基于YOLOv2视觉的路标检测系统与应用  

Road Sign Detection Method and Application Based on YOLOv2 Vision

作　　者：黄文冠 林华沐 周莹 李沛濠 周小燕 刘昱林 

机构地区：[1]广东理工学院电气与电子工程学院,广东肇庆526000

出　　处：《工业控制计算机》2023年第8期135-136,139,共3页Industrial Control Computer

基　　金：广东理工学院2022年大学生创新创业训练计划项目(省级):疫熠先锋(S202213720016)。

摘　　要：针对于封闭园区内物资和外卖递送,及疫情时期社区内消毒喷洒等情况下出现的人手不足和效率低的问题,提出了一种基于YOLOv2视觉的路标检测系统来帮助无人小车和机器人实现在封闭园区内的自主巡航功能。该系统基于YOLOv2算法实现,拟通过机器视觉识别路标来解决自主巡航的问题。YOLOv2算法通过FPS和MAP来表示路标检测的效率和准确性,其整个YOLOv2网络结构由19个卷积层、5个池化层和批量标准化层构成特征提取网络。通过该网络结构将路标物体中心点与分割出的网络单元格进行匹配,利用所得的IOU分数作为回归Label值,再通过载入模型网络平台不断训练得到的路标识别模型,完成机器视觉帮助无人物流车和机器人自主巡航运送物资的应用。In this paper,a signpost detection system based on YOLOv2 vision is proposed to help unmanned vehicles and robots realize the autonomous cruise function in the closed park.Based on YOLOv2 algorithm,the system aims to solve the problem of autonomous cruise by recognizing road signs through machine vision.The YOLOv2 algorithm uses FPS and MAP to represent the efficiency and accuracy of road sign detection.The obtained IOU score is used as the regression Label value,and then the road sign recognition model is loaded into the model network platform and constantly trained to complete the application of machine vision to help the unmanned logistics vehicle and robot autonomous cruise transport materials.

关 键 词：YOLOv2 视觉检测 路标识别 自主巡航 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] U463.6[自动化与计算机技术—计算机科学与技术]