中文摘要：

随着高分辨率遥感影像在矿山分类工作中的广泛应用，面向对象的多尺度分割技术成为研究热点。最优尺度的选取是此类技术的关键。为获取高分遥感影像最适宜的分割尺度，本文以赣州稀土矿山高分辨率遥感影像为源数据，选择与邻域绝对均值差分方差比方法RMAS（RatioofMeanDifferencetoNeighbors（ABS）toStandard Deviation）方法获取最优分割尺度。通过分割质量值对分割结果进行评价分析，研究验证了RMAS方法的可行性。

英文摘要：

Sincehigh resolution remote sensing image is widely used in the work of mining classification, the technologies of object - oriented multi - scale segmentation become a hot topic. The selection of the optimal scale is the key to such technology. In order to obtain the optimal segmentation scale for high resolution sensing image, high - resolution remote sensing images of Ganzhou rare earth mines are selected as source data and RMAS （Ratio of Mean Difference to Neighbors （ABS） to Standard Deviation） method to obtain the optimal scale in the paper. Segmentation results are evaluated by segmentation quality value, and the study proves the feasibility of RMAS method.