Passive Non-line-of-sight imaging of moving targets using Physical embedding and Event-based vision

Passive Non-line-of-sight (NLOS) imaging has shown promising applications in imaging occluded objects around corners. However, this inverse problem is highly ill-posed and results in poor reconstruction with traditional physical retrieval methods, particularly in moving target imaging. With the development of neural networks, data-driven methods have greatly improved accuracy, however, heavy reliance on data volume has put great pressure on data collection and dataset fabrication. We propose a physical embedded passive NLOS imaging prototype with event-based vision (PNPE), which induces an event camera for feature extraction of dynamic diffusion spot and leverages simulation dataset to pre-train the physical embedded model before fine-tuning with limited real-shot data. The proposed PNPE is verified by simulation and real-world experiments, and the comparisons of data paradigms also validate the superiority of event-based vision in passive NLOS imaging for moving targets.