摘要
为了提升无人机检测的精准度与工作效率,设计了一种以YOLOV4算法优化为基础的低空环境下的无人机检测与跟踪方法。此方法的最大特点是有效结合了卷积神经网络(CNN)的检测方法与追踪算法,实现了对无人机的动态化检测,针对性地优化改进了传统的YOLO网络结构,建立了检测与追踪模型,在自建数据集上进行对比实验。结果表明,创建的模型能辅助降低小目标的漏检概率,在确保检测与追踪实时性的基础上使检测精度增加到77.3%,确保了低空无人机追踪过程的相对稳定性。
In order to improve the accuracy and working efficiency of UAV,the study designs a UAV detection and tracing method in low altitude environment based on YOLOV4 algorithm optimization.The most significant characteristics of the method are the effective combination of the detection method and tracing algorithm of Convolutional Neural Network(CNN),the realization of the dynamic test of UAV,the optimization of traditional YOLO network structure,the establishment of detection and tracing model,and the contrast experiment of self-built database.The results show that the established model can assist to decrease false dismissal probability of the small targets,and increase detection precision to 77.3%based on the guarantee of instantaneity of detection and tracing.This ensures the relative stability of low altitude UAV tracing process.
作者
刘子瑞
Liu Zirui(Jinchu University of Technology,Jinmen 448000,China)
出处
《黑龙江科学》
2023年第6期70-72,共3页
Heilongjiang Science
