基于分布式微多普勒融合特征的高可用性微小型无人机智能探测技术

传统的微小型无人机探测技术存在虚警率高、漏警风险大、地物遮挡效应强、依赖多传感器协同等不足。本项目拟研究基于分布式微多普勒融合特征的高可用性微小型无人机雷达探测技术，瞄准城区和战场两大应用领域解决当前微小型无人机探测的迫切需求。在系统设计层面，研究高可用性分布式系统，根据区域特征合理规划雷达分布，规避高楼、树木等干扰物的遮挡，确保全方位探测，并在某部雷达出现故障后，其探测区域可被其他雷达接管，在适当的冗余覆盖下做到无漏警探测。在信号处理层面，利用辅助噪声多维经验模态分解从分布式回波中分离多维微多普勒信号，基于华为国产化AI框架MindSpore，建立分布式循环神经网络模型提取多维微多普勒融合特征，提高强杂波环境下识别目标的可靠性。研究成果在民用领域可解决无人机的“黑飞”问题；在军事应用中不排斥现有中远距离雷达探测体制，可级联配合使用，作为我国在近程防御系统中精准识别微小型无人机的终极屏障。

The performance of traditional detection technology is restricted by the high false alarm rate, big risk of missing alarm, strong shielding effect of ground objects, and dependence on the cooperation of multiple sensors. The core of the project is to study high availability detection of small unmanned aerial vehicles (UAVs) with distributed micro-Doppler (m-D) fusion features, aiming at two major applications for urban and battlefield to solve the urgent needs of the detection for small UAVs. At the system level, the distributed system with high availability is investigated, and the radar distribution is assigned reasonably according to the regional characteristics, so as to avoid the obstruction of high-rise buildings, trees and other interference objects, monitoring all-round the detection area. When certain radar fails, the detection area could be taken over by other radars in the system, and missing alarms could be eliminated under the appropriately redundant coverage. At the signal processing level, multi-dimensional m-D signals are separated from the distributed echoes by noise assisted multivariate empirical mode decomposition. MindSpore, the domestic AI framework of Huawei, is adopted to establish the distributed recurrent neural network (RNN) model to extract multi-dimensional m-D fusion features, which could improve the reliability of target recognition in strong clutter environment. The results of the project could solve the "black flight" problem of drones in the civil field; In military applications, the existing radar systems with medium and long detection ranges are not excluded, and the research of this project can be used in cascade, which could provide the ultimate barrier to accurately identify small UAVs in defense system of the short range.