APP下载

基于OIF和最优尺度分割的GF?2影像分类适用性研究

2018-04-13任金铜杨武年邓晓宇王蕾王芳

现代电子技术 2018年8期
关键词:分类研究

任金铜 杨武年 邓晓宇 王蕾 王芳

摘 要: 高分二号作为国产高分辨率遥感的代表,其影像数据对提高地物信息提取的质量和精度的作用值得研究和探索。通过分析高分二号多光谱数据特点,利用OIF指数选取最佳波段组合,选用ESP最优尺度分析算法获得研究区最优分割尺度,最后在最佳波段组合和最优分割尺度的基础上提取典型地物,并对分类结果进行精度验证。研究发现,高分二号多光谱数据最佳波段组合为134;利用最佳波段组合和最优分割尺度提取地物信息的总体分类精度均大于85%,Kappa系数均大于0.8,分类结果精度较高;从总体来看,当选用最优分割尺度为82时,分类结果精度最高,其总体精度为93%,Kappa系数为0.910;其次是最优分割尺度为31,其总体精度为89%,Kappa系数为0.859;最优分割尺度为42的分类结果精度表现最差,其总体精度为85%,Kappa系数为0.808。

关键词: 高分二号; OIF; 多尺度分割; 面向对象分类; KNN; 遥感卫星; 地物信息提取

中图分类号: TN911.73?34 文献标识码: A 文章编号: 1004?373X(2018)08?0072?06

Abstract: GF?2 is a representative of high?resolution remote sensing satellites of China, whose image data′s function to improve the quality and accuracy of ground object information extraction is worthy of research and exploration. Optimal waveband combinations are selected by analyzing the characteristics of GF?2′s multispectral data and using the OIF indexes. The ESP optimal scale analysis algorithm is selected to obtain the optimal segmentation scales in the research area. On the basis of optimal waveband combinations and optimal segmentation scales, the typical ground objects are extracted and the accuracy of classification results is verified. The research results show that the optimal waveband combination of GF?2′s multispectral data is 134; the overall classification accuracy of ground object information extracted by means of optimal waveband combinations and optimal segmentation scales is larger than 85%, the Kappa coefficient is larger than 0.8, and the accuracy of classification results is high; on the whole, when the selected optimal segmentation scale is 82, the accuracy of the classification results is the highest (the overall accuracy is 93% and the Kappa coefficient is 0.910); when the optimal segmentation scale is 31, the accuracy of the classification results comes to the second (the overall accuracy is 89% and the Kappa coefficient is 0.859); when the optimal segmentation scale is 42, the accuracy of the classification results is the lowest (the overall accuracy is 85% and the Kappa coefficient is 0.808).

Keywords: GF?2; OIF; multi?scale segmentation; object?oriented classification; KNN; remote sensing satellite; ground object information extraction

0 引 言

随着空间信息技术的不断发展,20世纪90年代以来高分辨率遥感卫星影像数据逐渐进入商业和民用领域,应用领域日益广阔[1]。2006年我国将高分辨率对地观测系统重大专项(高分专项)列入《国家中长期科学与技术发展规划纲要(2006—2020年)》,高分二号(GF?2)的成功发射,宣告了我国高空间分辨率遥感进入亚米时代[2]。

“高分專项”的实施使得我国自主获取高空间分辨率遥感卫星影像的能力越来越强。如何提高国产高分辨率遥感数据处理能力和信息提取精度值得进一步研究和探索。……

登录APP查看全文

猜你喜欢

分类研究
FMS与YBT相关性的实证研究
2020年国内翻译研究述评
辽代千人邑研究述论
分类算一算
垃圾分类的困惑你有吗
视错觉在平面设计中的应用与研究
EMA伺服控制系统研究
教你一招:数的分类
新版C-NCAP侧面碰撞假人损伤研究
给塑料分分类吧