多尺度多准则的遥感影像线状地物信息提取

ZY-3 linear features information extraction based on the multi-scales and multi-criterion

针对传统低层次单个像元纯光谱的信息提取方法存在的不足,以ZY-3高分辨率遥感影像为数据源,依据"种内一致性最大、类间异质性最大"原则,提出了基于多尺度多准则的ZY-3影像线状地物信息提取方法。利用基于最优分割尺度的影像分割、面向对象线状地物目标信息的提取方法,实现了河流、水渠、道路等线状地物信息的提取,并通过实验分析验证了可行性、有效性。结果表明:构建的RMAS指数指标,可以作为多尺度分割中选择最优分割尺度的一种很好的指示器,所提方法不仅有效避免了分割斑块面积对分割结果评价的影响,还能够有效识别ZY-3影像中的线状地物信息,研究结果对高分辨率遥感影像处理尺度依赖的技术难点及线状地物信息提取等研究具有重要的参考价值。

This paper analyzed the deficiency of traditional information extraction method based on the single pixel and pure spectral, with the ZY-3 high resolution remote sensing image as data source, accord- ing to the kind of consistency within the maximum class heterogeneity between maximum principle. The paper puts forward to the ZY-3 linear features information extraction method , which based on the multi- scales and multi-criterion, using image segmentation, which based on the optimal segmentation scale, object-oriented information extraction method of linear target features, and realizes the river, canal and road linear feature information extraction, the feasibility and validity is verified by the experiment. The re- sults indicates that RMAS index can be used as a good indicator of selecting the optimal segmentation scale in multi-scale segmentation, the proposed method not only effectively avoid the impact of segmentation plaque area on the segmentation result evaluation, also can effectively identify the linear feature informa- tion of ZY-3 images, These results provide a reference for the research of technological difficulty of high resolution remote sensing images processing which is scale-dependent and linear feature information.