Light field-guided optical synthetic aperture imaging system on master-slave UAVs

Area of Science:

• Optics and Photonics
• Remote Sensing Technology
• Robotics and Autonomous Systems

Background:

• High-resolution, large field-of-view (FoV) aerial imaging is crucial for remote sensing, astronomy, and geology.
• Dynamic relative poses of unmanned aerial vehicles (UAVs) challenge accurate large-FoV image stitching.
• Existing methods struggle with the dynamic nature of multi-UAV formations for imaging.

Purpose of the Study:

• To develop a light field-guided optical synthetic aperture imaging system using master-slave UAVs.
• To address the inaccuracy in large-FoV image stitching caused by dynamic UAV poses.
• To enable high-resolution and large-FoV aerial image and video generation.

Main Methods:

• A master-slave UAV formation captures cross-scale multi-view aerial images.
• Image stitching is reframed as novel-view synthesis on a large-FoV light field.
• A 5D light field is constructed using Gaussian splatting, guiding synthetic aperture imaging.

Main Results:

• The system successfully generates accurate, high-resolution images with a large FoV.
• The proposed method is extendable for generating large-FoV videos.
• Experimental results validate the system's capability in real-world scenarios.

Conclusions:

• The light field-guided synthetic aperture imaging system offers a robust solution for large-FoV high-resolution aerial imaging.
• This technology has significant potential for applications in remote sensing and beyond.
• The novel approach overcomes limitations of dynamic pose variations in multi-UAV systems.