Spatially enhanced spatio-temporal fusion model for heterogeneity regions
PI Xinyu, ZENG Yongnian, WANG Pancheng
2023, 52(10):
1714-1723.
doi:10.11947/j.AGCS.2023.20220519
Asbtract
(
)
HTML
(
)
PDF (16360KB)
(
)
References |
Related Articles |
Metrics
With the development of remote sensing technology, remote sensing data has been increased rapidly. However, due to the limitation of sensors and the influence of cloud and rain weather, it is difficult for a single sensor to obtain remote sensing images with high spatial-temporal resolution, which affects the study of global and regional environmental change to a certain extent. The development of spatio-temporal fusion theory and technology of remote sensing image provides an effective way to solve this problem. In recent years, a number of spatio-temporal fusion algorithms have been proposed. However, there are still challenges to spatio-temporal fusion for accuracy and spatial detail of heterogeneity areas. Therefore, this paper proposes a spatially enhanced spatio-temporal fusion model for heterogeneity regions. Firstly, based on the principle of spectral mixing analysis and the assumption of spatial characteristics invariance of remote sensing data, the low-resolution spectral changes are downscaled to high-resolution. Secondly, based on the assumption of spectral invariance relationship of remote sensing data with different resolutions, the final fusion image is obtained. The experimental results show that compared with the commonly used STARFM and FSDAF models, the spatially enhanced spatio-temporal fusion model for heterogeneity regions can not only predict the phenological change information of different ground features effectively, but also preserves the spatial details of the ground surface and enhances the spatial characteristics and fusion effect in heterogeneous surface area; The mean values of root mean square error (RMSE), correlation coefficient (r) and structural similarity index (SSIM) of the spatially enhanced spatio-temporal fusion model reached 0.024, 0.898 and 0.897, respectively. Compared with the commonly used STARF and FSDAF models, the mean value of RMSE decreased by 6.71% and 4.33%, respectively; the average value of r increased by 1.95% and 1.74%, respectively; the average value of SSIM increased by 2.33% and 2.08%, respectively. The proposed spatially enhanced spatio-temporal fusion model for heterogeneity regions has the advantages of high fusion accuracy, simple and easy operation, especially in the heterogeneous surface coverage area. Therefore, the spatially enhanced spatio-temporal fusion model for heterogeneity regions has good prospects in remote sensing applications.