高光谱图像降维压缩比自动检测数学模型仿真  

Mathematical Model Simulation of Dimension Reduction Compression Ratio Automatic Detection for Hyperspectral Images

作　　者：方知 FANG Zhi(Dongchang College of Liaocheng University,Liaocheng Shandong 252000,China)

机构地区：[1]聊城大学东昌学院,山东聊城252000

出　　处：《计算机仿真》2021年第4期487-491,共5页Computer Simulation

基　　金：山东省教育科学“十二五”规划课题(CBS15009)。

摘　　要：传统高光谱图像降维压缩时无法保留全部信息,还会混入冗杂数据,导致其检测精度不高。为此提出高光谱图像降维压缩比自动检测数学模型。对初始图像做分割处理,利用关联性进行波段子空间分割,设定适当目标,计算出局部数值,并通过图像的协方差矩阵,获取偏度值与峰度值,得到JSFK模型;在此基础上,通过波段预测系数,获得图像波段预测压缩比;选取适当比例区域,根据降维压缩比的主要成分及其物理特征,使用欧式距离计算法得出最优峰度数值,从而建立压缩比自动检测数学模型。通过检测模型验证,相对于传统方法,可以最大程度的保留原始数据,在压缩时还能去除图像初始光谱维的冗余数值,进一步减少检测时间,使精准度和效率都得到提升。It is unable for traditional hyperspectral images to retain all information when reducing the dimension, leading to low detection accuracy. Therefore, a mathematical model for automatically detecting the dimension reduction compression ratio of hyperspectral images was presented. First, the initial image was segmented, and the band subspace was segmented by relevance. Second, an appropriate target was set and the local value was calculated. Meanwhile, the deviation value and peak value were obtained through the image covariance matrix, and the JSFK model was built. On this basis, the band prediction compression ratio of the image was calculated using the band prediction coefficient. And then, the regions with an appropriate proportion were selected. According to the main components and physical characteristics of dimension reduction compression ratio, Euclidean distance was used to calculate the best peak value, and then the mathematical model of compression ratio automatic detection was established. Through model verification, the proposed method can retain original data to the greatest extent. In addition, this method can remove the redundant value of the initial spectral dimension of the image during compression, so as to further reduce detection time and improve the accuracy and efficiency.

关 键 词：高光谱图像 降维压缩比 自动检测 数学模型 

分 类 号：TP421[自动化与计算机技术]