首页|基于TDA4VM的疲劳状态实时检测系统设计

基于TDA4VM的疲劳状态实时检测系统设计

扫码查看
针对传统嵌入式平台疲劳状态检测系统识别精度低和实时性差的问题,设计了一种基于TDA4VM异构多核处理器的疲劳状态实时检测系统.TDA4VM嵌入式处理器通过摄像头获取图像并进行目标检测,STM32 微控制器控制外设模块,包括GPS模块、GSM模块和语音模块.在目标检测算法方面,先在YOLOX目标检测算法中引入注意力机制模块CBAM(Convolutional Block Attention Module),再对激活函数进行改进,并优化小滑窗替换算法.将训练后的YOLOX模型部署在硬件平台上,实际车载实验结果表明,在不同环境下疲劳状态检测精度可达到 95.3%,同时还实现了30 帧/s的实时检测.该检测系统具备精度高、实时性强和教学简易等特点,在实验教学和工程应用方面具有一定的参考价值.
Design of a Real-time Fatigue Status Detection System Based on TDA4VM
A real-time fatigue detection system based on the heterogeneous multi-core processor is designed to solve the problems of low recognition accuracy and poor real-time performance in fatigue state detection systems of traditional embedded platforms.The TDA4VM embedded processor acquires images from the camera and detect target.The STM32 microcontroller controls peripheral modules,including GPS module,GSM module and voice module.In the aspect of target detection algorithm,firstly,the attentional mechanism module CBAM is introduced into YOLOX object detection algorithm,then the activation function is improved and the shifting small sliding window is optimized.The trained YOLOX model is deployed on the hardware platform.Through actual vehicle experiments,the results show that the accuracy of fatigue detection can reach 95.3%in different environments,and real-time detection at 30 frames per second is achieved.This system has the characteristics of high accuracy,robust real-time performance,and easy teaching,which has certain popularization value in experimental teaching and engineering application.

fatigue testingdeep learningheterogeneous multi-coreTDA4VM processorYOLOX algorithm

付丽、滕召波、张一帆、罗钧、王浩程

展开 >

重庆大学光电技术及系统教育部重点实验室,重庆 400044

疲劳检测 深度学习 异构多核 处理器 YOLOX算法

2024

实验室研究与探索
上海交通大学

实验室研究与探索

CSTPCD北大核心
影响因子:1.69
ISSN:1006-7167
年,卷(期):2024.43(11)