AUTHOR=Hao Zhicheng , Qiu Jun , Zhang Haimiao , Ren Guangbo , Liu Chang 

TITLE=UMOTMA: Underwater multiple object tracking with memory aggregation

JOURNAL=Frontiers in Marine Science

VOLUME=Volume 9 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/marine-science/articles/10.3389/fmars.2022.1071618

DOI=10.3389/fmars.2022.1071618

ISSN=2296-7745

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