海面无人机(UAV)的检测属于海杂波背景下的小目标检测问题,多特征联合检测是解决此类问题的有效途径。针对已有的时频三特征检测方法在特征提取阶段计算复杂度过大、难以实现实时检测的问题,提出了一种基于快速时频图的海面无人机多特征检测方法。首先,对雷达复回波数据进行分段快速傅里叶变换,将计算得到的多普勒幅度谱沿多普勒维对齐拼接从而构建快速时频图;其次,对快速时频图进行归一化,达到杂波抑制和增强目标回波的目的,并基于归一化的快速时频图提取三种时频特征;然后,利用快速凸包学习算法训练给定虚警概率下的检测判决区域;最后,通过实测UAV 数据验证并分析了所提方法的有效性。

The detection of sea surface unmanned aerial vehicles (UAVs) belongs to the problem of small target detection in the background of sea clutter due to weak echoes, and joint multi-feature detection is an effective way to solve such problems. Aiming at the existing time-frequency (TF) tri-feature detection method, which has too much computational complexity in the feature extraction stage and is difficult to realize real-time detection, this paper proposes a fast TF-map-based multi-feature detection method for sea surface UAVs. First, the segmented FFT is performed on the radar complex echo data, and the computed Doppler amplitude spectrum is aligned and spliced along the Doppler dimension so as to construct a fast time-frequency map. Second, the fast TF map is normalized to achieve clutter suppression and enhancement of the target echoes, and three kinds of time-frequency features are extracted based on the normalized fast TF map. Third, the fast convex hull learning algorithm is utilized to train the decision judgement region under the given false alarm probability. Finally, the effectiveness of the proposed method is validated and analyzed by the measured UAV data.

TN957. 5;V279

国家自然科学基金资助项目(62371382)