海面目标跟踪过程受到强海杂波干扰、回波信号时变非平稳、点迹丢失和目标机动等影响,导致传统方法的航迹起始效果较差,出现混批、起始失败和误跟踪等严重后果。针对上述问题,提出了一种基于机器学习支持向量机的海面目标航迹起始算法,采用支持向量机作为分类器,通过样本训练实现对真实目标与虚假目标的分类,将航迹起始问题转化为真假目标的区分和鉴别问题。文中方法利用机器学习的数据驱动策略选择分界面方案代替传统技术中利用先验知识来人工选择门限分界,可以显著减少对先验信息的依赖;同时,门限具有自适应调整的能力,可大大提高算法的自适应性和鲁棒性。最后,利用雷达实测数据对算法的有效性进行了验证。

The tracking process of sea surface targets is often influenced by strong sea clutters, plots missing and target maneuver, which result in poor track initiation or even confusion, initiation failure and tracking mistake. In this paper, a type of sea surface targets track initiation method is proposed by using the machine learning support vector machine approach, which utilizes support vector machine as the identifier and transforms the track initiation problem to the classification of true and false targets. In the investigated machine learning algorithm a data driven approach rather than the manned method using priori information is chosen to define thresholds, which can reduce the dependence on priori knowledge. The thresholds can be adjusted adaptively and this may enhance adaptivity and robustness. Finally, the effectiveness of the proposed algorithm is verified based on real radar data.

TN957.51