/VisualOdometry

Visual Odometry using SIFT feature point detection, PnP and Linear Triangulation. Extra Project : ENPM673 @UMD College Park

Primary LanguageJupyter Notebook

Visual Odometry:

We use filtered SIFT keypoint correspondences from RANSAC utilising the epipolar constraints to obtain the essential matrix, which is used to compute the camera poses. Cheirality Check is performed to ensure that the physically correct pose is selected out of mathematically possible 4 poses.

download Oxford Robot Car dataset from here and extract in ./Data folder

Run

python3 VisualOdometry.py