基于高分一号影像的天津市宝坻区地表覆盖分类研究

Study on Landcover Classification in Baodi District Tianjin City Based on Gaofen-1 Satellite Image

由于国产高分一号卫星WFV相机的16 m数据获取具有免费、时效性好等特点而被广泛应用于遥感信息提取。面向对象的分类方法能够解决基于传统像素的分类方法所具有的破碎性强、易产生同谱异物现象的问题。本研究采用基于面向对象的分类思路,以天津市宝坻区为研究区域,通过获取高分一号卫星影像并进行预处理,利用遥感信息处理软件eCognition对该数据进行影像分割,并对影像分割斑块进行基于规则的分类,提取出宝坻区范围内植被、水体、建设用地、裸地四类地物,并利用高分辨率影像对比进行精度验证,分类总精度达到93%。结果表明:利用高分一号国产卫星16 m影像,采用面向对象的分类思路进行地表覆盖类别的信息提取,可以快速便捷地实现目标区域空间地理信息的获取,对于监测地表覆盖面积和分布状况起到指导作用。

The 16 m data of WFV camera of domestic Gaofen-1 satellite is widely used in remote sensing information extraction due to its free access and good timeliness.The object-based classification method can solve the problem that the traditional pixel-based classification method has strong fragmentation and is easy to produce the same spectrum of different objects.In this study,we use object-oriented classification method and taking Baodi district Tianjin city as the research target,acquiring and preprocessing GF-1 satellite image,using eCognition software to process segmentation of image,and classify the segmentation image with rule-based extraction method,and extract the vegetation,water,construction land and bare land of Baodi district.Using higher resolution image to verify classification result,the overall accuracy of accuracy verification results reaches 93%.Results showed that using 16 m resolution of GF-1 satellite images with the object-based method to extract landcover categories could be help extracting geographic information target region quickly and easily,and played a guiding role on monitoring the landcover area and distribution.