基于HJ-1星和GF-1号影像融合特征提取冬小麦种植面积  

Extraction of Winter Wheat Planting Area Based on Fusion Features of HJ-1 and GF-1 Image

作　　者：张宏 李卫国[1,2] 张晓东 卢必慧[1] 张琤琤[1] 李伟[3] 马廷淮[4] ZHANG Hong;LI Weiguo;ZHANG Xiaodong;LU Bihui;ZHANG Chengcheng;LI Wei;MA Tinghuai(Institute of Agricultural Information,Jiangsu Academy of Agricultural Sciences,Jiangsu Nanjing 210014,China;College of Agricultural Engineering,Jiangsu University,Jiangsu Zhenjiang 212013,China;Fluid Machinery Engineering Technology Research Center,Jiangsu University,Jiangsu Zhenjiang 212013,China;Institute of International Education,Nanjing University of Information of Science and Technology,Jiangsu Nanjing 210044,China)

机构地区：[1]江苏省农业科学院农业信息研究所,南京210014 [2]江苏大学农业工程学院,江苏镇江212013 [3]江苏大学流体机械工程技术研究中心,江苏镇江212013 [4]南京信息工程大学国际教育学院,南京210044

出　　处：《中国农业科技导报》2024年第2期109-119,共11页Journal of Agricultural Science and Technology

基　　金：国家重点研发计划项目(2021YFE0104400);江苏省农业科技自主创新资金项目(CX[20]2037)。

摘　　要：为提高基于国产环境与灾害监测预报卫星(HJ-1/CCD)影像大范围提取冬小麦种植面积的精度,以江苏省宿迁市沭阳县为研究区域,对冬小麦拔节期30 m×30 m的HJ-1/CCD多光谱影像和2 m×2 m的高分1号卫星全色影像(GF-1/PMS)进行融合与面向对象分类研究。将GF-1/PMS全色影像进行8、16和24 m重采样,得到4种空间分辨率(含2 m)的全色影像,分别与HJ-1/CCD多光谱影像利用光谱锐化法(Gram-Schmidt,GS)进行融合。通过对融合影像进行质量评价,选择适合研究区冬小麦种植田块格局的适宜尺度影像。将HJ-1/CCD多光谱影像重采样,得到与适宜尺度融合影像相同尺度的影像,在两景影像中分别选取包含光谱、纹理信息的训练融合影像样本(samples of fused image,SFI)和重采样影像样本(samples of resampling image,SRI),采用面向对象分类方法对适宜尺度融合影像(fused image,FI)和重采样影像(resampling image,RI)进行冬小麦种植面积提取。结果表明,16 m×16 m融合影像的效果优于2 m×2 m、8 m×8 m和24 m×24 m融合影像,其均值、标准差、平均梯度和相关系数分别为161.15、83.01、4.55和0.97。面向对象分类后,SFI对重采样影像RI16m分类的总体精度为92.22%,Kappa系数为0.90。SFI对融合影像FI16m分类的总体精度为94.44%,Kappa系数为0.93。SRI对重采样影像RI16m分类的总体精度为84.44%,Kappa系数为0.80。SFI对融合影像FI16m分类效果最好,说明基于融合影像和融合影像提取样本(SFI)结合的面向对象分类方法能准确提取冬小麦种植面积。另外,重采样影像和融合影像提取样本(SFI)相结合的面向对象分类方法也可较好提取冬小麦种植面积。为利用国产中空间分辨率HJ-1/CCD卫星和高分1号卫星融合影像有效提取大区域冬小麦种植面积信息提供了参考。In order to improve the accuracy of extracting large-scale winter wheat planting area from the data of domestic environment and disaster monitoring and forecasting satellite(HJ-1/CCD).This study took Shuyang County,Suqian City,Jiangsu Province as the research area.The fusion and object-oriented classification of the 30 m×30 m HJ-1/CCD multispectral image and the 2 m×2 m GF-1 panchromatic image(GF-1/PMS)at the jointing stage of winter wheat were carried out.The GF-1/PMS panchromatic images were resampled at 8,16 and 24 m,and panchromatic images with four spatial resolutions(including 2 m)were obtained,which were fused with HJ-1/CCD multispectral images by Gram-Schmidt (GS), respectively. Through the quality evaluation of the fused image, theappropriate scale image suitable for the pattern of winter wheat planting fields in the study area was selected. TheHJ-1/CCD multispectral image was resampled to obtain an image with the same scale as the appropriate scale fusedimage. In the 2 scene images, the training samples SFI (samples of fused image) and SRI (samples of resamplingimage) containing spectral and texture information were selected respectively, the object-oriented classificationmethod was used to extract the planting area of winter wheat from fused image (FI) and resampling image (RI). Theresults showed that the fusion effect of 16 m×16 m fused images was better than 2 m×2 m, 8 m×8 m and 24 m×24 mfused images, and the mean, standard deviation, average gradient and correlation coefficient were 161.15, 83.01,4.55 and 0.97. After object-oriented classification, the overall accuracy of SFI for the classification of resampledimage RI16m was 92.22%, and the Kappa coefficient was 0.90. The overall accuracy of SFI for the classification offused image FI16m was 94.44%, and the Kappa coefficient was 0.93. The overall accuracy of SRI for the classificationof resampled image RI16m was 84.44%, and the Kappa coefficient was 0.80. The classification effect of SFI for thefused image FI16m was the best, indicating that

关 键 词：HJ-1/CCD卫星影像 GF-1/PMS卫星影像 冬小麦种植面积 特征提取 影像融合 面向对象分类 

分 类 号：S127[农业科学—农业基础科学]