资源简介

摘要： 从遥感图像监督分类结果到矢量对象的转换是遥感技术领域的一个瓶颈问题。提出了一种“分类图斑自适应概括”(Self-adaptive Generalization of Classified patch,SGCP)的方法,是针对这一问题的新尝试。SGCP能实现从破碎图斑到完整图像对象的自动转换,它由如下运算组成:①以形态学开启和形状系数分离道路与其他不透水表面;②以面积过滤和数学形态学操作去除噪声,以使图斑完整;③以递归凸残差回补简化图斑边界;④以膨胀和面积占优方法消除图斑裂隙;⑤以凸节点减少率评估图斑概括度,并同时以面积保持和分类精度保持评估概括精度;直至形成指定概括度的对象。概括运算的主要参数(如结构元素尺寸、递归次数、邻域窗口尺寸等)均由计算机自适应确定,同时预留部分用户调节参数,在自动概括的同时,允许人工干预概括程度。经Matlab仿真测试,该方法可以在保持分类精度与获取概括对象之间取得较好平衡。当图斑简化度上升22.9%时,面积平均变化仅为2.7%,分类精度仅平均下降0.72%。 关键词： 图斑概括;  二值形态学;  凸残差;  形状系数;  自适应参数化     Abstract: Pixels and patches allocated by supervised classification are often very scattered and cluttered.To make these classified patches more complete to form image objects and with only tolerable errors,a new algorithm,named Self\|adaptive Generalization of Classified Patch (SGCP),has been proposed.It uses some typical algorithms of both image analysis and map generalization,and has sure progress in the combined applications of the two.By using SGCP,it has been achieved to conduct generalization of classified patches of land use in a self\|adaptive way.Meanwhile,a better balance between degree and precision of the generalization can be promised.There are six steps to conduct SGCP:①performing a supervised classification in a multi\|descriptor space;②separating road from the rest patches of impervious surface via binary morphology operations and by using shape descriptors;③removing noises and making these patches more complete via binary morphology operations and by filtering out smaller patches;④simplifying the boundary of a patch by recursively backfilling its convex residuals into a convex hill of the patch;⑤eliminating gaps between patches by merging the gaps into surrounding larger patches and ⑥ assessing the generalization degree of a patch by checking the vertex reduction rate of a convex hill of the patch and,in the same time,assessing the generalization accuracy by checking whether the area of each class and the global accuracy of all the classes are maintained as well as possible.Some main parameters,such as size of structure element,recursion times,neighboring size etc.,are self\|adaptively determined.In addition,there are several reserved parameters to allow the degree of generalization adjustable by user as the user is unsatisfied with the results of automatic generalization.Simulation tests with Matlab show that by using the algorithm proposed in this paper,a proper balance between the degree and accuracy of generalization can often be guaranteed.An increase of 22.9% in the generalization degree will lead only a decrease of 0.72% in the global accuracy and a change of 2.72% in the patch area.

相关资源

1. 水质遥感监测的关键要素叶绿素a的反演算法研究进展
2. 基于无人机遥感技术的森林参数获取研究进展
3. 基于无线传感器网络的森林生态系统观测试验平台构建
4. 基于无人机激光雷达的森林冠层高度分析
5. 多源时序国产卫星影像的森林火灾动态监测
6. 基于无人机遥感技术的台风灾害倒伏绿化树木检测

下载说明

·如果您发现该资源不能下载，请通知管理员.gissky@gmail.com

·为确保下载的资源能正常使用，请使用[WinRAR v3.8]或以上版本解压本站资源,缺省解压密码www.gissky.net ，如果是压缩文件为分卷多文件，请依次下载每一个文件，并按照顺序命名为1.rar,2.rar,3.rar...，然后鼠标右击1.rar解压.

·为了保证您快速的下载速度，我们推荐您使用[网际快车]等专业工具下载.

·站内提供的资源纯属学习交流之用,如侵犯您的版权请与我们联系.