基于Jetson TX2的老人跌倒监护设计  

Design of Jetson TX2 based fall monitoring for the elderly

作　　者：张晓华[1] ZHANG Xiaohua(School of Electronic Engineering,North China University of Water Resources and Electric Power,Zhengzhou 450046,China)

机构地区：[1]华北水利水电大学电子工程学院,河南郑州450046

出　　处：《现代电子技术》2022年第20期45-48,共4页Modern Electronics Technique

基　　金：华北水利水电大学一流本科课程建设基金项目:“嵌入式系统”(2020-90号);教育部高等教育司2021年第一批产学合作协同育人项目:引入RT-thread的一流本科课程《“嵌入式系统”的建设和提升》(202101303022);2021年河南省高等教育教学改革研究与实践项目:“专创融合”助力电子信息类一流专业建设的探索与实践(2021SJGLX410)。

摘　　要：为减少“空巢老人”因发生跌倒而遭受的后继伤害,助力人工智能在居家环境中的应用落地,文中提出一种基于通用边缘智能平台的老人跌倒监护设计。该设计以边缘智能平台为核心,实时采集高清摄像头的视频数据,并对视频帧进行运动目标提取和人体跌倒姿态识别,在跌倒发生时能实时给监护人发送告警信息,以便获得及时救助。同时,在Jetson TX2平台上构建机器视觉开发环境,利用Darknet深度学习框架和预训练权重模型实现人体识别和跌倒监测,基于OpenCV和腾讯云短信技术分别实现本地端图形方式和远程端短信方式的报警功能。测试结果表明,文中的设计方案能够实时、精确地监测跌倒异常,无需佩戴,适合老人的使用习惯,且具有实现简单、易于部署的优点,可为跌倒监护提供一种低成本的方案。In order to reduce the subsequent injury caused by falls of "empty nest elderly" and help the application of artificial intelligence in the home environment,a fall monitoring design based on a universal edge intelligence platform for the elderly is proposed. In the design,the edge intelligence platform is taken as the core,the video data of the high-definition camera is collected in real time,and extraction of the moving target and recognition of the human fall posture on video frames are carried out to send warning information to the guardian in real time when a fall occurs,so as to obtain timely assistance. The machine vision development environment is built on the Jetson TX2 platform. The Darknet deep learning framework and pretrained weight model are used to achieve human body recognition and fall monitoring. The warning functions of local graphical mode and remote SMS mode are realized respectively on the basis of OpenCV and Tencent cloud SMS technology. The testing results show that the design scheme can accurately monitor fall abnormalities in real time,which does not need the elderly to wear anything,is suitable for the use habits of the elderly,and has the advantages of simple implementation and easy deployment. It can provide a low-cost solution for fall monitoring.

关 键 词：跌倒监护 Jetson TX2 视频数据采集 机器视觉 运动目标提取 人体识别 告警信息发送 系统测试 

分 类 号：TN931.3-34[电子电信—信号与信息处理] TP212.9[电子电信—信息与通信工程]