/symbol-detector

Real-time shape detection with camera frame analysis.

Primary LanguagePythonMIT LicenseMIT

Symbol detector

Real-time shape detection with camera frame analysis.

This program was developed for my thesis in 2017 and the repo contains just a slightly modified version.

In that time I had no concept about TDD and Clean Code, that's why the code is messy and doesn't have tests.

Operation

The program analyzes each frame and finds the pointer. It stores the pointer's coordinates from appearing to disappearing in blocks. When a block is defined, it sends to detector worker.

It was only tested on Ubuntu based linux, it may require some minor modifications to run on Windows.

Requirements

  • python >= 3.8.0
  • poetry
  • tcl/tk

Usage

  • Create a python virtual environment.
  • Activate the environment and run poetry install.
  • Find an illuminating object that can be easily distinguished from the background.
  • Make sure that you have a camera.
  • Run symbol-detector-gui script.
  • Go to Settings -> Camera probe and calibrate the camera. The calibration is correct when only the expected object (pointer) can be shown on the screen.
    • Refresh the frame with r and adjust blur and exposition.
    • Find the center of the pointer
    • Left click
    • Find the border of the pointer
    • Right click
    • Middle click somewhere. If a red circle appears, the calculation was successful.
    • Press q
    • If you want to improve your selection, press d and go from beginning.
  • Select the symbol set.
  • Start the detection.