摘要
目标检测是高光谱数据处理的重要应用之一,高光谱图像中空间和光谱信息的充分利用对于目标检测率的有效提升非常关键。提出一种新的联合稀疏表示的目标检测方法,将混合范数理论和算法应用于高光谱目标检测,在联合高光谱图像空间和光谱信息的基础上,建立了基于联合稀疏性约束的混合范数正则化数学模型,并利用交替方向乘子法对模型进行了优化求解。仿真实验结果表明,该方法能有效提高高光谱目标检测的准确性,降低虚警率。
Target detection is one of the most important applications of hyperspectral imagery (HSI). The traditional target detection techniques usually discard the spatial information of the target, resulting in a lower accuracy of detection. A novel simultaneous sparse representation model is proposed for HSI target detection. The proposed approach applies the theory and algorithm of mixed-norm to the hyperspectral target detection. By considering the combination of spectral information and spatial context of HSI, a model with a mixed-norm regularizaton based on the simultaneous sparse representation is proposed. And this model is finally solved via alternating direction mehtod of multipliers (ADMM) efficiently. The effec- tiveness and accuracy of the proposed simultaneous sparse representation model and algorithm are demon- strated by experimental results on a real hyperspectral images.
出处
《兵工学报》
EI
CAS
CSCD
北大核心
2014年第6期834-841,共8页
Acta Armamentarii
基金
国家自然科学基金项目(61101194)
江苏省自然科学基金项目(BK2011701)
江苏省'六大人才高峰'项目(WLW-011)
高等学校博士学科点专项科研基金项目(20113219120024)
中国空间技术研究院创新基金项目(CAST201227)
中国地质调查局工作项目(1212011120227)
关键词
信息处理技术
高光谱图像
目标检测
混合范数
联合稀疏性
交替方向乘子法
information processing
hyperspectral imagery
target detection
mixed norm
simultaneous sparsity
altermation direetion mehtod of muhipciers