基于改进YOLOv8的路口多目标识别优化方法  

作　　者：唐继杰 欧晓放 

机构地区：[1]中国人民公安大学,北京100038

出　　处：《科技创新与应用》2024年第35期59-63,69,共6页Technology Innovation and Application

摘　　要：针对我国城市路口多目标识别效率不高,识别不够精准的问题,该文提出一种基于改进YOLOv8s模型的检测方法。对原始的YOLOv8s模型进行改进,增加了卷积分支,采用融合Diverse Branch Block模块的特征提取方式。结合城市路口场景构建数据集,实验结果表明,改进模型在精确度、召回率和平均准确度的指标上较原模型分别提升4.2%、3.1%和3.08%。改进模型表现出良好的性能,能够满足实际城市路口的多目标检测需求。In view of the problem that the efficiency of multi-target recognition at urban intersections in China is not high and the recognition is not accurate enough.This paper proposes a detection method based on the improved YOLOv8s model.The original YOLOv8s model is improved,convolution branches are added,and feature extraction methods that incorporate the Diverse Branch Block module are adopted.Combining urban intersection scenarios to build a dataset,experimental results show that the improved model improves accuracy,recall rate and average accuracy by 4.2%,3.1%and 3.08%respectively compared with the original model.The improved model shows good performance and can meet the multi-target detection needs of actual urban intersections.

关 键 词：多目标检测 深度学习 YOLOv8 交通安全 优化方法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]