Knowledge Resource Center for Ecological Environment in Arid Area
| 时间序列卫星数据的农作物分类研究及应用 ——以天山北坡为例 | |
| 其他题名 | Crop Classification Based on Time Series Satellite Imagery in Northern Slope of Tianshan Mountains |
| 黄双燕 | |
| 出版年 | 2018 |
| 学位类型 | 硕士 |
| 导师 | 陈曦 |
| 学位授予单位 | 中国科学院大学 |
| 中文摘要 | 当前，基于机器学习方法开展农作物分类研究，对于确保干旱区粮食安全和生态安全有着极为重要的现实意义。近年来，遥感技术在农作物分类和种植结构提取方面已取得长足的发展，但仍面临诸多困难与挑战：一是农作物种类自动识别方法有待进一步提升，二是时间序列多源数据融合技术有待进一步研究。本研究选择时间序列Sentinel 2A与Landsat 8卫星数据，结合野外调查数据，采用随机森林机器学习方法，探索基于地块基元和红边特征的多源遥感数据农作物自动分类方法，实现了天山北坡地区主要农作物种植结构的自动提取。主要内容包括：引入地块基元和红边特征，探讨不同分类特征组合对机器学习分类精度的影响；探讨分别在数据级、特征级联合利用中分辨率时间序列多源卫星数据对农作物分类精度有何影响，确定时间序列 Sentinel 2A数据和Landsat 8数据在干旱区分类应用过程中的最佳融合层次。结果表明：（1）基于时间序列Sentinel 2A，分别利用8种时间序列光谱、植被指数分类特征集对沙湾县进行农作物分类识别，总体精度均在89 % 以上。分类组最高精度达94.02 %，表明随机森林机器学习分类器可有效集成多维向量的优势，是一种行之有效的干旱区典型农作物分类方法。引入地块基元后，有红边光谱组及有红边指数组的分类精度分别提高3.13 % 和4.07 %，表明利用地块基元中心点作为同质地块的代表像元，提取的分类特征可供机器学习分类器使用，能较大幅度提高分类精度及效率，平滑地块内部同种作物间的“椒盐噪声”。红边特征的引入使光谱组及指数组的分类精度分别提高2.39 % 和1.63 %，对随机森林分类模型的精度提升十分有效。（2）探讨基于数据级、特征级的多源数据联合利用方式对农作物分类精度的影响，结果表明Landsat 8与Sentinel 2A的融合层次在植被指数特征级精度最高。在此基础上对温泉县进行农作物遥感分类识别的精度达92.92%。特征级分类特征的增加会促使随机森林分类器的总体分类精度及各作物精度提升。（3）引入地块基元与红边特征，基于Landsat 8与Sentinel 2A的时间序列植被指数特征级融合数据，采用上述基于随机森林的机器学习方法，对天山北坡主要农作物进行分类识别，其总体分类精度达到了81.51 %，符合大区域遥感监测应用需求。2016年天山北坡研究区（含兵团）内，小麦种植面积最大的区域是奇台县，占研究区小麦种植面积的18.66 %。春玉米种植面积最大的区域是呼图壁，占研究区春玉米种植面积的20.59 %。棉花种植面积最大的区域是沙湾县，占研究区棉花种植面积的31.16 %。提取的天山北坡2016年种植结构表明，棉花是天山北坡经济带种植面积最大的作物，主要分布在天山北坡的中部区域的冲积平原区及洪积冲积扇区。春玉米主要分布在天山北坡的山前丘陵区。小麦在整个天山北坡经济带的均有分布，但种植面积较少且地块较为零星破碎。研究结果具有科学意义和实用价值，为机器学习方法及时间序列多源遥感数据在干旱区农业遥感的应用提供了参考。部分研究结果在甘家湖棉花受灾案例中得到了应用。 |
| 英文摘要 | Currently, research on crop classification based on machine learning methods is of great practical significance for ensuring food security and ecological security in arid regions. In recent years, remote sensing technology has made great progress in the classification of crops and extraction of planting structures, but it still faces many difficulties and challenges. First, the method of automatic identification of crop types needs to be further improved. Second, the time series multi-source data fusion technology needs further study. The project choose the random forest machine learning method based on the selection of time series Sentinel 2A and Landsat 8 satellite data, and combined with field survey data, so that we can explore a multi-source remote sensing data crop classification method for automatic extraction of planting structure of major crops based on the characteristics of the ground block and red edge in the northern slope of the Tianshan Mountains. The main contents include several poins listed in the following: the introduction of land-based primitives and red edge features, and explores the effects of different classification feature combinations on the accuracy of machine learning; and the use of multiple-source satellite data pairs at the data level and feature level respectively. What effect does the crop classification accuracy have on the determination of the best integration level of time series Sentinel 2A data and Landsat 8 data in the arid zone? The results are listed in the following. Firstly, based on the time series Sentinel 2A, eight kinds of time series spectra and vegetation index classification feature sets were used to classify and identify crops in Shawan County, and the overall accuracy was over 89%. The highest accuracy of the classification group is 94.02 %, which indicates that the random forest machine learning classifier can effectively integrate the advantages of multidimensional vectors, and is an effective method for classifying typical crops in arid regions. After the introduction of the block-based primitives, the classification accuracy of the red-edge spectral group