Skip to content

Latest commit

 

History

History
74 lines (56 loc) · 2.29 KB

README.md

File metadata and controls

74 lines (56 loc) · 2.29 KB
Raw

Robo Videographer

Project for a automatic videographer system for recording sports matches.

Technical objective

The aim is to create a computer vision tracking system that can run on a Raspberry Pi using the camera module that tracks the action happening on a rugby pitch.

Implimentation goal

The final aim is to use the system to record a tag rugby matches with no human intervention.

Technology to use

  • Raspberry Pi
  • Raspberry Pi Camera module
  • OpenCV | SimpleCV
  • Video camera
  • Servo

Intial Design

  1. Get OpenCV/SimpleCV working on the Raspberry Pi with the camera.
  2. Build a program that tracks the pixel activity.
  3. Output activity value to a servo position.

Progress

Software

  • Get OpenCV/SimpleCV working on computer
  • Develope simple "action tracking"
  • Develope simple servo control on raspberry pi
  • Add "smooth action tracking"
  • Get OpenCV/SimpleCV working on raspberry pi
  • Merge action tracking with servo control
  • Optimise action finding code to run faster
  • Add concurrency between servo control and action detection using python processes

Hardware

  • Source tri-pod
  • Mount servo to tri-pod
  • Mount camera to tri-pod
  • Mount Pi to tri-pod
  • Mount Power to tri-pod
  • 3D model case
  • wire up power splitter

Extra Software

  • Load via new sd card
  • Add script auto-run with switch
  • Get Pi to act as wifi base
  • Connect android to Pi by wifi
  • View image from camera on android via wifi

Other processes

  • Collect training data
  • Build tool to help label training data
  • Label training data

Notes

Use sudo modprobe bcm2835-v4l2 to activate camera.

Resources

http://www.raspberrypi-spy.co.uk/2015/02/how-to-autorun-a-python-script-on-raspberry-pi-boot/

https://gpiozero.readthedocs.io/en/stable/recipes.html

https://en.wikipedia.org/wiki/Optical_flow

http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html

https://www.modmypi.com/blog/how-do-i-power-my-raspberry-pi

https://en.wikipedia.org/wiki/Recurrent_neural_network

Testing

  • Compare feature engineering with prediction speed and accuracy
  • Compare temporal framing sizes with prediction speed and accuracy