结合金字塔和局部二值模式的遥感图像分类  

Remote sensing image classification based on pyramid and local binary pattern

作　　者：吴庆岗[1] 赵伊兰 夏永泉[1] 李灿林[1] WU Qinggang;ZHAO Yilan;XIA Yongquan;LI Canlin(School of Computer and Communication Engineering,Zhengzhou University of Light Industry,Zhengzhou 450001,China)

机构地区：[1]郑州轻工业学院计算机与通信工程学院

出　　处：《现代电子技术》2019年第13期56-60,64,共6页Modern Electronics Technique

基　　金：国家自然科学基金项目(61502435);国家自然科学基金项目(61602423);国家自然科学基金项目(61702464);河南省教育厅科技攻关项目(14A520034);郑州轻工业学院博士基金项目(2013BSJJ041);郑州轻工业学院校青年骨干教师(13300093)~~

摘　　要：在住宅区遥感图像分类中,为了克服尺度变化和旋转变化带来的影响,提出一种结合金字塔原理和局部二值模式的图像分类算法。首先对原始住宅区遥感图像进行多次下采样以构建不同尺度的空间金字塔;然后利用局部二值模式提取不同尺度遥感图像的纹理特征,以消除旋转变化的影响;最后将不同尺度下的纹理特征融合到一起,利用支持向量机对住宅区遥感图像进行分类。在标准图像数据集上的实验结果表明,低尺度纹理特征将会降低住宅区遥感图像的分类精度,与单尺度纹理特征相比,多尺度融合的纹理特征提高了遥感图像分类精度,平均高达4.77%。In order to overcome the effects of scale and rotation variations for the classification of residential areas in remote sensing images, an image classification algorithm based on pyramid principle and local binary pattern (LBP) is proposed. The original residential remote sensing images are repetitiously down-sampled to construct the spatial pyramids with different scales. The local binary pattern(LBP) is used to extract the texture features of remote sensing images with different scales,so as to eliminate the effect of rotation variation. The extracted texture features with different scales are fused together, and the support vector machine (SVM) is adopted to classify the remote sensing images of residential area. The extensive experimental results obtained from standard remote sensing image datasets demonstrate that the texture features with low scale can reduce the classification accuracy of remote sensing image of residential area,and the texture features with multi-scale fusion can improve the classification accuracy of remote sensing image as high as 4.77% in comparison with the single-scale texture features.

关 键 词：遥感图像 局部二值模式 空间金字塔 特征融合 支持向量机 图像分类 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP751[电子电信—信息与通信工程]