基于边缘计算和YOLOv3的盲道识别方法  

Blind Road Recognition Method Based on Edge Computing and YOLOv3

作　　者：贾存南 刘新春[1] 李昀骏 陈刚[1] JIA Cunnan;LIU Xinchun;LI Yunjun;CHEN Gang(College of Information and Intelligence,Hunan Agricultural University,Changsha 410128,China)

机构地区：[1]湖南农业大学信息与智能科学技术学院,湖南长沙410128

出　　处：《现代信息科技》2023年第21期102-105,共4页Modern Information Technology

基　　金：国家级大学生创新训练计划项目(202110537027)。

摘　　要：当前我国盲人数量超1700万,而对盲道场景进行实时检测是确保盲人出行安全的重要前提。文章针对目前盲人出行的实际场景,采用基于边缘计算的YOLOv3算法,实现对盲道场景的实时识别。首先,引入了边缘计算模型保证盲道数据处理的实时性;然后,采用YOLOv3算法对盲道障碍物进行检测,通过进行数据增强来提高算法对盲道障碍物的检测性能;最后将YOLOv3部署在边缘计算设备上进行训练,通过GPU的并行计算来提高算法的实时检测效果;实验结果表明,该方法在盲道场景识别的检测效率上相比于传统的盲道识别方法有着大幅度的提升,经实际应用可满足盲人出行的导航需求。At present,there are more than 17 million blind people in China,and real-time detection of blind road scenes is an important prerequisite to ensure the safety of blind people.Aiming at the actual scene of the blind travel,this paper uses the YOLOv3 algorithm based on Edge Computing to realize the real-time recognition of blind road scenes.Firstly,the Edge Computing model is introduced to ensure the real-time performance of blind road data processing.Then,YOLOv3 algorithm is used to detect blind road obstacles,and data enhancement is performed to improve the detection performance of the algorithm for blind road obstacles.Finally,YOLOv3 is deployed on Edge Computing devices for training,and the real-time detection effect of the algorithm is improved by GPU parallel computing.The experimental results show that the detection efficiency of blind road scene recognition of the method is greatly improved compared with the traditional blind road recognition method,and the practical application can meet the navigation needs of the blind travel.

关 键 词：YOLOv3 边缘计算 盲道识别 数据增强 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]