Integrated Method for Pixel-Based Supervised Classification and Object-Based Segmentation: Identifying Agricultural Land and Renovated Grassland
Integrated Method for Pixel-Based Supervised Classification and Object-Based Segmentation: Identifying Agricultural Land and Renovated Grassland
复制标题
基于像素的监督分类和基于对象的分割的集成方法：识别农业用地和更新的草地
DOI：
10.11440/rssj.39.225
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
中嶋 康博
中科院分区：
文献类型：
--
作者：
佐藤 赳;村上 智明;中嶋 康博
为了提高草地畜牧业的饲料自给率和管理效率，必须通过草地整治来提高饲草料的质量和产量。遥感分析可以用来对大范围内的草地进行监测。一些研究调查了已修复的草原；然而，这些研究在已修复的草原和其他土地利用/土地覆盖之间存在误判。因此，在这项研究中，我们开发了一种基于像素和基于对象的图像分析相结合的方法来进行基于地块的估计，并将其应用于北海道康森高原的草原。首先，我们创建了一个农田部分。其次，将监督分类结果叠加，确定最终的土地利用/土地覆盖分类。与传统的基于Landsat 8 OLI和SPOT 6的监督分类结果相比，利用SPOT 6进行农田分割显著提高了kappa系数，分类精度也高于以往的研究。
To increase the feed self-sufficiency of livestock and management efficiency of dairy farming on a grassland, it is necessary to improve the quality and production of feed grass through grassland renovation. Remote sensing analysis can be used to monitor renovated grassland over a broad area. A few studies have investigated renovated grasslands; however, these contain a misjudgment between renovated grassland and other land use/land cover. Therefore, in this study, we developed a method to integrate pixel-and object-based image analysis to conduct plot based estimation and applied it to grasslands on the Konsen plateau in Hokkaido. First, we created a farmland segment. Second, we overlaid the supervised classification results and decided the final land use/land cover classification. Performing farmland segmentation using SPOT 6 enhanced the kappa coefficient significantly compared with the traditional supervised classification results obtained using both Landsat 8 OLI and SPOT 6. The classification accuracy is also higher compared with that achieved in previous studies.