This project is the capstone project for my bachelor’s degree @UM-SJTU Joint Institute, initiated by Prof. Mian Li. You can find the organization page on Github.
Design Overview
IIFire
, orIntelligent IoT Firefighting System
, is an integrated system deployed on government servers of Shanghai fire brigade, which detects potential fire hazard and report real-life fire alarms in buildings to save people’s life and money. Currently,IIFire
is connected with over 7 million IoT devices.IIFire
provides a solution for better pre-warning of fire hazard to prevent a fire disaster, which achieves higher accuracy, less false positives, and lower response time than previous fire alarm system.With
Internet-of-Thing
technologies andFuzzy Analytic Hierarchy Process
algorithm deployed in edge computing devices, millions of data generated by IoT devices (especially CV cameras) are processed and send toIIFire
data center immediately. The paper of our team regardingFAHP
can be found here and downloaded here.IIFire
has a follow-up work flow management system. Once a fire alarm or hazard are triggered, withIIFire
software packages including app & website, administrator can keep an eye on the fire status anytime anywhere on any devices, and maintainers can come on spot right away when notification is forwarded to their IIFire apps.
1. Problem Statement
Nowadays, the cost and damage brought by fire disasters are heavy. According to U.S Fire Administration and National Fire Protection Association, the death number, as well as the death rate, has increased by about 3.3% over the past decades. In 2014, 1.9% of the U.S. GDP is consumed in fire protection and losses. As a result, delivering an integrated firefighting system to prevent potential fire hazards with higher reliability, greater performance, and lower cost is needed.
2. System Design
The system architecture is shown in figure below.
- The main concept of our system is based on
IoT
, which stands forInternet-of-Things
. Raw data of firefighting devices obtained from different buildings can be connected and cooperate together. Building our own data center with databaseMangoDB
andFAHP
algorithms, cross-validations can be performed to millions of data acquired. Also, CV algorithmYolo
andShufflenetV2
are applied on real-time fire & smoke detection through camera flow, assisting other IoT devices to decrease the false-positive rate brought by human factors.
- With processed data, user interfaces are the next critical step for staffs to take action when fire hazards are detected. With
React Native
andReact.js
, we design routine data display and safety score calculation for supervision purpose as well as emergency management workflow for both firefighting managers and maintenance workers to be noticed once anomalies occurs. The app and website are real-time synchronized through our server written inPython
.