Research on Real-Time Tracking Algorithm for Multi-Objects of Shipboard Aircraft Based on Detection

Abstract

For safe guidance and real-time monitoring of the shipboard aircraft during the operation, for traditional detection-based object tracking algorithms has poor tracking performance and susceptibility to interference, proposed the multi-object real-time tracking algorithm of shipboard aircraft that combined YOLO v3 object detection algorithm and Kalman filtering. The anchor size of the original YOLO v3 algorithm is optimized by the K-means clustering algorithm, combined with the Kalman filter algorithm to achieve effective tracking of shipboard aircraft objects, and compared with object tracking algorithm based on optical flow and SORT multi-object tracking algorithm on self-built shipboard aircraft tracking dataset and MOT16 multi-object tracking dataset. The results show that the tracking algorithm proposed is more accurate, robust and stable, and has strong adaptability when the object frame width and height change suddenly.

Keywords

• |object detection|kalman filtering|hungarian algorithm|object tracking|deep learning