Three-dimensional computational ghost imaging using a dynamic virtual projection unit generated by Risley prisms.

Area of Science:

• Optics and Photonics
• Computational Imaging
• Machine Learning for Computer Vision

Background:

• Stereo vision computational ghost imaging (CGI) typically requires multiple spatially separated projection units or detectors for 3D imaging.
• Existing methods can be complex and bulky, limiting practical applications.

Purpose of the Study:

• To develop a compact and flexible 3D CGI system.
• To enhance the angular resolution and accuracy of 3D reconstruction in CGI.

Main Methods:

• A compact 3D CGI system utilizing Risley prisms, a stationary projection unit, and a bucket detector was designed.
• Multi-view ghost images were acquired by rotating the Risley prisms to simulate a dynamic virtual projection unit.
• A convolutional neural network (CNN) for super-resolution (SR) was employed to improve angular resolution.
• An optimized 3D CNN was implemented for disparity estimation and final 3D reconstruction.

Main Results:

• The system successfully generated multi-view ghost images for 3D imaging.
• The SR-CNN enhanced the angular resolution of the reconstructed images.
• The 3D CNN effectively performed disparity estimation and 3D reconstruction.
• Experimental validation confirmed the system's effectiveness and the method's potential.

Conclusions:

• The proposed compact 3D CGI system with Risley prisms offers a flexible and effective approach to 3D imaging.
• The integration of SR-CNN and 3D CNN significantly improves image quality and reconstruction accuracy.
• This technology shows promise for applications in navigation, detection, and other fields requiring 3D imaging.