Skip to main content

[ALTERNATIVE DOCUMENTATION] Lifeguard AI

image04-LifeguardAi (1).jpg

 

 

 

 

 

 

 

 

Contributors: Hendrik Hagala (Research)

Problem

Drowning is the 3rd leading cause of unintentional injury death worldwide, accounting for 7% of all injury-related deaths. There are an estimated 320 000 annual drowning deaths worldwide.

Solution

The idea is to create an ocean rescue robot that detects distressed swimmers, sends out alerts to the person responsible and drives autonomously to victims to help them by providing easy grip handles to grab and secure themselves.  

Industry

There are about 4 types of solutions on the market for rescuing distressed swimmers:

1) Detectors for pools that are detecting distressed swimmers and sending out alerts to lifeguards.

Examples: Coral Drowning Detection, SwimEye, Poseidontech, Laifeguard

2) Detectors for beaches, using cameras above the water to detect swimmers in distress.

Example: Sightbit

3) Beach and ocean rescue robots that do not detect drowning swimmers but that are remote controlled to to drive to distressed swimmers and they are working as a buoy where the person can grab and hold above the water.

Examples: EmilyRobot.  

This robot is for beach and ocean rescue. Can be thrown from helicopter, boat, from the beach. Detects and goes to distressed people so he/she can hold on something. Speed: 23 MPH.

4) Wristband that keeps track of guests while swimming or playing in the pool and if a user stays too deep for too long the Sentag Drowning Detection system will immediately alert lifeguards.

Example: Sentag

Below are a few drowning detection projects (using AI and body-mapping techniques) that might be helpful in your R&D:

 

https://www.diva-portal.org/smash/get/diva2:1337391/FULLTEXT02

This study first investigates to find the suitable deep learning algorithms that can be used to detect objects and then an experiment is performed with the chosen algorithms to state the possibility to detect humans in an underwater environment and then evaluate performance of algorithms. Analysis showed that the Faster RCNN algorithm gave the best results, being able to detect and track humans in an underwater environment with a maximum confidence level of 99% even when humans are seen in various angles, postures, and variable lighting conditions.  

 

https://github.com/Nico31415/Drowning-Detector

This program will detect if a person is drowning. This project is still a work in progress, so it can only be implemented with a computer's webcam, and doesn't work completely yet.

 

https://github.com/sswarnak77/LifeGuard.IO

Their website is off and not sure it’s possible to contact them but they are giving quite a good overview about their work progress.

 

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7036163/

This is a presentation of drowning detection system on coastal lines using image processing techniques and neural network

 

https://github.com/Reema1234ag/Drowning-Risk-Analysis

Using YOLO object detection, this program will detect whether a person is drowning or not. This software can be used with a Raspberry Pi Camera, which can then be placed underwater with an appropriate case.

 

Possible Equipment

AI camera: https://store.opencv.ai/collections/frontpage/products/oak-1 or https://www.mouser.com/ProductDetail/Luxonis/OAK-1?qs=%2Fha2pyFaduhFBpt8h%252Bork1LvGTDcd88Sc%252BpmKgBYCR0%3D suggested by Wesley

 

Product’s requirements

https://docs.google.com/spreadsheets/d/1ST-2q7jssu5LGNBINQJXhInl6gcsGW56cBtMkw3a0r8/edit?usp=sharing

ADVANCED

BASICS

FUNCTIONALITY

 
 

Software

 
 

Do the cameras have 360 degree view above the water?

 
 

Do the cameras have 360 degree view under the water?

 
 

Can it detect swimmers in distress 60 M from the SeaPod?

 
 

Can it detect swimmers in distress from the depth of 30 M?

 
 

Can it detect swimmers in distress in the dark?

 
 

Can it detect swimmers in distress when it's raining?

 
 

Can it detect swimmers in distress when it's foggy?

 
 

Can it detect swimmers in distress with high waves?

 
 

Is the detection accuracy at least 95%?

 
 

If there are more than one victim in separate locations then

is it smart enough to make the right choice based on condition, age, gender?

 
 
 
 
 

Rescue Robot

 
 

Is the rescue robot autonomous?

 
 

-approaches victims itself?

 
 

-bring the victim to previously specified location

 
 

Can you switch from autonomous mode to manual?

 
 
 
 
 

Is it working on electricity?

 
 

Is the duration 12-14 min full power run time (equal to 70-100 rescues at 100 yards)?

 
 

Is the battery operational range of 5 + miles at full power, 6 + miles at slower speed?

 
 

Does it have a quick charge?

 
 

Can it be thrown to the water from 10 m (landing on the right side is required)?

 
 

Is the top speed at least 35 km/h?

 
 

Can it drive accurately in the dark?

 
 

Can it drive accurately when it's raining?

 
 

Can it drive accurately when it's foggy?

 
 

Can it drive with high waves?

 
 

Can victims grab from the rescue robot's easy to grip handles and secure themselves?

 
 

Can the rescue robot bear 3 people at the same time?

 
 

Is the rescue robot stable enough for the victim to pull itself on top?

 
 

Is the propeller in a safe place so victims won't harm themselves?

 
 

Does it have a long flag on top to ensure easy visibility?

 
 

Does it have waterproof internal modular components?

 
 

Can the rescue robot grab from the victim when he/she is unconscious and floating on the water?

 
 

Can the rescue robot grab from the victim when he/she is unconscious on the ocean floor?

 
 

Can the rescue robot grab from the victim when he/she is unconscious under the water?

Project video link:

https://www.dropbox.com/s/qwvrapqwazhvcjl/LifeGuardAI.mp4?dl=0

 

This project is being developed as an open-source project with the following licensing: