Remote Sensing Target Tracking Method Based on Super-Resolution Reconstruction and Hybrid Networks.

Area of Science:

• Computer Vision
• Remote Sensing Technology
• Artificial Intelligence

Background:

• Remote sensing images present challenges like high complexity, distortion, and large-scale variations.
• Nonlinear motion features of remote sensing targets hinder accurate tracking with existing methods.
• Current algorithms often result in significant tracking errors, impacting data utility.

Purpose of the Study:

• To address the limitations of current remote sensing target tracking algorithms.
• To propose a novel target tracking method that enhances accuracy and reduces errors.
• To leverage super-resolution techniques for cost-effective high-resolution image acquisition and analysis.

Main Methods:

• A super-resolution reconstruction network is employed to enhance the resolution of remote sensing images.
• A hybrid neural network is utilized for estimating target motion post-detection.
• The Hungarian algorithm is applied for efficient identity matching of tracked targets.

Main Results:

• The proposed method achieved a tracking accuracy of 67.8%.
• The recognition identification F-measure (IDF1) value reached 0.636.
• Performance indicators demonstrate superiority over traditional target tracking algorithms.

Conclusions:

• The developed super-resolution hybrid network effectively improves remote sensing target tracking accuracy.
• The method meets the requirements for precise tracking of targets in complex remote sensing scenarios.
• This approach offers a cost-effective solution for obtaining high-resolution imagery for tracking applications.