Recent advances in drone-view object detection

作者：Jiaxu Leng, Mengjingcheng Mo, Yinghua Zhou, Yongming Ye, Chenqiang Gao, Xinbo Gao · 发表于：Journal of Image and Graphics · 年份：2023 · DOI：10.11834/jig.220836 · 被引用次数：22 · 研究领域：Video Surveillance and Tracking Methods、Robotics and Sensor-Based Localization、Advanced Neural Network Applications

在人工智能技术的支持下,无人机初步获得智能感知能力,在实际应用中展现出高效灵活的数据收集能力。无人机视角下的目标检测作为关键核心技术,在诸多领域中发挥着不可替代的作用,具有重要的研究意义。为了进一步展现无人机视角下的目标检测研究进展,本文对无人机视角下的目标检测算法进行了全面总结,并对已有算法进行了归类、分析和比较。1)介绍无人机视角下的目标检测概念,并总结无人机视角下目标检测所面临的目标尺度、空间分布、样本数量、类别语义以及优化目标等 5 大不均衡挑战。在介绍现有研究方法的基础上,特别整理并介绍了无人机视角下目标检测算法在交通监控、电力巡检、作物分析和灾害救援等实际场景中的应用。2)重点阐述从数据增强策略、多尺度特征融合、区域聚焦策略、多任务学习以及模型轻量化等方面提升无人机视角下目标检测性能的方法,总结这些方法的优缺点并分析了其与现存挑战之间的关联性。3)全面介绍基于无人机视角的目标检测数据集,并呈现已有算法在两个较常用公共数据集上的性能评估。4)对无人机视角下目标检测技术的未来发展方向进行了展望。;Given the support of artificial intelligence technology, drones have initially acquired intelligent sensing capabilities and have demonstrated efficient and flexible data collection in practical applications.Drone-view object detection, which aims to locate specific objects in aerial images, plays an irreplaceable role in many fields and has important research significance.For example, drones with highly mobile and flexible deployment have remarkable advantages in accident handling, order management, traffic guidance, and flow detection, making them irreplaceable in traffic monitoring.As for disaster emergency rescue, drones with aerial vision and high mobility can achieve efficient search and safe rescue in large areas, locate people quickly and accurately in distress, and help rescuers control the situation, thereby ensuring the safety of people in distress.This study provides a comprehensive summary of the challenges in object detection based on the unmanned aerial vehicle(UAV)perspective to portray further the development...