Depth Estimation Using Stereo Cameras

Overview

Using a stereo camera, depth is determined by the concept of triangulation and stereo matching. By constraining the problem on a 2D plane known as the epipolar plane, we can simplify the formal search to a line search along the epipolar line.

Approach

1. Identify similar points from feature descriptors.
2. Match feature correspondence using a matching cost function.
3. Using epipolar geometry, find and match correspondence in one picture frame to the other.
4. Compute disparity from known correspondence.
5. Compute depth from known disparity.

To run the code

git clone https://github.com/Prat33k-dev/Stereo_Depth_Estimation.git
cd Stereo_Depth_Estimation/code
python3 stereo.py --filePath ../data_files/ --dataset 1

Parameters

• FilePath - Dataset Path file path. Default :- '../data_files/'
• dataset - Dataset Number . Default :- '1'

Results

CALIBRATION

Matches

RECTIFICATION

Epipolar lines for unrectified images

Epipolar lines for rectified images

CORRESPONDENCE

Disparity Map

DEPTH

Depth Map