基于面向对象的北京市区城市内部用地信息提取

选取北京市四环以内为研究区域,以资源三号卫星遥感影像为数据源,针对类型多样、特征易混淆的城市内部用地以及高分辨率遥感影像海量信息、人工提取费时费力等特点,论文基于面向对象分类方法,探讨城市内部用地自动提取方法,并对分类结果进行精度评估.结果表明:利用不同地物的光谱、形状、纹理和空间关系等特征,通过多尺度分割和隶属度函数法,构建合理的分类层次,不仅精确提取出研究区内水体、绿地、建设用地和待开发用地,更独具创新地区分了城市建设用地内部各种地物类型,包括工业生产用地、低密度和高密度生活用地以及交通用地.该方法有效地利用了资源3 号卫星影像的光谱、纹理及空间信息特征,总体精度可达到87.00%,Kappa系数达到0.853 9,取得较好的分类效果.

Land Use Information Extraction in the Inner City of Beijing Based on Object-oriented Classification Method

Land use patterns in the inner cities are often complex and diverse. This paper firstly develops a technical method specially for extracting land information of the inner city based on the method of object-oriented classification, and then collects Resources Satellite No.3 Image to extract land information of the urban areas within the Fourth Ring Road in Beijing city and evaluates the accuracy of the classification result. The results show: by means of multiresolution segmentation, membership function and reasonable classification levels, and by use of the characteristics of spectrum, shape, texture and spatial relationships of different objects, objectoriented method can not only extract water, green land, construction land and undeveloped land, but also extract industrial land, low-intensity and high-intensity residential land, and land to be developed in the study area. The paper takes full use of the characteristics of the high resolution image. The overall accuracy achieves 87.00%, and the Kappa coefficient is 0.8539. The classification results are acceptable.