基于Xaiver NX的高光谱数据滤波并行加速算法研究

Parallel Acceleration Algorithm for Hyperspectral DataFiltering Based on Xaiver NX

高光谱遥感图像将空间图像维和光谱维结合形成数据立方体,高光谱图像数据的每个像素维度都有一条光谱曲线,而每个光谱曲线分类的运算都是重复进行的,数据冗余比较大,所以数据处理的速度,就决定了高光分类设备的最高运行速度。在高光谱数据分析之前需要进行滤波,为了减小分析的时间,需要对数据滤波运算进行加速,GPU正好具有并行运算的能力,可以显著提高运算的速度。首先,本文介绍了高光谱数据滤波的相关方法;其次,介绍了显示核心(Graphics Processing Unit,GPU)并行运算的原理;最后,以SG滤波为例,将算法与GPU硬件相结合并进行优化,显著提升了高光谱数据预处理的运算速度。

Hyperspectral remote sensing image combines spatial image and spectral dimension to form a data cube.Each pixel dimension of hyperspectral image data has a spectral curve,and the classification operation of each spectral curve is repeated,and the data redundancy is relatively large.Therefore,the speed of data processing determines the maximum operation speed of hyperspectral classification equipment.Filtering is required before hyperspectral data analysis.In order to reduce the analysis time,the data filtering operation needs to be accelerated.GPU just has the ability of parallel operation,which can significantly improve the speed of operation.Firstly,this paper introduces the related methods of hyperspectral data filtering;Secondly,the principle of parallel operation of graphics processing unit(GPU)is introduced;Finally,taking SG filtering as an example,the algorithm is combined with GPU hardware for optimization,which significantly improves the operation speed of hyperspectral data preprocessing.