[paparazzi-commits] [paparazzi/paparazzi] 07317f: [modules] edgeflow noisemeasurement fix (#1820)

[GitHub]

  Branch: refs/heads/master
  Home:   https://github.com/paparazzi/paparazzi
  Commit: 07317f1be26e0c70264a217a7a082fd4e9bb81b9
      
https://github.com/paparazzi/paparazzi/commit/07317f1be26e0c70264a217a7a082fd4e9bb81b9
  Author: knmcguire <address@hidden>
  Date:   2016-07-22 (Fri, 22 Jul 2016)

  Changed paths:
    M conf/airframes/TUDELFT/tudelft_bebop_opticflow.xml
    M conf/modules/cv_opticflow.xml
    M sw/airborne/modules/computer_vision/opticflow/opticflow_calculator.c
    M sw/airborne/modules/computer_vision/opticflow/opticflow_calculator.h
    M sw/airborne/modules/computer_vision/opticflow_module.c

  Log Message:
  -----------
  [modules] edgeflow noisemeasurement fix (#1820)

The more corners the algorithm tracked, the higher the noise measurement (est 
velocity error = tracked corners/ max_corners) becomes. However this is not 
wanted for the ins_hff, since there a higher estimated velocity error results 
in less trust in that measurementof the Kalman filter. I made a new equation, 
which actually gives a lower noise measurement with more corners.

In the future there needs to be a better, with optitrack groundtruth backed up, 
noise model, however this pull request will hopefully give an intermediate 
solution. I also added an optional median filter if the hff is not used or if 
more outliers needed to be filtered out.