省域精细地表覆盖数据智能化生产方案与实践

Scheme and Practice of Intelligent Production of Provincial Scope Fine Scale Land Cover Data

为了实现基于亚米级高分遥感影像的地表覆盖数据智能化生产,在“图谱耦合”遥感认知理论和“分区-分层”地学思想指导下,研究了省域范围精细地表覆盖数据的技术路线、地类样本设计、模型体系构建方案和矢量图斑后处理策略,并成功应用于江苏全省地表覆盖数据生产。通过比较验证:整体分类精度达到90%;耕地、建筑、林地、水面等典型地物的图斑形态接近人工勾画效果;相对于传统人工作业模式,效率提升8倍。实践表明,该生产体系技术路线可行,具有一定的推广应用价值。

In order to realize the intelligent production of land cover data based on sub-meter high resolution remote sensing images,the technical route,sample rules,model system construction scheme and vector data post-processing strategy of producing provincial fine scale land cover data were studied under the guidance of Coupled Spectral and Spatial geoscientific thought.These studies were successfully applied to the production of fine land cover data in Jiangsu Province.The evaluation result shows that the overall classification accuracy reaches 90%,the shape of typical features such as crop land,buildings,forest land and water surface are close to the effect of manual operation,the efficiency is improved by 8 times compared with the traditional manual operation mode.It can be seen that the technical route of this production system is feasible and has the value of promotion and application.