☎

The start

Today I had a talk with Toby. I got some (seemingly) simple instructions for now:

  1. Read out an IMU with a programming language or to file I got a MinIMU-9 v2 (L3GD20 and LSM303DLHC carrier) of Pololu, which you can hook up with an Arduino.
  2. Track a camera’s distance with respect to a marker Toby suggested to use ArUco, which should be easy. I can use my laptop’s webcam, or a Logitech C920.
  3. Perform a simple kind of fusion between the tracked position and the IMU data Show that the IMU can improve the pose estimate of the tracker, for instance, by either pre- or postprocessing ArUco’s output.

There is also some theoretical stuff to do:

  1. Read in on Kalman filters The online lectures of Cyrill Stachniss were recommended. I should also do the homework.
  2. In Caarls’ PhD thesis, read up on specific Kalman filters and „continuous time processes” Because the IMU generates more often new data than the cameras, integrating this needs investigation.
  3. Collect papers on IMU – (stereo)camera fusion If finished with all, search for papers on SLAM/PTAM methods that introduce some method of fusing these two sensors.

I started with the first practical step. I installed the arduino package for Ubuntu, and did what the related Pololu software told me to do. To see if the provided drift correction works properly, I taped it down the table and let it run for some time. Results will follow with and without drift correction.

References

  • Jurjen Caarls. Pose estimation for mobile devices and augmented reality. PhD thesis, Delft University of Technology, 2009. [ bib ]
    
@phdthesis{caarls2009pose,
  title = {Pose estimation for mobile devices and augmented reality},
  author = {Caarls, Jurjen},
  year = {2009},
  school = {Delft University of Technology}
}