无人机视角下基于深度学习的多目标跟踪研究进展

Research Progress of Multi-Target Tracking Based on Deep Learning from Perspective of UAV

基于无人机平台的多目标跟踪在智慧城市、农业生产、灾害预警与搜救等多个领域有广泛的应用前景。与普通视角下的多目标跟踪相对成熟不同,无人机视角下的多目标跟踪面临一系列尚未完全解决的问题,主要包括目标尺度变化、相似目标干扰、目标被遮挡和重叠以及目标分布不均等。梳理了近年来在普通视角下的经典多目标跟踪算法,并以基于检测的跟踪框架为主,综合分析了无人机视角下多目标跟踪领域的主要技术路线和最新方法。统计了性能评估方法和主流数据集,分析了当前无人机视角下多目标跟踪所面临的主要挑战,并对未来的研究趋势进行展望,旨在为后续相关研究提供参考。

Multi-object tracking based on UAV platforms has wide application prospects in various fields,including smart cities,agricultural production,disaster early warning and search,and rescue operations.Unlike the relatively mature multi-object tracking from the traditional perspective,multi-object tracking from the UAV perspective faces a series of challenges that have not been completely solved.These challenges mainly include target scale changes,interference from similar targets,target occlusion and overlap,and uneven target distribution.This paper compiles the classic multiobject tracking algorithms developed in recent years from a traditional perspective.It also comprehensively analyzes the main technical approaches and latest methods in the field of multi-object tracking from the perspective of UAVs,with a particular focus on the detection-based tracking framework.Additionally,it examines the performance evaluation methods and mainstream datasets used in this domain.Moreover,the paper analyzes the primary challenges faced in multi-object tracking from UAV perspectives and offers insights into future research trends,aiming to provide valuable references for further related studies.