基于面向对象法的方祥乡地类信息提取方案设计

Design of Object-Oriented Ground Class Information Extraction Scheme for FangXiang

地物信息的提取中应用到面向对象方法的软件eCongintion,利用地物相应特征来辅助地物信息的提取最后输出的矢量数据也易于与ArcGIS相结合。研究中使用的面向对象方法主要包括2个关键步骤:多尺度分割和分类,使用一种快速的分割方法得到最终分割结果,然后使用最邻近法进行对地物的分类,实现高分辨率遥感影像地物信息的提取。完成地物信息提取之后,开始对分类结果精度评价,使用得出的结果与基于像元的分类方法进行了对比,面向对象方法所得出的总体精度比传统方法所得出的总体精度高出18.15%,结果表明,面向对象方法得出分类结果精度高于基于像元的分类方法。

With the wide application of high-resolution remote sensing images,how to use it to extract the information of ground objects has obviously become one of the hotspots and difficulties in the field of remote sensing applications.In the traditional classification method,which is easy to ignore the image texture features and structural features in the process of extracting information,the classification result is prone to“salt and pepper”phenomenon.Therefore,in the classification of high-resolution remote sensing images today,the traditional classification methods are no longer practical.The extraction of feature information in this study is applied to the object-oriented classification software eCongintion,which uses features of the feature to assist in the extraction of feature information.The final output vector data is also easily combined with ArcGIS.The object-oriented method used in this study mainly includes two key steps,multi-scale segmentation and classification.A fast segmentation method is used to get the final segmentation result,and then the nearest neighbor method is used to classify the objects and achieve high-resolution remote sensing.After the feature information extraction is completed,the accuracy of the classification results is evaluated,and the results、are compared with the pixel-based classification methods.In contrast,the overall accuracy of the object-oriented method is 18.15%higher than that of the traditional method.The result shows that the object-oriented method has better classification accuracy than the pixel-based classification method.