Improving spatiotemporal data fusion method in multiband images by distributing variates.

Area of Science:

• Remote Sensing
• Geospatial Analysis
• Image Processing

Background:

• Spatiotemporal data fusion is crucial for generating continuous fine-resolution satellite imagery.
• Existing methods face challenges in accurately capturing seasonal variations and handling landscape heterogeneity.

Purpose of the Study:

• To introduce the Residual Distribution-based Spatiotemporal Data Fusion Method (RDSFM) for enhanced fusion accuracy.
• To address residuals from spatial and temporal variations in satellite imagery.
• To minimize data requirements by using only one high-resolution reference image.

Main Methods:

• Utilized the IR-MAD algorithm to estimate subpixel distribution weights.
• Incorporated multivariate data collected over time to address residuals.
• Benchmarked RDSFM against unmixing-based data fusion (UBDF) using real satellite images.

Main Results:

• RDSFM accurately predicts seasonal variations in red and NIR bands, vital for vegetation analysis.
• The method effectively handles heterogeneous landscapes and dynamic land cover changes.
• Visual and quantitative assessments confirmed RDSFM's strong performance.

Conclusions:

• RDSFM offers significant advantages over existing spatiotemporal data fusion techniques.
• The method enhances fusion accuracy, particularly for vegetation monitoring in complex environments.
• RDSFM provides a robust solution for generating high-quality, fine-resolution satellite imagery with reduced data input.