基于分割的多尺度城市绿地景观

利用多分辨率分割生成3种尺度的城市绿地景观.以50像元和300像元的目标分割值生成小尺度和大尺度的景观影像目标层，利用最近邻法分类两个目标层.对小尺度景观分类后的结果进行分割生成中间尺度的景观影像目标层并进行分类.从3个景观尺度的分类目标层中矢量化提取出绿地信息并计算6个景观指数：多样性、优势度、均匀度和分维数、破碎度和内缘比.多样性、优势度、均匀度和分维数的最大值分别为2.2、0.681、0.948和0.326，最小分别为1.641、0.122、0.707和0.113.表明多样性、均匀度和破碎度随着景观尺度的增大而减小，优势度随着景观尺度的增大而增大.基于分割产生多尺度景观的方法可以满足城市绿地景观研究的需求.

Segmentation-based multi-scale urban green space landscape

In this study, three scales urban green space landscapes were generated by multi-resolution segmentation. With 50 and 300 pixels as the object segmentation thresholds, the small- and large-scale landscape object image layers were produced, and the two object image layers were obtained by the nearest neighbor classification method. The result of small-scale landscape classification image was segmented into middle-scale landscape image, and then classified. Green space information was extracted through vector form of object image layers of three scales landscape classification. The landscape indexes diversity, dominance, evenness, fractal dimension, fragmentation, and interior to edge ration were calculated, with the largest values of the former four indexes being 2. 2, 0. 681, 0. 948, and 0. 326, and the smallest values being 1. 641, 0. 122, 0. 707, and 0. 113, respectively, indicating that the diversity, evenness and fragmentation decreased, while the dominance increased with increasing land- scape scale. The method of multi-resolution segmentation to generate multi-scale landscape could meet the needs of urban green space landscape research.