An object-based change detection approach using high-resolution remote sensing image and GIS data

Abstract

This paper proposed an automatic approach to change detection using GIS data and remote sensing images. The approach is based on an object-based SVM classification. A pixel-merge segmentation algorithm using spectral information and area size is utilized to generate image objects. Samples are calculated using remote sensing image and historical land use vector data automatically. Then, an object-based SVM classification is used on remote sensing images. Object boundaries originated from GIS are basic elements to calculating class percentage in per region. Comparing class percentage and historical class property, if the class percentage is large and different to historical property, these regions are identified as changed. The paper first introduced the general approach, and then defined and discussed the spectral channels used for the classification. The results of test areas are followed. Finally, experimental results confirmed the advantages and efficiency of the proposed approach.