基于Yolov5的交通信号灯智能识别程序开发
Development of intelligent recognition program for traffic signal lamps based on Yolov5
郑国荣 1张尊栋 1赵文芊 1柏卓茁 1贾菲儿1
作者信息
- 1. 北方工业大学电气与控制工程学院,北京 100144
- 折叠
摘要
交通信号检测是智能汽车识别交通环境的一项重要辅助技术,现有的算法能够解决单一交叉口环境下的信号检测问题,但需要在十字路口的复杂交通环境中提高算法的精度和干扰可靠性.文章以one-stage目标检测算法Yolov5的应用为研究基础,实现多场景下的交通信号灯自动检测与识别,使用Labeling进行图片标注,通过镜像、裁剪、反转、等运行增强数据集,不断地调参实验与迭代模型训练,目标检测精度达到80%.
Abstract
Traffic signal detection is an important auxiliary technology for intelligent vehicles to identify traffic environment.Existing algorithms can solve the problem of signal detection in a single intersection environment,but it is necessary to improve the accuracy and interference reliability of the algorithm in the complex traffic environment of intersections.Based on the application of one-stage target detection algorithm Yolov5,this paper realizes the automatic detection and recognition of traffic lights in multiple scenes.Labeling is used to annotate pictures,the data set is enhanced by mirroring,cropping,reversing,etc.,and the parameter adjustment experiment and iterative model training are continuously carried out.The target detection accuracy reaches 80%.
关键词
目标检测/Yolov5/Labeling图片标注/模型训练Key words
target detection/Yolov5/Labeling picture annotation/model training引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据