基于面向对象的珲春地区高分辨率遥感影像分类研究

以吉林省珲春市春化镇为研究区,以Pleiades、高分一号、资源三号影像为实验数据,利用面向对象信息提取方法实现了对3种遥感影像进行信息提取。利用3D Filter边缘检测算子对多尺度分割进行优化,通过对影像进行多次实验得出地物要素的最优分割参数,并且建立不同地物要素的分割层级。分析实验数据的特点构建了合理的分类层级,选取能区分各个地物要素的特征进行组合,利用阈值分类和模糊分类实现地物要素的信息提取。利用混淆矩阵对数据进行客观分析,得到3种影像的总体分类精度和kappa系数。分析结果表明:Pleiades影像分类精度较高,更适合本实验区的遥感影像信息提取。

High Resolution Remote Sensing Image Classification in Hunchun Area Based on Object Oriented

In this paper,Chunhua Town,Hunchun City,Jilin Province,as the study area,with GF-1 and ZY-3 and Pleiades sat ellite image of the experimental data.Completed using object-oriented method for high-resolution remote sensing images to extract information.Using Filter 3D edge detection operator to optimize the multi-scale segmentation,and get the optimal segmentation parameters of the ground object,and establish the different elements of the division level.Characteristics of experimental data to build their classification hierarchy,select the characteristics of each feature elements of feature combinations,using a threshold value classification and fuzzy classification realize feature extraction feature information.The overall classification accuracy and kappa coefficient of the three images are obtained by using the confusion matrix.The analysis results show that the accuracy of Pleiades image classification is the highest,and it is more suitable for the extraction of image information in the experimentation area.