| 摘要: |
| 【目的】 为利用中低分辨率遥感影像，精确获取县域尺度的农田分类结果，提供一定的方法参考。【方法】该文以中等分辨率的Landsat8 OLI遥感影像数据为数据源，采用面向对象的CART决策树分类法，对垫江县的农田进行了识别与提取，并与基于像元提取的最大似然法分类结果进行精度对比。【结果】研究表明：（1）与最大似然分类法相比，CART决策树分类法的精度更高，总体精度和Kappa系数分别达到88.8%和0.85，对于旱地和水田的制图精度高达90%以上；（2）在分割尺度的确定上，使用eCognition软件中的ESP工具能快速的确定最优分割尺度，提升了效率和科学性；（3）在县域尺度上，面向对象分类法对中等分辨率影像数据进行遥感提取也具有一定的适用性。【结论】基于landsat8 OLI遥感数据的面向对象分类法能够实现县域尺度低成本高精度农田分类的需要，也为缓解精度和成本、空间分辨率和提取方法的矛盾提供了一定的参考。 |
| 关键词: 遥感；面向对象分类；监督分类；县域尺度；农田提取 |
| DOI： |
| 分类号:S29 |
| 基金项目:国家科技支撑计划课题(No.2015BAD07B04) |
|
| Farmland Recognition and Extraction Based on Object-oriented Classification |
|
zhangwei
|
|
College of Geography and Tourism, Chongqing Normal University
|
| Abstract: |
| 【Objective】In order to provide a method for accurately obtaining farmland classification results at a county scale using medium and low resolution remote sensing images.【Method】Taking medium-resolution landsat8 OLI remote sensing image data as data source and using the object-oriented CART decision tree classification method to identify and extract the farmland in Dianjiang County, then comparison the classification results with maximum likelihood method which extract based on pixel.【Result】The research results showed that: (1) Compared with the maximum likelihood classification method, the CART decision tree classification method has higher precision, the overall classification accuracy and Kappa coefficient reached 88.8% and 0.85, respectively, and the mapping accuracy for dry land and paddy fields is over 90%; (2) In the determination of the segmentation scale, the ESP tool in the eCognition software can quickly determine the optimal segmentation scale and improving efficiency and scientificity; (3) On the county scale, the object-oriented classification method also has certain applicability to remote sensing extraction of medium resolution image data.【Conclusion】The object-oriented classification method based on landsat8 OLI remote sensing data can realize the need of low-cost and high-precision farmland classification at county scale, and also provides a reference for mitigating the contradiction between precision and cost, spatial resolution and extraction method. |
| Key words: remote sensing; object-oriented classification; supervised classification; county scale; farmland extraction |
