基于哨兵二号红边波段特征的作物分类  

Crop Classification Based on Sentinel-2 Red-Edge Band Features

作　　者：王美月 万红[1] 刘法军 刘晓君 WANG Mei-yue;WAN Hong;LIU Fa-jun;LIU Xiao-jun(College of information Science and Engineering/Shandong Agricultural University,Tai'an 271018,China;Shandong Fifth Institute of Geology and Mineral Exploration,Tai'an 271019,China)

机构地区：[1]山东农业大学信息科学与工程学院,山东泰安271018 [2]山东省第五地质矿产勘察院,山东泰安271099

出　　处：《山东农业大学学报(自然科学版)》2025年第1期11-20,共10页Journal of Shandong Agricultural University:Natural Science Edition

基　　金：国家青年基金(41501409);山东农业大学横向课题(140-381293)。

摘　　要：为探究红边波段特征在作物信息提取方面的潜力,本文基于2020年岱岳区哨兵二号光学影像数据,首先进行多尺度分割,然后综合利用红边波段特征的光谱特征和纹理特征,建立最优特征空间,最后在不同红边波段参与下进行大豆、小麦、果园和其他作物的分类识别。结果表明:(1)同时采用红边波段的光谱和纹理特征,相比仅使用单一类型的特征信息,显著提升了大多数作物类型的识别精度。(2)考虑全部红边波段后,分类的总体精度和Kappa系数分别达到87.1%和0.855,与不考虑红边波段相比,分别提升了9.1%和10.6%,显著改善了对作物类别的区分能力,减少了类别混淆。研究结果可为红边波段特征的深度分析和作物类别的精细提取提供技术参考。To explore the potential of red-edge band features in crop information extraction,this paper utilized Sentinel-2 optical image data of Daiyue District from 2020.Firstly,performed multi-scale segmentation was performed.Then,the spectral features and texture features of red edge band were comprehensively used to establish the optimal feature space.Finally,soybean,wheat,orchard and other crops were classified under the participation of different red-edge bands.The results show that:(1)The combined use of spectral and textural features from the red-edge bands significantly improves the recognition accuracy for most crop types,compared to using only a single type of feature information.(2)Considering all red-edge bands,the overall accuracy and Kappa coefficient of the classification reach 87.1%and 0.855,respectively,representing an increase of 9.1%and 10.6%compared to the case without red-edge bands.The results can provide technical reference for in-depth analysis of red-edge band and fine classification of crop types.

关 键 词：红边波段 面向对象 作物分类 多尺度分割 特征选择 

分 类 号：S127[农业科学—农业基础科学]