Baby Suffocation Monitor

Baby Suffocation Monitor

Team members: Daniel Maryakhin, Vika Baranovsky, Matan Fintz.

Background

Babies are exposed to many hazardous situations in their first months.
One of them is suffocation during sleep while in their crib when they are with their nose towards the mattress without clear air way.
There are many other gadgets on the market for monitoring the baby such as Baby Sense, but they only alert AFTER the baby is not breathing and not before.

Our goal

We wanted to offer an alternative product that can foresee the danger before it occurs.
The idea was having 2 simple everyday webcams that can be found in most houses,
using them to detect the baby while in his crib and follow his face’s angles relatively to the mattress and alert when he might be going to turn his face down to the mattress or cover his face.

About the project

Equipment:

  • 2 webcams
  • 1 pc
  • 1 chessboard pattern

The setup (cameras and baby):

Baby only

Whole set

Cameras only

General process:

  1. Calibration of the cameras (using a chessboard pattern).

  2. Capture of inital (stereo) frame, straight frontal view and detection of the colored stickers.

  3. Triangulation and 3D representation of initial colors locations.

  4. Live capture of new (stereo) frames and detection of the colored stickers.

  5. Triangulation and 3D representation of the new color locations.

  6. Detection of the rotation/translation between the initial and the new frame using Kabsch’s algorithm.

  7. Representation of the baby’s location with 6 degrees-of-freedom.

  8. Monitoring and responding to dangerous angles or lack of detection.

Source code

Github project
Written in Python 3.5 using OpenCV

Demonstration video

The future…

Here are some possible ways to take the idea to the next level:

  • Use smart feature detection and get rid of the stickers.
  • Upgrade the system to support 3D models (and real babies) more accurately.
  • Upgrade the system to work in environments with any lighting conditions (maybe even total darkness, using IR or night vision, for example).
  • Add a small IOT computer device (such as Odroid, Arduino, Raspberry pi etc.) to make a stand-alone system.
  • Upgrade the method of danger notification.
  • Apply the concepts in this system to animal tracking.
Accessibility
Close Close