GitHub - teezeit/monocular_invehicle_visual_odometry: This code uses the images of an in-vehicle dashcam to estimate the position of the car in a series of consecutive images. It uses the FAST keypoint detection algorithm combined with the opticalFlowLK keypoint matching algorithm.

teezeit / monocular_invehicle_visual_odometry Public

Notifications You must be signed in to change notification settings
Fork 1
Star 1

This code uses the images of an in-vehicle dashcam to estimate the position of the car in a series of consecutive images. It uses the FAST keypoint detection algorithm combined with the opticalFlowLK keypoint matching algorithm.

1 star 1 fork Branches Tags Activity

Star

Notifications

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 8 Commits
output		output
pics		pics
cars.xml		cars.xml
documentation.pdf		documentation.pdf
f.py		f.py
get_positions.py		get_positions.py
readme.txt		readme.txt

Repository files navigation

Check the documentation.pdf to see what this is about!

make sure you have packages:

numpy
opencv (3)!
matplotlib
scipy
mpl_toolkits
math
os


then just go and run: python get_positions.py
f.py contains all the helper functions doing the dirty work.

I work on an Ubuntu 15.10


The script:

Reads in the images
Findes the keypoints
matches the keypoints
estimates the Rotation and Translation
Visualises the result

creates an ./output/ folder with the results


OUTPUT:
./kps/          - the keypoints found in the images
./matches/      - the visualisation of the flow of the keypoints between consecutive images
./result/       - the 2D mapping and the 3D- coordinate plot
./logfile.txt   - the indivdicual rotation and translation of each step and information if the sanity check was negative.