Impervious Surface Extraction Based on Improved Support Vector Data Description

Abstract

The continuous expansion of urban impervious surfaces has brought negative impacts on the urban environment. In order to quickly extract urban impervious surfaces to monitor urban development, this paper proposes a multiresolution segmentation-based impervious surface extraction method. The method is an improvement on the deep support vector data description method. The study is carried out to validate the method using some areas of Shenzhen as the experimental area. The experimental results show that the improved DSVDD method has enhanced all accuracy indicators, while its landscape pattern index reflects that the improved model has less fragmentation.