摘要
首先将一幅100像元×100像元的高光谱图像划分为以5像元×5像元构成的区域,作为大像元的图像,大像元中25个像元特征值的均值作为一个大像元的特征值。每组图像有175幅图像,实验中将他们分成四段,发现第三、四段的信息熵较小(其数值仅为二或三位数的数量级),仅用第一、二段的数据进行分类。选用图像信息熵≥1 500的图幅,进行该类图像信息熵均值的计算。农田、山体、居民地、水体4类不同地物的信息熵的均值和4种地物中某个地物的属性,利用其信息熵值与4种不同地物信息熵的均值比较,差值最小者的属性,即为待定地物的属性。
Firstly,a hyperspectral image of 100×100 pixels is converted into a region composed of 5×5 pixels,as a large pixel image,the mean of the eigenvalues of 25 pixels in a large pixel is regarded as the eigenvalue of a large pixel.The information entropy of a hyperspectral image is obtained on the basis of a large pixel image.There are 175 images in each group.In the experiment,they are divided into four segments.It is found that the information entropy of the third and fourth segments are smaller(their value is only the order of magnitude of two or three digits),only the data in the first and second segments are adopted.We choose the images whose information entropy is more than 1 500,and calculate the mean value of information entropy of this kind of images.There are four kinds of different land objects whose mean information entropy are known:farmland,mountain,residential area and water area.To determine the attributes of a certain land object in the four,we compare its information entropy with the mean values of information entropy of four different land objects,and the attributes of the smallest difference are that of the land object to be determined.
作者
郑肇葆
郑宏
ZHENG Zhaobao;ZHENG Hong(School of Remote Sensing and Information Engineering,Wuhan University,Wuhan430079,China;School of Electronic Information,Wuhan University,Wuhan430072,China)
出处
《测绘地理信息》
2019年第5期8-10,共3页
Journal of Geomatics
基金
深圳市基础科研项目(JCYJ20150422150029095)
关键词
图像信息熵
高光谱图像分类
大像元
image information entropy
hyperspectral image classification
large image pixels
