Extracting Impervious Surfaces From Multi-source Remote Sensing Data

Impervious surface is a major indicator of rapid urbanization,which leads to urban waterlogging.The urban impervious surface extraction using remote sensing image is of great significance to analyse the urban spatial pattern and guide the sponge city construction.This paper takes the advantages of multi-spectral satellite imagery and LiDAR data based on Grabcut to extract impervious surfaces.The main research is carried out as follows:(1)The characteristics of spectral features and LiDAR point cloud are analyzed.Spectral features are extracted from multi-spectral satellite imagery by index counting and Tasseled Cap transforming,and roads are extracted by filtering from elevation and intensity information of LiDAR point cloud.(2)Graph cut and Grabcut algorithm are studied.To take the advantages of multi-source features,this paper improve Grabcut from three respects as follows:energy function,initialization and feature space.(3)The experiments of impervious surfaces extraction are carried out.The results of this paper's algorithm and other traditional algorithm are contrasted and discussed.Taking the city of Guangzhou province as the example,the results showed that the method gained higher overall accuracy and robustness compared with traditional single-source method.