基于雷达、可见光、红外、声学、无线电侦测等多传感器融合的无人机目标探测与识别技术成为无人机探测预警系统的发展趋势,深度学习网络具有强大的表征学习能力,善于从多源异构数据中提取各种复杂特征。文中在对比各类传感器探测能力与性能优劣的基础上,回顾了以深度学习为主的特征提取算法针对各类传感器数据的应用成果及存在的问题,讨论了多源信息融合技术与多传感器系统搭建方案。最后,推荐了一种基于四类传感器的无人机探测系统搭建方案,并得出结论。

The target detection and recognition technology based on multi-sensor fusion of radar, visible light, infrared, acoustics, radio detection has become the development trend of unmanned aerial vehicle (UAV) detection system. Deep learning network has a strong representation learning ability, and is good at extracting various complex features from multi-source heterogeneous data. On the basis of comparing the detection ability and performance of various sensors, this paper reviews the application results and existing problems of the feature extraction algorithms headed by deep learning for multi-sensor data, and discusses the multi-source information fusion technology and the construction scheme of multi-sensor system. Finally, a scheme of UAV detection system based on four kinds of sensors is recommended and a conclusion is drawn.

TN95

国家自然科学基金委员会-中国民航局民航联合研究基金资助项目(U1933135);国家重点研发计划资助项目(2016YFC0800406)