基于面向对象的多尺度山地城市土地覆盖信息提取——以重庆市大学城城区为例  

Research on Multi-scale Mountainous City Land Cover Information Extraction Based on Objection-oriented——Taking Chongqing University Town as an Example

作　　者：何顺兵 牟凤云[1] 

机构地区：[1]重庆交通大学建筑与城市规划学院,重庆400074

出　　处：《重庆工商大学学报（自然科学版）》2016年第1期83-88,共6页Journal of Chongqing Technology and Business University:Natural Science Edition

基　　金：国家科技支撑计划课题(2012BAH28B04);国家自然科学基金(41201546)

摘　　要：以重庆市大学城区的Quick Bird高分辨率遥感影像为数据源,基于多尺度分割算法获得内部异质性较小的对象,构建典型地物影像特征库,实现不同地类特征的针对性描述,采用级联方式综合最邻近分类算法和模糊分类算法优点,实现山地城市土地覆盖信息的面向对象提取,并把提取结果与传统监督分类方法进行对比分析;结果表明:综合使用地物的光谱、纹理、几何等特征的面向对象分类法提取结果稳定性好,分类效率高;面向对象分类能够有效避免"椒盐"噪声,分类结果具有很好的区域性和连贯性;面向对象分类法对于山地城市中光谱特征特别相近或相同的耕地和草地、建筑用地和道路的分类,精度分别提高了28.67%、9.24%、10.38%和8.79%,有效地克服了同谱异物和同物异谱现象;根据对比分析结果得知,面向对象分类法在山地城市土地覆盖信息提取中具有较高的可行性和一定的应用价值,同时由于研究区域的复杂性和研究对象的普适性,算法研究成果可以推广应用到其他山地城市区域。By taking QuickBird high-resolution remote image of Chongqing University Town as data source, based on multi-scale segmentation algorithm to achieve inner object with small difference, the typical land image feature database is constructed to implement the pertinent description of different land features, the cascade method is used to comprehend the advantage of nearest neighboring algorithm and fuzzy classification algorithm, the objectoriented extraction of mountainous city land cover information is implemented, and the extracted results arc analyzed and compared with traditional monitoring classification method. The results show that the object-oriented classification of comprehensive utilization of spectral features, texture features and geometry features has higher classifying efficiency and stability, that this technology can effectively overcome the salt and pepper noise, that its classification result has good area feature and connection, that the classification accuracy of the object-oriented to arable land, grass land, construction land and road land with near or the same spectral features in mountainous city increases 28.67%, 9.24%, 10.38% and 8.79% respectively to effectively overcome the phenomenon of the same spectrum from different materials and the same materials with different spectrum. Comparison and analysis reveal that object-oriented classification method has good feasibility and certain application value in the mountainous city land cover information extraction, meanwhile, because of the complexity of the studied area and the universality of studied object, this algorithm research can be used in other mountainous cities.

关 键 词：多尺度分割 面向对象 模糊分类 山地城市 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]