These are a few of the codes for the projects that I did as part of the Mobile Robotics Course.
All the codes are documented and there is a report which provides certain details of the implementation and results
I wrote a code to perform monocular camera calibration using the Zhang's method in order to find the K matrix.
The accuracy of the intrinsics was evaluated using reprojection error computation.
I wrote a code to perform two view sparse reconstruction of an image captured by a calibrated monocular camera.
Numerical stability issues in SVD were solved by normalizing the points. The sample output is given below.
This code is used to reconstruct a 3D cube using multi-view triangulation. For this, projection matrices of all 8 views were used to set up a least square system which was solved using SVD. The results for the same (reconstruction) are shown below:
I also implemented the trust-region strategy of LM algorithm for non-linear least square problems.