论文部分内容阅读
通过一种卷积算法,从遥感分类数据中计算城市用地比例,并以该比例作为划分城市和非城市区的阈值,进而提取城市范围。方法构建了评价提取结果的指标体系,以确定卷积模板大小和阈值的选取。以北京为研究区,利用2007年SPOT5多光谱影像分类数据,对方法的可行性进行了验证。结合统计数据中的建成区面积,采用504种模板和阈值组合进行提取试验,讨论了模板和阈值对提取结果的影响。最终选择(205,51)的模板和阈值组合,对城市范围进行提取。研究结果表明:(1)该方法解决了遥感监测的城市用地分布区与城市范围存在差异的问题,为城市研究基础数据的准备提供了新的方法;(2)阈值过高或过低会造成城市范围的明显缩小或扩大;较小的模板不利于消除随机误差,较大的模板则会导致结果过于平滑;(3)对于SPOT5多光谱影像分类数据,模板大小在193—205像元(约4km2)、阈值接近50时,提取结果最好。
Through a convolution algorithm, the proportion of urban land from remote sensing classification data is calculated, and the ratio is used as the threshold for dividing urban and non-urban areas to extract the urban area. Methods The index system of evaluation results was constructed to determine the size of convolution template and the threshold selection. Taking Beijing as a research area, the feasibility of the method was verified by using the 2007 SPOT5 multi-spectral image classification data. Combined with the area of the built-up area in the statistical data, 504 templates and threshold combinations were used to extract the experiment, and the influence of template and threshold on the extraction results was discussed. The final selection (205, 51) of the template and the threshold combination, the city range extraction. The results show that: (1) This method solves the problem of the difference between urban land use and remote sensing monitoring urban area, and provides a new method for the city to prepare basic data; (2) The threshold is too high or too low (3) For SPOT5 multi-spectral image classification data, the size of the template is in the range of 193-205 pixels 4km2). When the threshold is close to 50, the extraction result is the best.