基于YOLOv5s的盲人出行辅助系统研究  

Research on Travel Assistance System for the Blind Based on YOLOv5s

作　　者：李岳 夏益民[2] 蔡述庭[2] 邢延[2] 

机构地区：[1]广东工业大学自动化学院,广东广州510006 [2]广东工业大学集成电路学院,广东广州510006

出　　处：《工业控制计算机》2023年第12期15-17,共3页Industrial Control Computer

基　　金：车规级关键系统基础芯片研发与应用(2022B0701180001);教育部产学合作协同育人项目(202102172024);粤澳联合创新项目(2021A0505080006);广东工业大学大学生创新训练项目(xj2022118450214);珠海市产学研合作项目(ZH22017001200095PWC)。

摘　　要：针对传统盲人出行辅助工具灵活性低、交互性差的问题,提出了一个基于目标检测算法YOLOv5s的盲人出行辅助系统。系统采用双目摄像头实时捕获周围环境信息,通过目标检测算法检测出周围的物体,并进行距离测量,然后将检测数据和距离信息相结合并进行分析,最后以语音的形式反馈给盲人。实际测试结果表明,系统在白天或光线较好的场景下,检测准确率在95.5%以上,在夜晚或光线较弱的场景下,通过低照度增强算法,检测准确率在93.7%以上,并且系统使用多线程并行结构,使得处理视频流的平均帧率为31.34 fps。A blind travel assistance system based on the object detection algorithm YOLOv5s is proposed to address the issues of low flexibility and poor interactivity of traditional blind travel assistance tools in this paper.The system uses binocular cameras to capture real-time environmental information,detects surrounding objects through object detection algorithms,and performs distance measurements.Then,the detection data and distance information are combined for analysis,and finally feedback to the blind person in the form of voice.The actual test results show that the detection accuracy of the system is above 95.5%in daytime or well lit scenes,and above 93.7%in nighttime or weakly lit scenes through low illumination enhancement algorithms.Moreover,the system uses a multi-threaded parallel structure,resulting in an average frame rate of 31.34 fps for processing video streams.

关 键 词：盲人出行辅助 YOLOv5s 距离测量 低照度增强 多线程并行 

分 类 号：R319[医药卫生—基础医学] TP391.41[自动化与计算机技术—计算机应用技术]