首页|期刊导航|四川轻化工大学学报(自然科学版)|基于YOLOv8n-ByteTrack的无人机检测与跟踪方法

基于YOLOv8n-ByteTrack的无人机检测与跟踪方法OA

Drone Detection and Tracking Method Based on YOLOv8n-ByteTrack

中文摘要英文摘要

针对现有检测与跟踪算法在复杂场景中存在漏检率高、身份切换频繁及运动轨迹碎片化等问题,提出一种基于YOLOv8n-ByteTrack的无人机检测跟踪方法.在检测阶段,引入小目标检测层以增强浅层特征利用,设计多尺度注意力机制对多尺度特征进行加权优化,并采用轻量化 C2f_Star 模块提升特征融合效率,对 YOLOv8n 进行优化;在跟踪阶段,结合全局运动补偿矫正相机位移干扰,引入分阶段重识别模块降低遮挡误匹配风险,并提出基于运动特性的边界框自适应调整策略以应对复杂运动,对ByteTrack进行改进.实验结果表明:改进后的YOLOv8n平均精度达90.9%,较基线算法提升5.8个百分点;改进后的ByteTrack在IDF1指标上达54.0%,较原跟踪算法提升了28.5个百分点.通过检测与跟踪模块的联合优化,系统在跟踪场景下的IDF1进一步提升至59.3%,显著提升了对低空无人机的可靠检测与持续追踪能力.

In view of the problems of high missed detection rates,frequent identity switching and fragmented motion trajectory in existing detection and tracking algorithms under complex scenes,a drone detection and tracking method based on YOLOv8n-ByteTrack is proposed.In the detection stage,a small target detection layer is introduced to enhance the utilization of shallow features,a multi-scale attention mechanism is designed to optimize the weights of multi-scale features,a lightweight C2f_Star module is used to improve the feature fusion efficiency,which improves YOLOv8n;in the tracking stage,the camera displacement interference is corrected by combining global motion compensation,and a staged re-identification module is introduced to reduce the risk of occlusion mismatching.A bounding box adaptive adjustment strategy based on motion characteristics is proposed to cope with complex motion,and ByteTrack is improved as well.Experimental results show that the average precision of the improved YOLOv8n reaches 90.9%,which is 5.8 percentage points higher than that of the baseline algorithm;the improved ByteTrack reaches 54.0%in the IDF1 index,which is 28.5 percentage points higher than that of the original tracking algorithm.Through the joint optimization of the detection and tracking modules,the system's IDF1 in the tracking scenario is further improved to 59.3%,significantly improving the ability to reliably detect and continuously track low-altitude drones.

叶雨灵;姚娅川

四川轻化工大学 自动化与信息工程学院,四川 宜宾 644000||智能感知与控制四川省重点实验室,四川 宜宾 644000四川轻化工大学 物理与电子工程学院,四川 宜宾 644000||智能感知与控制四川省重点实验室,四川 宜宾 644000

信息技术与安全科学

YOLOv8ByteTrack无人机检测无人机跟踪多尺度注意力机制

YOLOv8ByteTrackdrone detectiondrone trackingmulti-scale attention mechanism

《四川轻化工大学学报(自然科学版)》 2026 (4)

103-116,14

四川省科技厅重大专题项目(2018GZDZX0045)

10.11863/j.suse.2026.04.09

评论