面向对象的GF-1卫星影像苹果树遥感提取研究-以山地丘陵地区的栖霞市为例

国产高分卫星1号（简称GF-1）可提供高空间分辨率遥感数据，这为实现精准农业提供良好的数据保障。但是，其在复杂地形作物遥感提取方面的适用性仍需要进一步探索和分析。因此，以山地丘陵地形为主的栖霞市为研究区，基于GF-1影像数据利用面向对象分类法提取苹果树的种植面积和空间分布信息。研究结果表明：引入植被覆盖度和坡度信息的面向对象决策树分类方法（简称优化分类方法）能够有效提高苹果种植面积的遥感提取精度为94.1%，生产精度和用户精度分别为87.3%和90.3%。相比于最大似然监督分类和无辅助信息的面向对象决策树分类方法，本研究的优化方法遥感提取面积精度分别提高了17.4%和0.3%，生产精度分别提高了10.1%和2.6%，用户精度分别提高了10.1%和2.8%。该研究成果以期为GF-1卫星数据在山地丘陵地区的果园信息遥感提取提供基本的技术支撑和实践经验。

Apple tree extraction from GF-1 remote sensing image based on object-oriented classification method-taking Qixia city in hilly areas as an example

The domestic high resolution satellite 1 (GF-1) can provide high spatial resolution remote sensing data, which provides good data protection for precision agriculture. However, its applicability to crop extraction with remote sensing in complex terrain still needs further exploration and analysis. Therefore, based on the high resolution satellite GF-1 image data, Qixia city in mountainous and hilly areas was taken as the research area to explore the extraction of Apple tree planting area and spatial distribution information by object-oriented classification method. The results show that the object-oriented decision tree classification method with vegetation coverage and slope information can effectively improve the extraction accuracy of Apple planting area by remote sensing (94.1%), and the production accuracy and user accuracy are 87.3% and 90.3%.Compared with the maximum likelihood supervised classification and object-oriented decision tree classification method without auxiliary information, the extraction accuracy of remote sensing is increased by 17.4% and 0.3%, the production accuracy is increased by 10.1% and 2.6%, and the user accuracy is increased by 10.1% and 2.8%, respectively. The research results are expected to provide basic technical support and practical experience for extracting orchard information from GF-1 satellite data in Hilly areas by remote sensing.