Some problems when collecting patterns

[yangkaiarvin]

I created a subscriber to listen to the topic "/camera/pattern/stereo", but there is something wrong when it comes tolast_stereo_pattern_ = get_pattern_buffers.result()[itmStereo][0].asJson (true);in ensenso_grabber.cpp with the camera seeing the calibration pattern. The program stopped with no response. How can I get the pattern data from "/camera/pattern/stereo"?

Best Wishes!

[fsuarez6]

Yes, this is not working as expected. Could you please share the details of your system? In particular the following:

	Version
Camera	N35-805-16-BL
NxLib	2.0.147
uEyeDriver	4.81.18
CodeMeter	6.40.2402

You can get these details running:

$ nxView

and hovering the mouse over the lower-right corner

[yangkaiarvin]

Version
Camera N30-602-16-BL
NxLib 2.0.147
uEyeDriver 4.82.16
CodeMeter 6.40.2402

When we use the collect_srv_ = nh_.advertiseService("collect_pattern", &EnsensoDriver::collectPatternCB, this);, sometimes we will have getPatternCount()>0 but mostly the patterncount was 0. And the program keep running with no response.

[fsuarez6]

We have two use cases for the /camera/collect_pattern service:

Camera-robot calibration or handeye calibration

You want determine the position of the camera relative to a robot. The camera can either be fixed with respect to the robot's base or it can be mounted on the robot's hand and move with it.
More details:

• How to: Handeye calibration
• Pattern buffer

MWE for the calibration

Start the ensenso driver:

$ roslaunch ensenso viewer.launch rqt:=False rviz:=False

Configure the driver to only stream the images and switch off the projector. The front light will depend of your illumination conditions:

$ rosrun dynamic_reconfigure dynparam set /camera/ensenso_driver "{'Images':true, 'Cloud':false, 'Projector':false, 'FrontLight':false}"

The /camera/left/image_raw topic contains the raw image and when the pattern has been detected an overlay will appear in the image:

rosrun image_view image_view image:=/camera/left/image_raw

Collect the pattern observations. For the first call, clear the buffer:

$ rosservice call /camera/collect_pattern "add_to_buffer: true
clear_buffer: true
decode: true
grid_spacing: 0.0"
# Output
success: True
pattern_count: 1
grid_spacing: 18.75

For the subsequent calls, do not clear the buffer:

$ rosservice call /camera/collect_pattern "add_to_buffer: true
clear_buffer: false
decode: true
grid_spacing: 0.0"
# Output
success: True
pattern_count: 2
grid_spacing: 18.75

The pattern_count should increase with every observation.
For each observation you must know the robot pose.
Once you have enough observations (we usually collect 25) you can call the /camera/calibrate_handeye service.

Pattern pose estimation

You just need the pattern pose with respect to the camera frame.

MWE for the pattern pose estimation

Start the ensenso driver, configure it to only stream the images and switch off the projector.

Collect the pattern:

$ rosservice call /camera/collect_pattern "add_to_buffer: true
clear_buffer: true
decode: true
grid_spacing: 0.0"
# Output
success: True
pattern_count: 1
grid_spacing: 18.75

Estimate the pattern pose:

$ rosservice call /camera/estimate_pattern_pose "average: false" 
# Output
success: True
pose: 
  position: 
    x: 0.140828834732
    y: 0.154744567362
    z: 1.01843195859
  orientation: 
    x: -0.158285623859
    y: -0.183593665657
    z: -0.708303946633
    w: 0.662981558111

Hope this helps!

[yangkaiarvin]

Problems solved, thank you!