POI辅助下的高分辨率遥感影像城市建筑物功能分类研究

发布时间：2018-08-08 18:03

【摘要】：城市建筑物是城市的重要组成部分,对城市建筑物进行功能分类可以为城市功能区划分提供有利依据,辅助政府部门对城市规划、土地利用、资源、人口等方面的分布与分配进行管理与决策,有助于推进城镇化建设的可持续发展。仅利用目前的遥感分类技术难以对高分辨率遥感影像的城市建筑物信息进行功能分类,然而将遥感、互联网兴趣点(Point of Interest,POI)数据以及GIS技术有效地结合在一起,可以更为细致地分析城市信息,不仅实现了建筑物功能分类,而且提高了分类的准确率与可信度。本文首先选取卷积神经网络(Convolutional Neural Networks,CNN)方法对高分辨率遥感影像数据进行建筑物提取;然后,对POI数据的城市商服、公建和住宅用地进行核密度分析;最后分别统计每个建筑物3种用地的核密度平均值,并将该值设置为此建筑物的属性值,并结合POI数据的实际情况选择具有最佳功能分类精度的属性值作为阈值提取3种用地信息,从而完成不同功能的城市建筑物分类。精度评价结果表明,该方法对3种用地的提取效果良好,分类精度达到86%以上。
[Abstract]:Urban buildings are an important part of a city. The functional classification of urban buildings can provide a favorable basis for the division of urban functional areas and assist government departments in urban planning, land use, and resources. The distribution and distribution of population will help to promote the sustainable development of urbanization. It is difficult to classify urban building information in high-resolution remote sensing images by using only the current remote sensing classification technology. However, remote sensing, (Point of Interest-POI) data and GIS technology are effectively combined. It not only realizes the classification of building function, but also improves the accuracy and credibility of classification. In this paper, we first select convolutional neural network (Convolutional Neural Networks) method to extract the high-resolution remote sensing image data, and then analyze the nuclear density of the urban commercial service and the public housing land for the POI data. Finally, the average nuclear density of three kinds of land for each building is counted, and the value is set to be the property value of the building. According to the actual situation of POI data, the attribute value with the best classification precision is selected as the threshold value to extract three kinds of land information, so as to complete the classification of urban buildings with different functions. The accuracy evaluation results show that the method has a good effect on the extraction of three kinds of land, and the classification accuracy is over 86%.
【作者单位】：中国科学院遥感与数字地球研究所;中国科学院大学;
【基金】：高分辨率对地观测系统重大专项(03-Y20A04-9001-15/16) 国家自然科学基金项目(41601387)
【分类号】：P237

【相似文献】