Detecting spatio-temporal and typological changes in land use from Landsat image time series

Abstract

With the development of earth observation technology, the spatial and temporal resolution of remote sensing images have improved dramatically, providing abundant data for detecting land use change in detail. However, extracting spatio-temporal and typological changes in land use from remote sensing image time series is still challenging. Landsat image time series and land survey data are combined to develop a method to detect in detail the time and types of land use change. From the Landsat images, the time series of normalized difference vegetation index variation slope (TSNVS) on each pixel was constructed. The average TSNVS of sample pixels was used as the reference series. Based on TSNVS, the change in land use at six time points (S1 to S6) was detected by measuring the similarity of the time series on undetermined pixels and reference series. Land survey data were used to identify detailed types of land use change. The 16 different types of land use change were identified. The results indicate that the proposed method is effective at detecting spatio-temporal and typological changes in land use from Landsat image time series. After verification, the overall accuracy was 92%, and the kappa coefficient was 0.9.