A dynamic target tracking model for uavs based on the fusion of twin networks and deep learning

Abstract

Abstract In this paper, a deep learning model incorporating twin networks is proposed to address the challenge of dynamic target tracking by UAVs in complex environments. By exploring the combination of twin networks and convolutional neural networks (CNNs) in detail, this study aims to develop an efficient algorithm capable of accurately tracking dynamic targets under various environmental conditions. The design of the model focuses on utilizing the strong feature-matching capability of twin networks and the advantages of CNNs in image processing in order to improve tracking accuracy and response speed. Experimental simulations show that the model in this paper exhibits better accuracy and response time than the existing popular algorithm YOLOv3 in dynamic target tracking tasks at different difficulty levels. In particular, the present model demonstrates higher robustness and adaptability under different conditions. In addition, the practical application potential of the model is discussed in this paper, pointing out its wide application prospects in the fields of unmanned aerial surveillance and environment sensing of self-driving vehicles.