聚焦、检测、跟踪、再聚焦:基于经典跟踪前检测框架的实时小目标检测技术

Focus, Detect, Track and Re-Focus: A Dim Target Detection Technique Using the Classical Detect before Track Architecture

隐身技术的发展对空空导弹的检测跟踪性能提出了较大的挑战。弱小目标检测跟踪算法作为弹载信号处理的核心,需要同时兼顾检测性能与执行效率。经典的跟踪前检测(Detect Before Track, DBT)方法属于一种级联结构,具有较高的实时性。本文在DBT架构下,提出聚焦(Focus)、检测(Detect)、跟踪(Track)、再聚焦(Re-Focus)弱小目标检测跟踪算法FDTR。基于实测数据的实验结果表明,该方法具备优异的检测和跟踪性能,代码执行效率高实时性强。FDTR完全基于Python和标准库实现,接口规范,架构清晰。基于FDTR,可方便的替换并评估算法中的聚焦模块,并可进一步升级为检测前跟踪(TBD)架构,提升信噪比利用率。

The development of stealth technology has brought greater challenges to the detection and tracking performance of air-to-air missiles. As the core of missile-borne signal processing, the detection and tracking algorithm of dim targets needs to have high performance and real-time execution.Based on the classic Detect Before Track(DBT) framework, a focus, detect, track and re-focus algorithm called FDTR is proposed. FDTR is highly real-time and also has excellent detection and tracking performances. The implementation of FDTR is based on Python and standard libraries, which make both the interface and the architecture clear and easy to understand. The focus module in FDTR can be easily replaced and evaluated using advanced focus technique and the Track Before Detect(TBD) architecture can be integrated to further enhance the signal-to-noise accumulation ratio of dim targets.