Intelligent In-Vehicle Safety System Based on Yolov5  

作　　者：Biyun Chen Wei Cai Jinwen Zhu Biyun Chen;Wei Cai;Jinwen Zhu(College of Computer Information and Engineering, Yancheng Teachers University, Yancheng, China)

机构地区：[1]College of Computer Information and Engineering, Yancheng Teachers University, Yancheng, China

出　　处：《Journal of Computer and Communications》2024年第3期207-218,共12页电脑和通信（英文）

摘　　要：In order to reduce the occurrence of traffic accidents and assist drivers to avoid dangerous driving. This paper presents a smart in-vehicle safety system that utilises the Yolov5 algorithm. Yolov5 algorithm is used to anticipate driver fatigue and distraction behaviours, and remind drivers to pay attention to safe driving in time. The system continuously splits the frames and analyses the frame content through the video feedback from the front camera, compared to the traditional machine learning, Yolov5’s mosaic data is enhanced, resulting in a batch size enhancement of 92.3%, and it also uses the Drop Block mechanism to prevent overfitting. The hardware of this system uses STM32 microcontroller and uses system DMA interrupt control and buzzer alarm device to warn about dangerous driving behaviour.In order to reduce the occurrence of traffic accidents and assist drivers to avoid dangerous driving. This paper presents a smart in-vehicle safety system that utilises the Yolov5 algorithm. Yolov5 algorithm is used to anticipate driver fatigue and distraction behaviours, and remind drivers to pay attention to safe driving in time. The system continuously splits the frames and analyses the frame content through the video feedback from the front camera, compared to the traditional machine learning, Yolov5’s mosaic data is enhanced, resulting in a batch size enhancement of 92.3%, and it also uses the Drop Block mechanism to prevent overfitting. The hardware of this system uses STM32 microcontroller and uses system DMA interrupt control and buzzer alarm device to warn about dangerous driving behaviour.

关 键 词：Behaviour Detection STM32 Pyside2 Yolov5 Dlib Open Source Library Perclos Model 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]