窄带蜂窝物联网的嵌入式高光谱图像压缩方法

Embedded Hyperspectral Image Compression Method Based on Narrowband Cellular Internet of Things

针对当前嵌入式高光谱图像压缩方法未考虑高光谱图像空间相关性,导致压缩效果较差,解压精度较低,高分辨率图像占用空间大,造成网络传输效率低的问题,提出了窄带蜂窝物联网的嵌入式高光谱图像压缩方法。根据窄带蜂窝物联网上行、下行链路传输基本标准,通过具有统计特性的KLT算法,去除高光谱图像间谱间相关性,使待处理图像内容更加紧密,利用雅克比迭代算法,分解计算矩阵中的特征值,将多光谱图像每一波段,分成二维波形以及大小相等且不重叠的图像块,将光谱方向上相同位置的所有图像块转换成KLT,消除多光谱图像空间和光谱相关完成高光谱图像压缩。实验结果表明,所提方法具有较高的压缩精度,能够有效去除高光谱图像空间相关性,压缩效果较好,减少高光谱图像占用空间内存,缩短网络传输时间,提高网络传输效率。

At present,the embedded hyperspectral image compression method ignores the spatial correlation of hyperspectral image,resulting in poor compression effect,low decompression accuracy and low network transmission efficiency.Therefore,an embedded method of hyperspectral image compression for narrow-band cellular Internet of Things(IoT)was presented.According to the basic standards of NBC-IoT uplink and downlink transmission,the KLT algorithm with statistical characteristics was used to remove the spectral correlation between hyperspectral images,so that the content of the image to be processed was more compact.The eigenvalues in matrix were decomposed and calculated by Jacobian iterative algorithm.Each band of the multispectral image was divided into two-dimensional waveforms and non-overlapping image blocks with the same size.Moreover,all image blocks at the same position in the spectrum direction were converted into KLT,thus the multispectral image space was eliminated.Finally,the hyperspectral image compression was completed.Simulation results show that the proposed method has higher compression accuracy,which can effectively remove the spatial correlation of hyperspectral images.In addition,the method has better compression effect,so it can reduce the space memory occupied by hyperspectral images,shorten the network transmission time and improve network transmission efficiency.