AVOIDDS: A dataset for vision-based aircraft detection

Aircraft collision avoidance systems rely on sensor information to detect and track intruding aircraft so that they may issue proper collision avoidance advisories. While typical surveillance sensors for manned aircraft include transponders and onboard radar, autonomous aircraft will require additional sensors both for redundancy and to replace the visual acquisition typically performed by the pilot. As a result, the community has proposed detecting other aircraft using vision-based sensors such as cameras. These sensors require the development of techniques to process images of the environment to detect intruding aircraft. To boost this development, this artifact provides a dataset of 72,000 labeled images of intruder aircraft with various lighting conditions, weather conditions, relative geometries, and geographic locations. For more information on the structure of this dataset as well as benchmark models and a full simulator, see https://github.com/sisl/VisionBasedAircraftDAA.

航空器避碰系统（Aircraft collision avoidance systems）依托传感器信息探测并跟踪入侵航空器，以发出恰当的避碰告警指令。尽管载人航空器的典型监视传感器包括应答机（transponder）与机载雷达（onboard radar），但自主航空器需要增设额外传感器，一方面实现冗余备份，另一方面替代飞行员通常执行的目视目标识别环节。为此，航空学界已提出利用摄像头等基于视觉的传感器（vision-based sensors）探测其他航空器。此类传感器需要开发相关技术以处理环境图像，从而识别入侵航空器。为推动该领域的研发进展，本数据集包含72000张标注图像，涵盖不同光照条件、天气状况、相对几何构型与地理区位下的入侵航空器场景。如需了解该数据集的结构、基准模型以及完整模拟器的更多信息，请访问https://github.com/sisl/VisionBasedAircraftDAA。