楊智文1,2 ,陳金云2 ,張志遠3

　　(1.武漢大學遙感信息工程學院，湖北 武漢 430079;

　　2.重慶大學土木工程學院，重慶 400030;

　　3.水利部信息中心，北京 100053)

　　摘 要：針對傳統建立水庫相關數據庫的方法多為線下人工統計，受到諸多因素制約的缺陷，將網絡信息、電子地圖及遙感影像等數據綜合利用起來，實現水庫自動提取。利用網絡爬蟲從官方水利政務網站抓取相關信息，篩選出最新的水庫名稱及所屬地等數據，再根據水庫名稱及所屬地，調用百度、高德等地圖網站提供的應用接口，獲取水庫的空間坐標信息，并利用水庫坐標對遙感影像做緩沖區分析，提取出水庫水體范圍，得到1個含有水庫名稱、坐標、面矢量圖的水庫基礎數據庫。經過驗證，提取結果的查準率為0.9735，查全率為0.6113，作為二者的調和平均值達到0.751，能夠完整監測并提取出大中型水庫，但對小型水庫的監測提取效果一般，可解決傳統水庫提取方法需要先驗知識的問題，提高對水庫的區域性動態監測能力。

　　關鍵詞：水庫;自動提取;網絡信息;爬蟲;遙感影像;水體提取;地圖API

　　Research on automatic reservoir extraction method based on network information and remote sensing image

　　YANG Zhiwen1,2，CHEN Jinyun2，ZHANG Zhiyuan3

　　(1. School of Remote Sensing Information Engineering，Wuhan University，Wuhan 430079，China;

　　2. School of Civil Engineering，Chongqing University，Chongqing 400030，China;

　　3. Information Center， Ministry of Water Resources，Beijing 100053，China)

　　Abstract： In view of the shortcomings of traditional methods of developing reservoir-related database, which are mostly offline manual statistics and restricted by many factors, the study comprehensively uses data such as network information, electronic map and remote sensing image to realize automatic extraction of reservoirs’ information. The method uses web crawlers to grab relevant information from official water sector websites. The latest data such as the name and location of reservoirs are screened, and then, according to the name and location of the reservoir, the application interface provided by the map websites such as Baidu and Gaode is used to obtain the spatial coordinate information of the reservoir. The reservoir coordinates are used for buffer zone analysis on the remote sensing image to extract surface water area of the reservoir. Then a basic database of reservoir is obtained with regard to the name, coordinates and vector map of the reservoir. After verification, precision ratio of the extracted result is 0.9735, recall ration is 0.6113, and harmonic mean value of the two ratios is 0.751. The large and medium reservoirs can be completely monitored and extracted, however, the performance for small reservoirs is fair. This method solves the problem that priori knowledge is needed in traditional reservoir extraction methods and it improves the regional dynamic monitoring capability of reservoirs.

　　Key words：reservoir;automatic extraction;network information;crawler;remote sensing image;water body extraction;map API