基于大数据分析的光学遥感图像场景分类研究  

Research on scene classification of optical remote sensing image based on big data analysis

作　　者：张颖 万厚冲 ZHANG Ying;WAN Houchong(Handan University,Handan Hebei 056000,China;Hengshui University,Hengshui Hebei 053000,China)

机构地区：[1]邯郸学院,河北邯郸056000 [2]衡水学院,河北衡水053000

出　　处：《激光杂志》2021年第10期99-103,共5页Laser Journal

基　　金：河北省科技支撑计划项目(No.15210337)。

摘　　要：目前光学遥感图像场景分类方法,未曾确定图像场景分类隶属度函数,导致光学遥感图像场景分类效果差,图像场景分类指标值低,还存在较低的准确率和较高的错误率,为此提出基于大数据分析的光学遥感图像场景分类研究。确定光学遥感图像场景分类流程,利用大数据分析技术,分解图像场景特征,批量预测图像场景分类结果,在分类预测结果的基础上,设计图像场景分类隶属度函数,控制图像场景分类模糊度,完成图像场景分类。确定实验图像和图像场景分类对比指标计算公式,对比四组方法的实验结果可知,本方法对光学遥感图像分类可以达到100%准确识别遥感图像场景类别,且具有98.31%的准确率,且总体精度、制图精度、用户精度和Kappa系数等四个图像场景分类指标较高,分别为0.817 4、0.966 8、0.961 7、0.794 5。At present,the classification method of optical remote sensing image scene has not determined the membership function of image scene classification,which leads to poor classification effect,low index value of image scene classification,low accuracy and high error rate. Therefore,this paper puts forward the research of optical remote sensing image scene classification based on big data analysis. Determine the process of image scene classification of optical remote sensing,decompose the image scene features by using big data analysis technology,predict the image scene classification results in batches,and design the image scene classification membership function based on the classification prediction results to control the ambiguity of image scene classification and complete the image scene classification. The formula for calculating the comparison index of experimental image and image scene classification is determined. By comparing the experimental results of four groups of methods,it can be seen that the classification of optical remote sensing image by this method can achieve 100% accurate recognition of remote sensing image scene category with an accuracy rate of 98. 31%,and the four image scene classification indexes such as overall accuracy,drawing accuracy,user accuracy and Kappa coefficient are higher,which are 0. 817 4,0. 966 8,0. 961 7 and 0. 794 5 respectively.

关 键 词：大数据分析 光学遥感 图像场景 场景分类 

分 类 号：TN958.98[电子电信—信号与信息处理]