建筑物半自动提取方法的研究

发布时间：2018-06-20 14:59

本文选题：建筑物半自动提取 + 种子填充　；参考：《中国测绘科学研究院》2013年硕士论文

【摘要】：随着航空航天技术越来越成熟,各种高分辨率卫星传感器相继发射。面对大量的遥感影像数据,高分影像提取建筑物的生产领域中,主要通过作业员的人工识别,人工逐点、逐段进行采集勾绘建筑物,劳动强度大,工作繁琐；另一方面,全自动方式的地物识别提取,难度高,问题多,精度低且极其复杂,因此到目前为止还尚未实现。就目前技术水平而言,采用人机交互方式进行半自动提取是比较现实可行的。本文中首先研究了目前建筑物自动提取中最新提出的几种自动提取算法,包括GLC树算法和像元形状指数算法；接下来研究了比较传统的几种半自动提取算法,包括有优化路径的种子填充算法、区域增长算法和改进的区域填充扫描线算法。提出一种改进的建筑物半自动提取策略,并根据研究的半自动提取算法进行一定的改进,提出一种基于建筑物半自动提取的改进的种子填充算法。改进的建筑物半自动提取策略,首先针对高分辨率遥感影像中,建筑物自身形状的复杂性,分别利用GLC树算法和eCognition中规则集算法进行建筑物自动提取,对两种算法经过算法流程简便性和提取结果精度之间的比较,最终选择GLC树算法的提取结果作为建筑物自动提取的最终结果；对提取结果中的误分、漏分斑块采用改进的种子填充算法进行人工干预修正；半自动提取的矢量数据作为参考数据参与到建筑物自动提取中,再次对建筑物进行自动提取,保证建筑物提取的精度。该建筑物半自动提取实现解决现实生产中人工勾画提取建筑物过于繁琐的问题,并提高生产中建筑物提取的工作效率。最终形成一套人机交互半自动提取建筑物的工作流程和基于FeatureStation GeoEX智能解译平台的建筑物半自动提取插件。
[Abstract]:As aerospace technology becomes more and more mature, various high-resolution satellite sensors are launched one after another. Faced with a large number of remote sensing image data, high-score image extraction of buildings in the field of production, mainly through the manual identification of operators, manual point by point, section by section to collect and draw buildings, labor intensity, work tedious; on the other hand, The automatic method of feature recognition and extraction is very difficult, has many problems, low precision and extremely complex, so it has not been realized yet. In terms of current technical level, it is practical and feasible to use human-computer interaction for semi-automatic extraction. In this paper, we first study several new automatic extraction algorithms, including GLC tree algorithm and pixel shape index algorithm, and then study some traditional semi-automatic extraction algorithms. It includes seed filling algorithm with optimized path, region growing algorithm and improved region filling scan line algorithm. An improved semi-automatic extraction strategy for buildings is proposed, and an improved seed filling algorithm based on semi-automatic extraction of buildings is proposed. Based on the improved semi-automatic extraction strategy of buildings, first of all, in view of the complexity of the shape of buildings in high-resolution remote sensing images, the GLC tree algorithm and rule set algorithm in Cognition are used to extract buildings automatically, respectively. The GLC tree algorithm is chosen as the final result of building automatic extraction after comparing the simplicity of the algorithm flow and the precision of the extraction results. The improved seed filling algorithm is used for artificial intervention and correction; the semi-automatic vector data is used as reference data to participate in the automatic extraction of buildings, and then the building is automatically extracted again to ensure the precision of building extraction. The semi-automatic extraction of the building can solve the problem that the artificial drawing and extracting the building is too cumbersome and improve the efficiency of the building extraction in the production. Finally, a set of semi-automatic building extraction workflow based on human-computer interaction and a semi-automatic building extraction plug-in based on FeatureStation GeoEX intelligent interpretation platform were formed.
【学位授予单位】：中国测绘科学研究院
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P237

【相似文献】