and the red-edge index group increased by 3.13% and 4.07% respectively. It indicats that the use of the center point of the block-based block as the representative pixel of the homogeneous block, the extracted classification The feature can be used by the machine learning classifier to greatly improve the classification accuracy and efficiency, and to smooth the "pepper and salt noise" among the same kinds of crops within the block. The introduction of the red edge feature improved the classification accuracy of the spectral group and the exponent group by 2.39 % and 1.63 %, respectively, which was very effective in improving the accuracy of the random forest classification model. Secondly, explore the impact of the combined use of multi-source data based on the data level, and the feature level on the accuracy of crop classification. The results show that the integration level of Landsat 8 and Sentinel 2A has the highest accuracy in the vegetation index. On this basis, the accuracy of remote sensing crops classification reached 92.92% in Wenquan county. The increase of feature-level classification features will promote the overall classification accuracy of random forest classifiers and improve the accuracy of various crops. Finally, incorporating the characteristics of land parcels and red edges, based on the characteristics of Landsat 8 and Sentinel 2A time series vegetation indices and fusion data, using the above-mentioned machine learning method based on random forests to classify and identify the main crops on the northern slope of the Tianshan Mountains. The overall classification accuracy reached 81.51%, which is in line with the needs of large-scale remote sensing monitoring applications. In the 2016 Tianshan North Slope Research Area (including the Bingtuan), the largest wheat planting area was in Qitai County, accounting for 18.66% of the wheat planted area in the study area. The area with the largest spring maize planting area is Hutubi, which accounts for 20.59% of the planting area of spring maize in the study area. The area with the largest cotton planting area is Shawan County, accounting for 31.16% of the cotton planted area in the study area. The extracted planting structure of the northern slope of the Tianshan Mountains in 2016 shows that cotton is the largest crop planted in the economic belt on the northern slope of the Tianshan Mountains. It is mainly distributed in the alluvial plains and alluvial sectors in the central area of the northern slope of the Tianshan Mountains. Spring corn is mainly distributed in the hilly area on the northern slope of the Tianshan Mountains. Wheat is distributed throughout the northern slope of the Tianshan Mountains, but the planting area is small and the land is broken up. The research results have scientific significance and practical value. It provides a reference for machine learning methods and time series multi-source remote sensing data for agricultural remote sensing applications in arid regions. Some research results have been applied in the case of cotton disaster in Ganjia Lake. |
| 中文关键词 | 机器学习 ; 随机森林 ; 农作物分类 ; 地块基元 ; 多源数据 |
| 英文关键词 | Machine learning Random forest Crop classification Parcel data set Multi-sources data |
| 语种 | 中文 |
| 国家 | 中国 |
| 来源学科分类 | 测绘工程 |
| 来源机构 | 中国科学院新疆生态与地理研究所 |
| 资源类型 | 学位论文 |
| 条目标识符 | http://119.78.100.177/qdio/handle/2XILL650/288192 |
| 推荐引用方式 GB/T 7714 | 黄双燕. 时间序列卫星数据的农作物分类研究及应用 ——以天山北坡为例[D]. 中国科学院大学,2018. |
| 条目包含的文件 | 条目无相关文件。 | |||||
| 个性服务 |
| 推荐该条目 |
| 保存到收藏夹 |
| 导出为Endnote文件 |
| 谷歌学术 |
| 谷歌学术中相似的文章 |
| [黄双燕]的文章 |
| 百度学术 |
| 百度学术中相似的文章 |
| [黄双燕]的文章 |
| 必应学术 |
| 必应学术中相似的文章 |
| [黄双燕]的文章 |
| 相关权益政策 |
| 暂无数据 |
| 收藏/分享 |
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。
