Drone vs Bird (Drone vs Bird Detection Challenge)

小型无人机的威胁越来越大，因为它们可能被滥用于走私毒品等非法活动以及使用爆炸物或化学武器进行的恐怖主义袭击。目前正在研究几种监视和检测技术，它们在复杂性、范围和能力方面有不同的权衡。 “小型无人机监视、检测和对抗技术国际研讨会”（WOSDETC）旨在汇集学术界和工业界的研究人员，分享该领域的最新进展。结合起来，提出了无人机对鸟类检测挑战。事实上，鉴于无人机的特性，无人机很容易与鸟类混淆，这使得监视任务更具挑战性，尤其是在鸟类数量可能很大的海域。使用视频分析可以解决这个问题，但需要有效的算法能够在不利条件下运行，即弱约束、远距离、降低能见度等。此外，实际系统需要在远距离识别无人机，以便留出反应时间。因此，必须识别非常小的物体并将其与结构化背景和其他具有挑战性的图像内容区分开来。
该挑战旨在通过提供可能难以获得的视频数据集（无人机飞行需要特殊条件和许可，以及在岸需要考虑的问题的区域）。挑战目标是检测在短视频序列中某个时间出现的无人机，其中也有鸟类：算法应该只在无人机出现时发出警报并提供位置估计，而不是对鸟类发出警报。数据集在挑战的连续分期中不断增加，并在之后提供给社区。

The threat posed by small unmanned aerial vehicles (small UAVs) is growing, as they may be misused for illegal activities such as drug smuggling and terrorist attacks involving explosives or chemical weapons. Several surveillance and detection technologies are currently under research, with varying trade-offs in terms of complexity, coverage and performance. The International Workshop on Surveillance, Detection and Countermeasures Technologies for Small Unmanned Aerial Vehicles (WOSDETC) aims to bring together researchers from academia and industry to share the latest advancements in this field. Compounding this issue, the task of detecting UAVs against bird backgrounds poses a critical challenge. Indeed, due to the inherent characteristics of UAVs, they can be easily mistaken for birds, which renders surveillance tasks far more challenging, particularly in maritime areas where bird populations can be extremely large. Video analysis can be employed to address this problem, but effective algorithms are required to operate under adverse conditions, such as weak surveillance constraints, long detection ranges and degraded visibility. Furthermore, practical systems need to identify UAVs at long distances to allow sufficient reaction time. Consequently, it is necessary to detect extremely small objects and distinguish them from structured backgrounds and other challenging image contents. This challenge aims to provide a video dataset that may be otherwise difficult to obtain, covering areas where UAV flights require special conditions and permits, as well as regions with onshore operational considerations. The goal of the challenge is to detect UAVs that appear at certain timestamps within short video sequences that also contain birds: algorithms should only issue alerts and provide position estimates when UAVs are present, rather than triggering false alarms for birds. The dataset will be continuously expanded across successive phases of the challenge, and subsequently made available to the global research community.

搜集汇总

背景与挑战

背景概述

该数据集是一个用于无人机与鸟类检测挑战的视频数据集，旨在解决小型无人机在监控中易与鸟类混淆的问题，通过提供视频序列帮助开发算法在复杂条件下准确检测无人机。数据集由OBSS Technology于2021年发布，并持续更新以支持社区研究。