Dynamic vision-based underwater optical signal detection system in a degraded environment using multi-dimensional...

Area of Science:

• Optics
• Computer Vision
• Signal Processing

Background:

• Underwater optical signal detection is difficult in turbid and occluded conditions.
• Existing frame-based methods struggle with severely degraded underwater environments.

Purpose of the Study:

• To propose and evaluate a novel dynamic vision-based system for underwater optical signal detection.
• To assess the system's performance in degraded underwater conditions like turbidity and partial occlusion.

Main Methods:

• Utilized an event camera array for dynamic vision sensing.
• Employed multi-dimensional integral imaging to capture optical signals.
• Processed event sequences into multi-dimensional event videos.
• Classified videos using a Vision Transformer and Gated Recurrent Unit network (ViT-GRU).

Main Results:

• The dynamic vision system demonstrated superior detection performance compared to frame-based approaches.
• Performance was evaluated using Matthew correlation coefficient and error symbol count.
• The system successfully detected optical signals transmitted through degraded underwater environments.

Conclusions:

• The proposed dynamic vision-based system with integral imaging and ViT-GRU network is effective for underwater optical signal detection.
• This represents the first report of dynamic vision-based underwater optical signal detection utilizing multidimensional integral imaging.