面向对象的特征自动选择的建筑物信息提取

Extraction of Building Information Based on Object-oriented Feature Automatic Selection

针对建筑物信息提取存在低效率、高成本的问题,提出了一种利用分离阈值算法(seperability and thresholds,SEaTH)的高精度建筑物信息提取方法。采用高分二号遥感影像,通过半自动化信息提取构建分类规则的方法对天津市西青区的建筑物信息进行提取。通过运用SEaTH算法构建知识规则,选取训练样本并输出训练样本的特征值,将输出的特征值运用SEaTH算法进行自动确定阈值和特征优选,进而采用像素对象调整优化建筑物轮廓。将基于面向对象的最邻近分类法与该方法进行了精度评价对比。结果表明,该方法在提取建筑物信息时出现的错分漏分现象较少,且总体精度和Kappa精度都要高于基于面向对象的最邻近分类法,验证了其在提取建筑物信息方面的可行性。

In view of the low efficiency and high cost of building information extraction,a high-precision building information extraction method using seperability and thresholds(SEaTH)algorithm is proposed.This method uses GF-2 remote sensing image,and uses semi-automatic information extraction to construct classification rules to extract building information in Xiqing district,Tianjin.By using the SEaTH algorithm to construct knowledge rules,selecting training samples and outputting the feature values of the training samples,the output feature values are automatically determined by the SEaTH algorithm to automatically determine the threshold and feature optimization.It then adjusts and optimizes the building outlines by using pixel objects.The object-oriented nearest neighbor classification method is compared with the proposed method.The comparison results show that:the proposed method has fewer errors and misses in the extraction of building information,and the overall accuracy and Kappa accuracy are higher than that of the object-oriented nearest neighbor classification method,which verifies the feasibility of this method in extracting building information.