sensorfeaturetracking.jl's Introduction

SensorFeatureTracking

Algorithms to track features of interest, such as KLT. Please see documentation, and file issues or make suggestions as you see fit.

Note Features in this package are not yet optimized for speed, but the start of implementing machine/computer vision sparse feature functions that are useful to robotics.

Installing

This is an unregistered Julia package, but can be readily installed from a Julia console:

Pkg.clone("https://github.com/<YOURFORKEDREPO>/SensorFeatureTracking.jl.git")

Upstream master is at: https://github.com/JuliaRobotics/SensorFeatureTracking.jl.git

Examples

Please see, as part of the development, two cases in the examples folder. These examples currently extract corner features from either an image file, or a webcam image sequence.

Please note this code is not yet optimized in any way. Performance can be improved.

Unit Test

Basic unit tests are currently under development in the test folder. Unit tests can be run by typing:

Pkg.test("SensorFeatureTracking")

Contributors

This package uses several dependencies (thank you) and put together by: A. Hattingh, J. Terblanche, D. Fourie

sensorfeaturetracking.jl's People

Contributors

Stargazers

Watchers

sensorfeaturetracking.jl's Issues

implement equivalent of solvePnP

xref (thanks @yakir12 for pointing this out!):

JuliaGeometry/CoordinateTransformations.jl#30

In OpenCV this is called the solvePnP function, I've looked at it before, and its very long and hard to maintain. I've long wanted to rewrite

Also see:

JuliaRobotics/AprilTags.jl#83

Copying quote from @yakir12 (hope you don't mind):
"""
I feel like [JuliaRobotics/]CameraModels.jl package, together with the mechanics already in CoordinateTransformations.jl and Rotations.jl, plus some clever usage of autodifferentiation may very well be the ultimate way forward. If it's really solid we could build a similar interactive tool for calibration like AprilCal...
"""

I totally agree! I'm putting it all together in a common project, so we can track to see how what where is happening on this front.

Run femtocleaner

update to julia 1.0

update to compatible julia 1.0, please!!

Epic: Revive Minimum Sparse Feature Tracking

Can we run solvePnP on sparse features from Images.jl, do ransac, and compare with KLT feature tracking. Sparse features is old school. Future work can build out the semi-dense odometry methods, so point of this epic is just restore the best of the old school techniques for general use, boot strapping or comparison when developing newer techniques.

Project board: https://github.com/orgs/JuliaRobotics/projects/10/views/1

Recommend Projects