3D Stereo Vision research with Python and OpenCV

This is part of my code about my research with 3D Stereo Vision

Note: the project is from year ~2015 and this is my 1st experience with Python and C++. Anyway the examples may help you.

Advices

• Before start using real cameras, try to do the things with stamps from internet (photos and calibration)
• Before buying expensive industrial cameras, try to work with cheap USB cameras (it is super easy to use them in python with cv2.VideoCapture(0))
• If you intend to buy more expensive cameras check what SDK they have and is it ok for you to use it
• Don't start with fish-eye lenses (in fact don't use them), the less distortion you have, the easier it is to calibrate and examine
• Read little bit about lens
• Matlab Stereo Camera Calibrator is very useful when doing your research

C++ shooter

Uses the FlyCapture2 SDK. AsyncTriggerEx.cpp has hard-coded variable filesPrefix which is the directory where files will be saved. The program can works with array of cameras allCameras (You can push one or infinity cameras) and all will be triggered at the same time (with less than 0.001 diff) The output format is pgm which as far as I remember it was suitable for python.

v3 python scripts

1st check config.json and set your preferences

0.webcams_view.py

Will show the USB web cam live streaming

1.create_calibrations_from_images.py

Make calibration stamps when press the space key

1.create_calibrations_from_webcams.py

Live/automatic shooting from web cameras (you have to walk around like a monkey with a calibration board)

2.calibration.py

Crate calibration from the generated files/stamps

3.webcam_examples.py

No idea, I do not remember what that files do.

4.make_3d.py

The interesting part when you generate the 3D

5.extract-shapes.py

I'm not entirely sure if this script is working properly, but it should extract the 3D shapes

Bonus

2 Python scripts for car plate recognition