基于深度学习的无人机目标识别与反制  

Deep Learning Based UAV Target Recognition and Countermeasures

作　　者：侯琛 董俞伯 HOU Chen;DONG Yubo(Shaanxi Vocational College of Police Officers,Xi'an Shaanxi 710021;Xi'an Public Security Bureau,Xi'an Shaanxi 710075)

机构地区：[1]陕西警官职业学院,陕西西安710021 [2]西安市公安局信息通信处,陕西西安710075

出　　处：《软件》2024年第5期161-164,180,共5页Software

摘　　要：随着低空无人机在军事和民用领域的广泛应用,其安全隐患亟需关注。本文提出一种基于改进YOLOv7模型的检测方法,并引入注意力机制,强化模型对目标区域特征的表达能力。同时,提出一种改进的StrongSORT跟踪算法,优化跟踪性能。这些研究成果提高了检测和跟踪的准确性和实时性,通过云台主动跟踪控制算法扩大了监控视野,增强了系统的跟踪灵活性。最终实现了一套完整的红外无人机检测与跟踪系统,满足了实时跟踪的需求,并探讨了其在民用领域反无人机系统中的潜在应用。With the wide application of low-altitude UAVs in military and civil fields,their safety hazards need urgent attention.In this paper,a detection method based on the improved YOLOv7 model is proposed,and an attention mechanism is introduced to strengthen the model's ability to express the characteristics of the target area.An improved Strong SORT tracking algorithm is also proposed to optimize the tracking performance.These research results improve the accuracy and real-time performance of detection and tracking,expand the surveillance field of view through the gimbal active tracking control algorithm,and enhance the tracking flexibility of the system.A complete infrared UAV detection and tracking system is finally realized,which meets the real-time tracking requirements and explores its potential applications in anti-UAV systems in the civil sector.

关 键 词：深度学习 无人机目标识别 改进YOLOv7 注意力机制 StrongSORT跟踪算法 云台主动跟踪控制算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]