Computer Vision exercises

Computer vision exercises using OpenCV

Summary

1. 3D Perception (folder: Perception3D)
   1. Calibration
   2. Rectification
   3. Monovision
   4. Stereovision
2. Segementation & Registration (folder: Segmentation_Registration)
   1. Segmentation
   2. Registration
      1. RANSAC
      2. ICP
3. 3D Localisation (folder: Localisation)
   1. Pose estimation
   2. Tracking

Papers

Report

A report (in french) for every part can be found here. You can find mathematical explanations and visual results.

3D perspective

An article (in french) explaining the 3D Perspective mostly used for the Calibration can be found here

Usage

For convenient execution, you may use a python venv and then install the requirements with pip.

Create a python venv

$ python -m venv .venv
$ source .venv/bin/activate

Install the requirements

pip install -r requirements.txt

3D Perception

Script:

• Monovision: Perception3D/MonoMain.py
• Stereovision: Perception3D/StereoMain.py

Both script will do the calibration first and the rectification in second. The matrices cameraMatrix and distCoeffs will be printed.

Segementation & Registration

Script:

• Segmentation_Registration/Crop.py to get the reference object in a point cloud
• Segmentation_Registration/Main.py executes the registration from a source_file to a target_file (variable in the script)

Localisation

Script:

• Localisation/Main.py
  • Change the cameraMatrix and distCoeffs computed from the calibration.
  • The script will by default track a EPITA student card.
  • Feel free to track anything, the reference image must loaded in img_object variable