摘要
高光谱遥感影像通常包含几十或上百个光谱波段,其海量的数据给影像的存储、传输以及后续处理带来了挑战。针对这一问题,根据高光谱遥感影像谱间相关性强的特性,提出了一种结合双参考波段线性预测的基于压缩感知的高光谱遥感影像重构方法。首先,将高光谱遥感影像的波段进行分组,每组确定两个参考波段,使用正交匹配追踪(OMP)算法重构每组的两个参考波段。其次,根据重构恢复的组内的两个参考波段,建立了一个基于双参考波段的线性预测模型,用来计算该组内非参考波段的预测值;然后,使用OMP算法重构实际测量值与预测测量值的差值,得到差值向量;最后,利用得到的差值向量迭代修正预测测量值,直到恢复该波段原始图像。仿真实验结果表明,该方法提高了高光谱遥感影像的重构效果。
Hyperspectral remote sensing images usually contain tens or hundreds of bands. Its massive data has been a challenge to the storage, transmission and subsequent processing. To solve this problem, a new method of hyperspectral remote sensing image reconstruction based on compressive sensing is proposed. This method combines the characteristics of high spectral correlation of hyperspectral remote sensing image. Firstly, the spectral bands of hyperspectral remote sensing images are divided into two reference bands. Each band is determined by the orthogonal matching pursuit（OMP） algorithm, which is used to reconstruct two reference bands in each group. Secondly, according to the two reference bands reconstructed within the group, a prediction model is established. This linear double reference band model based on the calculation of the group is used to predict non reference band values; then, the OMP algorithm is used to get the difference of measured values and predicted values; finally, the predicted values are modified with the difference vector iteration until the original band image is restored. The simulation results show that the proposed method improves the reconstruction effect of hyperspectral remote sensing image.
作者
王晗
王阿川
WANG Han,WANG A’chuan(College of Information and Engineering, Northeast Forestry University, Harbin 150040, Chin)
出处
《红外技术》
CSCD
北大核心
2018年第6期556-562,共7页
Infrared Technology
基金
黑龙江省自然科学基金资助项目(C201414)
哈尔滨市优秀学科带头人基金项目(2014R FXXJ040)
关键词
高光谱遥感影像
压缩感知
线性预测
图像重构
hyperspectral remote sensing image
compressive sense
linear prediction
image reconstruction
