Skip to content

A prototype smart evacuation and rescue system that leverages data from WiFi, CCTVs and sound sensors to locate crowds, identify trapped or vulnerable occupants, compute optimal escape routes and provide real-time visualization of vital information for quick and effective decision-making by rescuers. 1st Prize @ the SCDF X IBM - Lifesavers' Inno…

License

Notifications You must be signed in to change notification settings

shingkid/DrWatson-ToTheRescue_SCDFXIBM

 
 

Repository files navigation

SCDF Call for Code 2020 - Dr. Watson

Problem Statement: Integrating with a Smart Environment

Team Members: Jane Seah, Lim Pei Xuan, Ong Yan Kai, Patrick Lim, Vinnie Chu

Short description

What is the problem?

In times of emergencies, such as a fire breakout, there is a need for more on-the-ground intelligence for effective decision-making to make evacuation and rescue safer for all.

People within the Building

Currently, there are a few possible scenarios that could put these people at risk of running to a chokepoint or dangerous zones during a panicky evacuation:

  • Lack of familiarity with the building layout
  • Lack of real-time information of blockages from the fire and alternative exit routes
  • Stranding due to mobility constraints

First Responders: Security, Management, Community First Responders

For this group of people, the priority would be to rush to the key areas that need help directing people to the nearest safest exit or areas where there is a need for assistance, such as people with disabilities. However, current CCTVs are not intelligent in tracking areas and people that need assistance, wasting security and management’s time for an effective on the ground evacuation efforts.

SCDF First Responders

For the SCDF emergency team rushing to the incident site, they also face a series of challenges in planning and executing a quick and effective rescue plan:

  • There is an information lag as they rely on the information provided by the caller/witnesses at the scene (i.e. Occupants at risk may have moved from their last known location)

  • Insufficient real-time overview of the situation and damage to formulate a clear rescue plan

    • Lack of information of structural damages and blockages increase time to search for rescue routes
    • No clear latest situation of where the trapped people are

Having all these vital information for the different parties have great potential in significantly reducing the casualties in times of emergency like a fire breakout.

Our Solution

Our team believes that with the rise of "Smart" Infrastructure, we can leverage existing and upcoming technologies to address these challenges. We have built a prototype evacuation and rescue system that makes use of smart intelligence collected from various data sources in the smart building to provide a clear, real-time situation on the ground through a 3-D building dashboard, highlighting the vital information for quick and effective decision-making by the first-responders and rescuers. The real-time intelligence also provides data to generate the safest and fastest evacuation routes for the people in the building.

To efficiently and safely evacuate occupants when a fire is detected, the system will make use of WiFi Triangulation to locate occupants and serve personalised optimal escape routes directly to their devices. The algorithm will tap on the data from sensors and CCTV cameras to determine which zones are unsafe/blocked and compute the best route to safety accordingly.

Upon the actuation of the fire alarm, the responders will have access to a dashboard with real-time updates of the situation at the site. The CCTV footage from the site will be processed to automatically identify (Computer Vision) zones that still contain a high number of occupants, as well as the zones where people with mobility impairments were detected (e.g. wheelchair bound occupants). Distress signals are also picked up by our sound sensors in the event that CCTVs' visibility is low. The dashboard will also contain a route finder that computes the fastest and safest way to move between any zone in the building, taking into account damage and blockages.

Our solution prototype could be implemented in existing or new buildings, as long as they have the infrastructure required (WiFi, CCTV Cameras and smoke/fire sensors).

Pitch Video

Dr Watson Pitch Video

Solution Architecture

Detailed Solution

Read more

Getting Started

Distress Identifier

README.md

Mobility Aid Detection

README.md

People Counter

README.md

Route Finder (Flask RESTful Web Service)

Setup & Run locally

cd api
python3 -m venv create flask
source flask/bin/activate
flask/bin/pip install requirements.txt
./app.py

Find optimal route

Example 1: From Exit 4 to Corridor 3-1

curl -i "https://dr-watson.us-south.cf.appdomain.cloud/route?origin=Exit_4&destination=C3-1"

Example 2: To Corridor 3-1

curl -i "https://dr-watson.us-south.cf.appdomain.cloud/route?destination=C3-1"

Example 3: From Corridor 3-1

curl -i "https://dr-watson.us-south.cf.appdomain.cloud/route?origin=C3-1"

Update compromised zones

curl -X PATCH "https://dr-watson.us-south.cf.appdomain.cloud/compromised?zones[]=C2-1&zones[]=C3-1"

Reset zones

curl -i "https://dr-watson.us-south.cf.appdomain.cloud/refresh"

Proof of Concept

Link to our dashboard: Dr Watson: To The Rescue

Tools and Technologies

  • Cloud Computing
    • IBM Cloud Foundry
    • IBM Watson Visual Recognition
    • IBM Developer Model Asset Exchange: Audio Classifier
    • AWS Simple Storage Service
  • Frameworks
    • React
    • Flask
  • Programming Languages
    • Python 3
    • HTML
    • CSS
    • JavaScript
  • IoT
    • CCTVs
    • Sound sensors
    • Raspberry Pi (WIP)

About

A prototype smart evacuation and rescue system that leverages data from WiFi, CCTVs and sound sensors to locate crowds, identify trapped or vulnerable occupants, compute optimal escape routes and provide real-time visualization of vital information for quick and effective decision-making by rescuers. 1st Prize @ the SCDF X IBM - Lifesavers' Inno…

Topics

Resources

License

Stars

Watchers

Forks

Languages

  • SCSS 52.4%
  • CSS 23.4%
  • JavaScript 18.7%
  • Python 5.3%
  • HTML 0.1%
  • Dockerfile 0.1%