摘要
针对土地利用动态遥感监测中存在的图斑边界提取问题,探索了将前时相的地类图斑的边界和土地利用类别作为先验知识约束分割过程的方法。该方法利用图斑的边界约束分割的空间范围,利用图斑的土地利用类别动态确定分割参数,是一种图斑边界控制的多参数分割方法。在此基础上利用多点接口(message passing interface,MPI)并行分割策略加速分割处理流程,有效提高了分割速度。实验结果表明,该技术能够获得更高的分割精度,为后续的土地利用现状信息提取和土地利用变化检测奠定良好的基础。
To overcome the block boundary extraction problem in remote sensing dynamic monitoring of land use, the authors have designed a land use patch boundary controlled mean shift segmentation algorithm making full use of the former land use vector data. This method uses the spatial extent of the patch polygons as the segmentation boundary constraint and selects the optimal segmentation parameter dynamically according to the land use type of the patch, thus being a multi -parameter segmentation method. The segmentation processing is also accelerated using MPI platform, which improves the running speed effectively. The experimental result shows that this method is capable of obtaining higher accuracy and can provide a good foundation for the subsequent land use information extraction as well as land use change detection.
出处
《国土资源遥感》
CSCD
北大核心
2013年第4期160-165,共6页
Remote Sensing for Land & Resources
基金
高分信息产品"生产线"系统先期攻关项目(编号:E0104/1112)
国土资源高分应用示范系统先期攻关项目(编号:E0202/1112)共同资助
关键词
土地利用动态遥感监测
均值漂移
MPI
分割参数
remote sensing dynamic monitoring of land use
mean shift
MPI
segmentation parameters
