A Moving Object Detection and Tracking Framework Based on YOLOv8 Model and Improved OC-SORT Algorithm in Complex Marine Environment

Moving target detection and tracking is an important application field of Marine detection, which can provide important support for maritime safety and monitoring maritime search and rescue, maritime traffic management and optimization and civil surveillance. Compared to conventional methods for detecting and tracking moving objects, deep learning-based approaches can achieve higher accuracy in complex Marine environments. The complex environment at sea usually includes changing light conditions, waves, reflections, etc., which have a great impact on traditional methods, and deep learning models can learn these complex features through a large number of training data, so as to improve the recognition and tracking ability of targets. This paper proposes a moving object detection and tracking framework based on YOLOv8 model and improved OC-SORT algorithm. First, the performance of YOLOv8 models of various sizes in detecting objects is evaluated to select the most appropriate model for multi-target motion detection in complex marine environments. Then, an enhanced OC-SORT algorithm is employed to track the detected targets, thereby completing the entire detection and tracking procedure. The experimental outcomes demonstrate that this framework not only boasts high accuracy in both detection and tracking but also possesses the capability for real-time tracking.