Added pictures

parent 818e2a51
== Vulcano Drilling Demo
This document describes the work done to adapt the drilling demo from the hi-shear drilling template to the new one, designed for the *Setitec driller*.
=== Detection of the template
Initially, the template's detection was almost never possible. The main problem appeared to be the differenciation between the background plane and the template plane.
Here is a list of the attempted changes :
* Change the *projector position* to different heights. Best results with projector pointing at the middle of the template. But low detection rate (less than around 25%)
* *Parameters tuning* in demo_tecnalia_lab package, _~/config/detection_inertia_param_torso_tecnalia.yaml_ (left arm) and _~/config/detection_inertia_param_arm_tecnalia.yaml_ (right arm).
Param tested : max and min distances from the scene, canny edge threshold. No significal improvement observed.
* Use a *white background* (with sheets of paper). Detection rate was improved (more than 50%), as the projected pattern is better seen compared to the dark initial color of the wing. However, it appeared that hiding the writing of Aernova (just above the template) made the detection impossible... Possible explanation : the different edges of the writing may be used by both cameras to detect similarities and construct the depth map.
* Add plastic wedges to create an *offset* and put more distance between the two planes that need to be differenciated. Tested with the dark background. Detection rate was improved (still less than 50%).
* *The dark coloured wing* used so far as a background was *removed*. We fixed the template directly on the support, at a distance of around 5 cm between the template and the wall to be able to insert an aluminium plate for real drilling tests. With this offset, the results were similar to the previous step.
PHOTO
[#img-sunset]
.A mountain sunset
image::sunset.jpg[Sunset,300,200]
* *White panels* were added to supress the offset between the background and the template to be more close to the final application. With this, we obtained the same random results as before, with a detection rate smaller than 50%.
PHOTO
* *Update camera* software and firmware to improve the smoothing of the depth map. Resulted in this output :
PHOTO () PHOTO
The distinction of the general shape of the template seems clear, with few outliers. However, the template is still not detected more than 50 % of the time
=== TCP of the new driller
It appeared that the driller was not correctly aligned with the holes. We followed these steps to correct it :
* New *TCP calibration*. It resulted in the exact same transform, so no improvement obtained
* *Offset* of the transform tuning. By adding small offset to the xyz transform, the driller was then pointing at the center of the holes when they are detected
These operations were repeated as we added the force sensors *OptoForce* at the end of the arm.
This diff is collapsed.
This diff is collapsed.
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment