U-net分割网络雷达弱小目标检测方法

Radar dim-small target detection based on U-net segmentation network

常规雷达检测方法难以满足当前弱小目标检测需求,而检测前跟踪(TBD)方法在雷达弱小目标检测中存在计算复杂度高、复杂场景性能严重退化的问题.为此,本文提出了一种U-net分割网络的雷达弱小目标检测方法.首先,利用过低门限的多圈检测点形成待分割的图片;然后,利用U-net网络对实体目标进行分割,实现对弱小目标的有效探测.雷达实测数据测试结果表明,本文提出的方法能够大幅提升雷达对远距离小目标的跟踪距离,有效抑制虚假短航迹,为解决雷达弱小目标检测提供了新的解决思路.

Conventional radar detection methods are unable to meet the current requirements for detecting dim-small targets,and the tracking before detection(TBD)method has the problem of high computational complexity and severe degradation of performance in complex scenes in detecting the radar dim-small targets.To this end,the paper presents a radar dim-small target detection method using U-net segmentation network.Firstly,multi-circle detection points with excessively low threshold are used to form an image to be segmented.Then,the target entities are segmented by using U-net network to realize the effective detection of dim-small targets.The test results of radar measured data show that the method proposed in the paper can greatly increase the tracking range of radar on remote small targets,effectively suppress false short tracks,and provide a new solution for radar detection of dim-small targets.