UMOTMA: Underwater multiple object tracking with memory aggregation

Underwater multi-object tracking (UMOT) is an important technology in marine animal ethology. It affected by complex factors such as scattering, background interference, and occlusion, which makes it a challenging computer vision task. As result, the stable continuation of trajectories among different targets has been key to performance UMOT tasks. To solve challenges, we propose underwater algorithm based on memory aggregation (UMOTMA) effectively associate multiple frames with targets. First, long short-term (LSTM)-based module (LSMAM) enhance utilization between frames. Next, LSMAM embeds LSTM into transformer structure save aggregate features Then, image enhancement M E introduced process original images, improves quality visibility images so that model can extract better from images. Finally, are integrated backbone network implement entire framework, fully utilize historical information tracked Experiments datasets fish school show UMOTMA generally outperforms existing models maintain stability target trajectory while ensuring high-quality detection. The code available via Github.