摘要
当前无人机多目标跟踪检测中还存在较多问题,比如精确度低以及占用内存大的问题。基于YOLOv5算法和DeepSort算法开发无人机软件,设计跟踪模板更新策略,通过检测模块将目标置信度、置信度等相关信息传输到跟踪模块实现跟踪。在YOLOv5算法和DeepSort算法的应用下,实现对目标估计的预测和分配预测结果,确保无人机跟踪效果和速度。实验结果发现,基于YOLOv5算法和DeepSort算法的无人机软件在多目标跟踪中,可以将跟踪精度提升22%,效果显著。
There are still many problems in the current multi target tracking and detection of unmanned aerial vehicles,such as low accuracy and large memory usage.Based on this,the YOLOv5 algorithm and DeepSort algorithm are used to complete the development of unmanned aerial vehicle software,and a tracking template update strategy is designed.Through the detection module,relevant information such as target confidence and confidence are transmitted to the tracking module to implement tracking.Under the application of YOLOv5 algorithm and DeepSort algorithm,the prediction of target estimation can be achieved,and the prediction results can be allocated to ensure the tracking effect and speed of unmanned aerial vehicles.The final experimental results found that the drone software based on YOLOv5 algorithm and DeepSort algorithm can improve tracking accuracy by 22%in multi target tracking,with significant results.
作者
李瑞
郭斌
焦义贵
LI Rui;GUO Bin;JIAO Yigui(Nanjing Multi Base Observation Technology Research Institute Co.,Ltd.,Nanjing 211500,China)
出处
《电子测试》
2024年第1期101-104,共4页
Electronic Test
关键词
跟踪检测
多目标
无人机软件开发
算法
tracking and detection
multiple objectives
drone software development
algorithm
