An Improved Spatiotemporal Fusion Approach Based on Multiple Endmember Spectral Mixture Analysis

Area of Science:

• Remote Sensing
• Geospatial Analysis
• Image Processing

Background:

• High spatial and temporal resolution remote sensing data is crucial for land use/cover mapping and biophysical parameter retrieval.
• Limitations in sensor performance and weather conditions hinder the acquisition of high-resolution remote sensing images.
• Spatiotemporal fusion models, like STARFM, are vital for overcoming these limitations.

Purpose of the Study:

• To address the uncertainty in neighboring similar pixel selection within existing spatiotemporal fusion models, particularly in heterogeneous areas.
• To improve the accuracy of remote sensing image fusion.
• To propose an improved Enhanced Spatial and Temporal Adaptive Reflectivity Fusion Model (I-ESTARFM).

Main Methods:

• Modification of the neighboring similar pixel selection procedure in the ESTARFM method.
• Utilizing land cover endmember types and fraction values from spectral mixing analysis for accurate pixel selection.
• Experimental validation and comparison with STARFM and ESTARFM models.

Main Results:

• The I-ESTARFM method demonstrates more accurate selection of neighboring spectrally similar pixels compared to STARFM and ESTARFM.
• Fused images generated by I-ESTARFM show increased correlation coefficients and decreased mean square errors against actual images.
• Significant improvements in fusion accuracy were observed, especially in spatially heterogeneous regions.

Conclusions:

• The I-ESTARFM method effectively reduces uncertainty in spectral similar neighborhood pixel selection.
• The proposed method enhances the precision of spatial-temporal fusion for remote sensing data.
• I-ESTARFM offers a more reliable approach for land cover information extraction and biophysical parameter inversion.