当前位置:网站首页>Introduction of safety helmet wearing recognition system
Introduction of safety helmet wearing recognition system
2022-08-11 06:17:00 【Beijing Fuwei Image】
The safety helmet wearing recognition system developed by Fuwei Image, based on artificial intelligence image recognition technology, can quickly analyze whether the person in the image is wearing a safety helmet, and alarm in real time.The system meets the requirements of low-cost deployment, large-scale access, multi-algorithm concurrency, small sample training, and multi-channel alarming.
I. Snapshot Rules
When the person in the image is not wearing a helmet, the Hard hat wearing recognition system automatically captures the image or video from the camera video stream and alarms.The long-distance image requires that the height of the human body is greater than 1/10 of the overall image, that is, the human eye can distinguish; the short-distance image requires at least the upper body to be exposed.The hats worn by workers should be site/factory safety helmets in red, yellow, blue, white, orange.
Second, how it works
The working principle of the helmet recognition system is to analyze whether the helmet is worn in real time, record live video, identify, track and alarm. The helmet recognition system does not depend on other sensors, and directly analyzes and warns through the video in real time and uploads it to the managementThe system server, and then the server analyzes the video stream in real time, and accurately determines whether there is a violation of regulations and not wearing a helmet through precise calculation and identification.While reminding supervisors, the system will automatically save the time, location and corresponding photos as the basis for punishment. Safety supervisors can remotely or on-site correct and supervise wearing safety helmets when they see the uploaded data and graphics.
Three, product advantages
Hard hat wearing recognition system does not require GPU, but only needs CPU to run, using existing equipment, greatly reducing deployment costs; suitable for large-scale large-scale scenariosOne host can simultaneously access dozens of video channels for analysis to meet the needs of large-scale analysis; the system can support multi-algorithm concurrency, and function modules can be built flexibly according to the needs of the scene.match.The system uses the latest artificial intelligence deep learning and big data technology, and the training model independently developed by Fuwei Image conducts small sample training, continuously improves the development efficiency of the algorithm, and makes the customized development algorithm for new scenarios mature.
4. Camera requirements and parameters
Hard hat wearing recognition system Supports Hikvision, Dahua IP cameras or DVRs, or other RTSP/RTMP/ONVIF video streams.The camera erection height h is about 2.0-3.5 meters, and when the long-distance telephoto shooting and the shooting distance is greater than 30 meters, it can be erected at 3.5-5 meters.The angle between the camera's line of sight and the horizontal plane is <60° (ie, don't look down too much).The shooting distance d is related to the focal length of the selected lens. The recommended focal length is 8mm, and the recognizable distance is 100 meters.If the focal length is 6mm, the recognition distance is 50 meters.
V. Operating environment requirements
System requirements: 64-bit Windows operating system, Win10 recommended;
CPU requirements: Core i3/i5/i7/i9 4th generation or above;
Memory requirements: 4G or above; recommended according to the specific number of channels and functions;
Network requirements: The official version of the software does not need to be connected to the Internet, but it must be connected to the camera.
The helmet wearing recognition system is independently developed by Fuwei Image, and is suitable for helmet recognition tasks in road/construction sites, substations, workshops, factories, mining areas, etc.The system has been released for more than a year, and the accuracy rate is more than 95%. The effect is optimal without picking scenes.
边栏推荐
猜你喜欢
随机推荐
CVPR2022——A VERSATILE MULTI-VIEW FRAMEWORK
智慧工地 安全帽识别系统
Thread Handler
order by注入,limit注入,宽字节注入
Redis哨兵模式
目标检测——Faster-RCNN 之 RCNN
更新GreenDAO实体类导致的编译错误
Realize data exchange between kernel and userspace through character device virtual file system (passed based on kernel 5.8 test)
安全帽识别系统-为安全生产保驾护航
.Net6 MiNiApi +EFCore6.0高B格操作的WebApi
2021-05-10
>>数据管理:读书笔记|第一章 数据管理
第七届集美大学程序设计竞赛(个人赛)题解
DNS外带注入SQLMAP
经纬度距离
梅科尔工作室-HarmonyOS应用开发第四次培训
RecycleView
Redis分布式锁
【mysql】查询不区分大小写(用户密码登录不区分大小写)
Maykle Studio - Second Training in HarmonyOS App Development