基于多时相合成孔径雷达与光学影像的冬小麦种植面积提取
发布时间：2018-12-18 13:02
【摘要】：小麦是中国最重要的农作物之一,准确、及时掌握小麦种植面积具有重要意义。以探索合成孔径雷达(synthetic aperture radar,SAR)与光学数据对种植结构复杂地区冬小麦识别的能力,提高识别精度为目的。该研究以多时相SAR(Sentinel-1A)和光学影像(Landsat-8)为数据源,选取种植结构复杂的都市农业区为研究区。构建不同特征向量组合,利用支持向量机(support vector machine,SVM)提取冬小麦种植面积。通过对比分析基于不同特征向量组合的冬小麦识别精度,结果表明:1)使用SAR后向散射数据得到85.7%的制图精度和87.9%的用户精度;2)添加SAR数据纹理信息,总体精度高达90.6%,比单独使用后向散射数据在制图精度和用户精度上分别提高7.6%和6.7%;3)当SAR数据和光学影像结合时,总体精度高达95.3%(制图精度97%,用户精度98.4%),比单独使用SAR数据在制图精度和用户精度上分别提高3.7%和3.8%。因此,基于SAR数据的都市农业区冬小麦分类,有着较高分类精度,纹理信息和光学影像的添加能有效提高识别精度。研究结果可为SAR数据的农作物识别和应用提供理论基础。
[Abstract]:Wheat is one of the most important crops in China. The purpose of this paper is to explore the ability of synthetic Aperture Radar (synthetic aperture radar,SAR) and optical data to identify winter wheat in areas with complex planting structure and to improve the recognition accuracy. In this study, multi-temporal SAR (Sentinel-1A) and optical image (Landsat-8) were used as data sources, and the urban agricultural region with complex planting structure was selected as the study area. Different feature vector combinations were constructed and the planting area of winter wheat was extracted by support vector machine (support vector machine,SVM). By comparing and analyzing the recognition accuracy of winter wheat based on different eigenvector combinations, the results show that: 1) 85.7% mapping accuracy and 87.9% user accuracy are obtained by using SAR backscatter data; 2) adding the texture information of SAR data, the overall accuracy is as high as 90.6, which is 7.6% and 6.7% higher than that of backscatter data used alone in cartographic accuracy and user accuracy, respectively; 3) when SAR data and optical image are combined, the overall accuracy is up to 95.3% (97% for cartography and 98.4% for users), which is 3.7% and 3.8% higher than that of using SAR data alone. Therefore, the classification of winter wheat in urban agricultural area based on SAR data has higher classification accuracy. The addition of texture information and optical image can effectively improve the recognition accuracy. The results can provide a theoretical basis for crop identification and application of SAR data.
【作者单位】： 南京农业大学资源与环境科学学院;南京农业大学公共管理学院;
【基金】：江苏高校优势学科建设工程资助项目(PAPD)
【分类号】：S127;S512.11
，
本文编号：2385889
[Abstract]:Wheat is one of the most important crops in China. The purpose of this paper is to explore the ability of synthetic Aperture Radar (synthetic aperture radar,SAR) and optical data to identify winter wheat in areas with complex planting structure and to improve the recognition accuracy. In this study, multi-temporal SAR (Sentinel-1A) and optical image (Landsat-8) were used as data sources, and the urban agricultural region with complex planting structure was selected as the study area. Different feature vector combinations were constructed and the planting area of winter wheat was extracted by support vector machine (support vector machine,SVM). By comparing and analyzing the recognition accuracy of winter wheat based on different eigenvector combinations, the results show that: 1) 85.7% mapping accuracy and 87.9% user accuracy are obtained by using SAR backscatter data; 2) adding the texture information of SAR data, the overall accuracy is as high as 90.6, which is 7.6% and 6.7% higher than that of backscatter data used alone in cartographic accuracy and user accuracy, respectively; 3) when SAR data and optical image are combined, the overall accuracy is up to 95.3% (97% for cartography and 98.4% for users), which is 3.7% and 3.8% higher than that of using SAR data alone. Therefore, the classification of winter wheat in urban agricultural area based on SAR data has higher classification accuracy. The addition of texture information and optical image can effectively improve the recognition accuracy. The results can provide a theoretical basis for crop identification and application of SAR data.
【作者单位】： 南京农业大学资源与环境科学学院;南京农业大学公共管理学院;
【基金】：江苏高校优势学科建设工程资助项目(PAPD)
【分类号】：S127;S512.11
，
本文编号：2385889
本文链接：https://www.wllwen.com/kejilunwen/nykj/2385889.html
