The goal of this exercise is to superimpose a virtual cube on a video of a planar grid viewed from diffeerent orientations using perspective projection, change of coordinate systems and lens distortion.
Given a dataset of images, with a set of 2D-3D correspondences for each image, and the camera matrix K. we have implement the DLT algorithm to estimate the camera pose [R/t] for each image.
We have implemented Harris corner detection, matching Descriptor and keypoint tracking on the popular KITTI dataset.
We have implemented the SIFT algorithm (descriptor) as simplified version of the original SIFT algorithm