High spatiotemporal resolution remote sensing images can provide on landform changes information fast and accurately, which have a wide range of applications and needs in many areas such as agricultural monitoring and urban planning and construction. However, due to the limitations of sensor hardware, remote sensing images have the phenomenon that both high spatial resolution and high temporal resolution are not compatible. In view of the complementary advantages of information between images from different sensors, the fusion of remote sensing images is a very meaningful direction. Based on sparse representation theory, this paper proposes a double-layer spatiotemporal fusion framework suitable for single-phase high-resolution remote sensing images. The Landsat8 OLI and MODIS remote sensing image are used as experimental data to fully analyze the method, and compared with the classical spatiotemporal fusion methods of STARFM, and analyze the impact of down sampling the multiple of down sampled in the double-layer on the fusion results. The experimental results show that our method has higher prediction accuracy, and the experimental results are best when the multiple of the down-sampled is four.
|Number of pages||9|
|Journal||Journal of Network Intelligence|
|Publication status||Published - Aug 2019|
- Landsat8 OLI
- RSparse representation
- Spatiotemporal fusion