Abstract:Visual multi-object tracking on the maritime surface based on unmanned surface vessels (USVs) is a crucial technology for realizing intelligent maritime surveillance. However, videos obtained from the maritime surface often encounter challenges such as scene blurring, oscillations of USV observation platforms, and non-linear movements of objects, significantly affecting tracking performance. In response to these challenges, the article builds upon ByteTrack to introduce the Maritime Object Motion Tracker (MOMT), employing YOLOv7 as the object detector. Diverging from ByteTrack, MOMT proposes a new metric function and incorporates Gaussian cascade matching for a more refined handling of low-score and unmatched bounding boxes. Additionally, an augmented observation Kalman filter module is developed to diminish the effects of nonlinear motion in objects. Specifically, First, in establishing the relationship between adjacent frames, the Fuse-distance metric function is proposed by synthesizing information on IOU distance, object category, and size, to determine the optimal matching between detection boxes and Kalman filter prediction boxes. Secondly, the introduction of the Gaussian cascade matching module better addresses the issue of frequent ID switches caused by significant shaking of the USV observation platform. Additionally, to cope with the impact of non-linear object motion, the combination of camera motion compensation and the Observation-centric Kalman filter is employed to mitigate estimation errors in the Kalman filter. Finally, MOMT is compared with other state-of-the-art tracking algorithms on the Singapore Maritime Dataset, achieving competitive performance with 46.0 MOTA and 66.7 IDF1, and operating in real-time at a speed of 44.6 FPS. Experimental results demonstrate that MOMT can achieve real-time and effective multi-object tracking in complex maritime scenarios, providing technical support for intelligent maritime surface monitoring.

Text-Guided Multi-Modal Fusion for Underwater Visual Tracking

A Robust Underwater Multiclass Fish-School Tracking Algorithm

Closed-Loop Tracking-by-Detection for ROV-Based Multiple Fish Tracking

Embedded Online Fish Detection and Tracking System Via YOLOv3 and Parallel Correlation Filter

Improving Underwater Visual Tracking With a Large Scale Dataset and Image Enhancement

Underwater Long-Term Object Tracker for Marine Organism Capture

Semi-supervised Visual Tracking of Marine Animals Using Autonomous Underwater Vehicles

Real-time Visual Object Tracking with Natural Language Description

Object Detection And Tracking Of Unmanned Surface Vehicles Based On Spatial-Temporal Information Fusion

A Unified Framework for Joint Moving Object Detection and Tracking in the Sky and Underwater

More Perspectives Mean Better: Underwater Target Recognition and Localization with Multimodal Data via Symbiotic Transformer and Multiview Regression

Long-Term Visual Tracking of an Unmanned Surface Vehicle in Complex Environment Based on Adaptive Multi-Feature Fusion with Correlation Filtering

Simple Online and Realtime Underwater Multiple Object Tracking

MOMT: A Maritime Real-Time Visual Multi-Object Tracking Algorithm Based on Unmanned Surface Vehicles

Novel Pipeline Integrating Cross-Modality and Motion Model for Nearshore Multi-Object Tracking in Optical Video Surveillance

Text-Guided Multi-Class Multi-Object Tracking for Fine-Grained Maritime Rescue

FishTrack23: An Ensemble Underwater Dataset for Multi-Object Tracking

Underwater Target Tracking Using Forward-Looking Sonar for Autonomous Underwater Vehicles.

Visual-Linguistic Representation Learning with Deep Cross-Modality Fusion for Referring Multi-Object Tracking

Joint Visual Grounding and Tracking with Natural Language Specification

Multi-AUV Assisted Seamless Underwater Target Tracking Relying on Deep Learning and Reinforcement Learning