.png){kind=icon}
.png){kind=icon}
Return
Three-dimensional computational ghost imaging with only single-pixel detection based on deep learning preprocessing pseudo-thermal light
.png){kind=icon}
PRE
AI
DOI：10.1016/j.optlastec.2025.113680.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Conventional three-dimensional computational ghost imaging (3D-CGI) methods, such as time-of-flight (ToF) LiDAR and multi-camera systems, typically rely on expensive high-speed detectors or complex multi device setups, which limits their scalability and practicality in different application scenarios. In this paper, we propose a simple and effective approach that combines deep learning optimization of speckle signals with cross-correlation ranging model for 3D reconstruction. The method eliminates the need for high temporal resolution detectors or multi-sensor configurations while enabling accurate multi depth decoupling. Experimental results demonstrate that, the proposed scheme achieves sub-meter ranging accuracy at kilometer-scale distances even under the interference of scattering medium. Moreover, 3D reconstruction simulations verify that the method maintains high accuracy and robustness to target offset (±20 %) and scale variation (±30 %). These results provide an effective solution for real-time 3D perception in complex environments.
Keywords:
deep learning
computational ghost imaging
3D reconstruction
cross-correlation ranging
speckle signals
Journal
O
IF:
5
Papers:
1.9K
Citations:
3.5W