Correlation-Based Methods for Tracking Moving Objects Using UAS Systems

The task of continuous visual tracking of moving objects by unmanned aerial systems (UAS) is one of the most relevant challenges in the fields of computer vision and autonomous robotics. This paper presents the development and testing of a lightweight real-time tracking system based on correlation filter methods, particularly the CSRT tracker. The system is integrated with a Software-in-the-Loop (SITL) simulation using PX4, MAVSDK, and Gazebo Classic, enabling realistic flight modeling without physical hardware. An initial manual selection of the target object is performed by the operator, after which autonomous tracking is executed. The solution emphasizes minimal computational load, rapid response to object movement, and autonomous operation without cloud service dependency. Experimental results demonstrate high stability of object tracking under moderate appearance changes and validate the efficiency of the proposed approach for low-power platforms. Future work will focus on expanding the system to multi-object tracking and enhancing robustness under complex environmental conditions. The work serves as a foundation for further integration into real UAVs, including C++ implementation and PX4 firmware deployment.