摘 要：随着无人机技术的不断发展，目标跟踪已成为无人机应用的关键技术之一。针对无人机目标跟踪中，目标易发生遮挡、形变、尺度变化以及多视角变化等问题，提出一种基于频域特征和Transformer的无人机目标跟踪算法。首先，采用蒸馏后的Transformer深度网络提取图像空间全局特征，随后利用自适应频域感知网络提取频域细节特征；同时在输入端增添学习图像作为补充，以捕获目标模块与搜索区域之间的相关性，用于更新初始目标模板，增强对目标的表征能力。其次，提出一种基于互信息最大化的多视角不变特征学习策略。通过最大化目标模板和搜索模板之间的互信息设计新的损失函数，提升跟踪网络处理目标变化的能力。最后，根据学习图像特征响应确定目标位置。仿真实验结果表明，该算法能够有效提升无人机目标跟踪的精度，具有较好的鲁棒性。

Abstract: With the rapid development of unmanned aerial vehicle (UAV) technology, object tracking has become one of the key techniques in UAV applications. To address challenges such as occlusion, deformation, scale varia-tion, and multi-view changes in UAV object tracking, this paper proposes a UAV tracking algorithm based on fre-quency-domain feature and Transformer architecture. First, a distilled Transformer network is employed to extract global spatial features from images, while an adaptive frequency-domain perception network captures fine-grained frequency details. In addition, a learning image is introduced at the input stage to capture the correlation between the target template and the search region, thereby updating the initial target template and enhancing target repre-sentation. Second, a multi-view invariant feature learning strategy based on mutual information maximization is pro-posed. By maximizing the mutual information between the target template and the search template, a novel loss function is designed to improve the network’s robustness against target appearance variations. Finally, the target position is determined according to the feature responses of the learning image. Simulation results demonstrate that the proposed algorithm effectively improves UAV object tracking accuracy and exhibits strong robustness under complex scenarios.