Fusing Multi-Temporal Context for Image Super-Resolution Reconstruction in Cultural Heritage Monitoring

Area of Science:

• Remote Sensing
• Geospatial Analysis
• Cultural Heritage Management

Background:

• Effective conservation of World Heritage Sites requires precise, continuous monitoring.
• Challenges like cloud cover and sensor resolution limit acquiring high-resolution imagery for temporal analysis.
• Accurate characterization of heritage site evolution is hindered by data gaps.

Purpose of the Study:

• To develop a temporal change-aware super-resolution reconstruction model.
• To overcome limitations in acquiring high-resolution imagery for key years.
• To enable accurate temporal evolution characterization of heritage landscapes.

Main Methods:

• Proposed a multi-branch architecture incorporating target year low-resolution images and adjacent year high/low-resolution images.
• Employed parallel encoding branches for spatial feature extraction and explicit temporal change pattern modeling.
• Developed a temporal change encoder to capture landscape evolution between years.

Main Results:

• The model successfully reconstructs high-quality imagery for a target year using temporal information.
• Experimental results on Weiyang Palace core area data demonstrated superior performance.
• Outperformed traditional single-image super-resolution and models lacking explicit temporal change modeling.

Conclusions:

• The proposed temporal change-aware super-resolution model effectively enhances heritage site monitoring.
• Utilizing temporal evolution information is a key strategy for reconstructing high-quality imagery.
• The method offers a significant advancement for dynamic monitoring and conservation of World Heritage Sites.