(RC2013XK009003)
中国博士后科学基金资助项目(2014M551221)
关键词
高光谱图像
稀疏解混
正交匹配追踪
广义Dice系数
丰度约束
hyperspectral imagery
sparse unmixing
orthogonal matching pursuit
generalized Dice coefficient
abundance constraints
