城郊高分影像中利用阴影的建筑物自动提取

发布时间：2018-06-24 20:02

本文选题：高分辨率遥感影像 + 建筑物提取　；参考：《武汉大学学报(信息科学版)》2017年10期

【摘要】：提出了一种充分利用阴影实现自动分类与后处理相结合的建筑物自动提取方法:首先根据阴影和植被自动检测结果并选定裸地样本确定预分类CMap图,并设计了基于偏移阴影分析的建筑物样本自动提取方法,结合支持向量机(support vector machines,SVM)分类模型将影像分为阴影、植被、建筑物、裸地4大类以提取建筑物初始结果;通过形态学处理提升区域完整性,区域增长补充漏检区域,利用设计的相交边界阴影比率筛除无阴影的非建筑物等措施,进行后处理优化获取最终结果。实验表明,充分利用阴影信息,不仅能准确、全面地获取各类样本,保证分类精度,与后处理优化策略紧密结合,大幅度提高了正确率和完整度;并且自动化程度得到有效提高,更适用于城郊区域建筑物的提取。
[Abstract]:In this paper, a building automatic extraction method which combines automatic classification and post-processing with shadow is proposed. Firstly, according to the results of shadow and vegetation automatic detection, the pre-classified CMap map is determined by selecting bare land samples. An automatic extraction method of building samples based on offset shadow analysis is designed, and the image is classified into four categories: shadow, vegetation, building and bare land in order to extract the initial results of the building with support vector machine (SVM) classification model. Through morphological processing to improve the integrity of the region, regional growth to supplement the missing areas, the design of the intersection boundary shadow ratio to screen out the shadow of the non-building, post-processing optimization to obtain the final results. Experiments show that the full use of shadow information can not only accurately and comprehensively obtain all kinds of samples, ensure the classification accuracy, and closely combine with the post-processing optimization strategy, and greatly improve the accuracy and integrity; And the degree of automation has been effectively improved, more suitable for the extraction of suburban buildings.
【作者单位】：长江大学地球科学学院;长江水利委员会长江科学院;武汉市房产信息中心;
【基金】：国家自然科学基金(41371343,31571565)~~
【分类号】：P237

【参考文献】