WARNING ! I developed this projects before following the Python projects at 42. The standard and the style may shock :)
This is a simple tool for gear tooth (not only) counting. It helps people counting occurences on rotation-regular shapes on pictures.
This is a presonal project built to try an idea : doing FFT on pictures.
demo.mp4
| Step | Result |
|---|---|
| 0) Raw image |  |
| 1) Shape is isolated using filters (Color selection, offset...), generating a new "black and white" picture. |  |
| 2) Image is "unrolled" using filtered image center of gravity as center (gears are always balanced) |   |
| 3) Gear profile is deduced by summing raws pixels |  |
| 4) Fast Fourrier Transform is applied to deduce harmonics. Amplitude of harmonics is used as "confidence level" |  |
| 5) Rays are plotted on raw image to allow user to vsualize and validate gear count |  |
Run with Python3 python3 SimpleGearToothCounter.py
Here are some examples of objects that can be "counted"
This requires Python 3
Following standard Python libraries are required :
numpy open-cv matplotlib imageio scipy thinker pyautogui skimage
For Windows installation, please copy and paste the following text in your Windows command prompt (CMD)
pip install numpy
pip install open-cv
pip install matplotlib
pip install imageio
pip install scipy
pip install thinker
pip install pyautogui
pip install scikit-image
Linux :
sudo apt-get install python3
sudo apt install python3-pip
pip3 install scikit-image
pip3 install opencv-python
pip3 install skimage
pip3 install pyautogui
pip3 install scipy
pip3 install imageio
pip3 install matplotlib
pip3 install numpy
sudo apt-get install scrot