Abstract
Intelligent visual tracking is a hot research topic in the field of computer vision, which not only has high scientific research value but also has broad application prospects. The development speed of aerial or ground vision tracking systems is fast, and the relevant technology is also becoming more and more perfect. They play an important role in the detection and tracking of suspicious vehicles and pedestrians in the field of security. However, the visual detection and tracking of underwater suspicious targets face the great challenge of weak observation and low light images, and related research is still immature. This paper proposes a cross-spectral proximity detection and tracking method based on the latest Segment Anything model, and verifies the effectiveness of the method in low light pool experiments.