Camera position  

  RSS
smassard
(@smassard)
New Member

Hi,

After building the Fabscan i realized the camera position isnt centred to the turntable:

As a result i cant see the whole turntable and the center of the turntable is not in the center of the preview image (crossed lignes). 

Is this a problem?

Regards

M.S.

Quote
Posted : 27/02/2019 12:25 pm
ErikVdbroeck
(@erikvdbroeck)
Active Member

Related question here...

- No problem to align the laser vertically against the backplate and going through the center of the turntable,

but I'm reading contradictory comments so can't figure out:

* if both lines of the green crosshair have to go through the center point C of the turntable ?

* Or if only the vertical green crosshair line has to cross the center of the turntable and the horizontal one is covering the line T where the bottom plate is touching the back plate ?

- I had mounted the camera/ledring plate in the predrilled holes, but then the crosshair is way above the C-point and also way above the T-line.

I tried to use an adjustable spring system, but that makes me incline the camera plate about 25 degrees down, so not even half of the calibration sheet was shown on the cam image. 

Should I bring the camera plate 4,5 cm down ?? This would mean the bottom of the camera/ledring-plate will almost be touching the floor of the cube. Any ideas about this please ?

ReplyQuote
Posted : 27/02/2019 8:42 pm
rene liked
rene
 rene
(@rene)
Member Admin

We will write a detailed tutorial about the calibration process. Some things need better documentation, but maybe there are also some bugs. We need to take some time for further investigation. If things are obvious bugs, please create a github issue.

ReplyQuote
Posted : 28/02/2019 10:22 am
ErikVdbroeck
(@erikvdbroeck)
Active Member

Looking forward to the tuto!

But maybe you could give us a quick hint, there, about where to point the camera crosshair ??

I only get uncoherent scan results.

1551378457-06-results-still-bad.jpg
ReplyQuote
Posted : 28/02/2019 6:27 pm
ErikVdbroeck
(@erikvdbroeck)
Active Member

(continued)

I only get uncoherent scan results.

See pics of  vertical aligned laser going through center of turntable, spring-adjustable cam-module from standard predrilled holes position so that vertical green crosshair line is cutting laser line in center of turntable, horizontal green line projected towards back plate, calibration using 3d printed calibration stand and the cal-board and waited until I got the message 'calibration finished'.

I also tried to point the camera more downward, to make the crosshair cover the line where the bottom plate is touching the back plate, but then only 2.5 lines of calibration squares are visible and the lower half of the cam image shows the empty bottom zone in front of the turntable: no positive result either.

I even tried a 3rd camera positioning so that the crosshair is in the middle of the white center square on the cal board and recalibrated.

Still lousy scan result afterwards.

1551378721-tryouts1-2-3-4-5.jpg
ReplyQuote
Posted : 28/02/2019 6:32 pm
ErikVdbroeck
(@erikvdbroeck)
Active Member

(continued II)

The calibration log file looks Ok and so does the default.config.json file with the calibration values. I include them here !

{
"folders": {
"www": "/usr/share/fabscanpi/",
"scans": "/home/pi/scans/"
},
"laser": {
"numbers": 1
},
"scanner_type": "laserscanner",
"calibration": {
"weight_matrix": [],
"dist_camera_matrix": [],
"pattern": {
"square_size": 11,
"rows": 6,
"columns": 8,
"origin_distance": 37.2
},
"camera_matrix": [
[
1326.592,
0.0,
637.799
],
[
0.0,
1319.824,
843.751
],
[
0.0,
0.0,
1.0
]
],
"distortion_vector": [
0.284,
-2.132,
0.002,
-0.002,
8.442
],
"laser_planes": [
{
"deviation": 0.05261286110996645,
"distance": 88.15394578052423,
"normal": [
0.8271677915811131,
-0.004239319279079676,
0.5619390293820804
]
}
],
"platform_translation": [
-1.006600040698955,
61.116599267151074,
159.2657002048133
],
"platform_rotation": [
[
0.0,
0.9997605399160551,
-0.021882934601150067
],
[
0.01914342610107397,
-0.021878924505851983,
-0.9995773316255127
],
[
-0.9998167478278771,
-0.0004189143414117509,
-0.019138842014652813
]
]
},
"keep_raw_images": "True",
"meshlab": {
"path": "/usr/bin/"
},
"process_numbers": 4,
"turntable": {
"steps": 6400,
"radius": 70
},
"camera": {
"rotate": "True",
"type": "PICAM",
"vflip": "False",
"preview_resolution": {
"width": 800,
"height": 600
},
"hflip": "True",
"resolution": {
"width": 1640,
"height": 1232
}
},
"discoverable": "True",
"serial": {
"plattform_type": "fabscanpi",
"baudrate": 57600,
"autoflash": "True",
"port": "/dev/ttyAMA0"
},
"texture_illumination": 40
}

 

ReplyQuote
Posted : 28/02/2019 6:33 pm
mario
(@mario)
Member Admin

Hello, 

the holes in the case are aligned in a way that the cam should exactly match the center of the table. One thing i realised in the last two years is that not all lenses are glued perfectly onto the pcb, this is generally regarded to the v2 module. Have a closer look to your camera module. Sometimes it helps to press the "lens module" gently into the right direction. I will try to add this as soon as possible to the documentation.

- Mario

ReplyQuote
Posted : 02/03/2019 6:40 pm
ErikVdbroeck
(@erikvdbroeck)
Active Member

Thanks for the info, Mario. I've got a model 2 like this, but using the cam doesn't look towards the center of the table at all. Not a matter of a just a few degrees, as it's looking straight towards the middle of the opposite wall. So, following your description, I would have to point it about 30 degrees downwards...

Can't do any better than this: see attached pic (unless I completely misunderstand?)

A picture of the actual calibration crosshair would tell more than words, I guess. Looking forward to it

1551896461-ScreenHunter_-2019-03-0626.jpg
ReplyQuote
Posted : 06/03/2019 6:21 pm
ErikVdbroeck
(@erikvdbroeck)
Active Member

...tried this way, no success: but as I mentioned in my first post, the cal-board comes on the screen for about 5 squares high, cal-process starts and announces 12 minutes left, but halfway (6 minutes left) it seems to choke as I hear stepping motor continues to be driven every like 10-15 secs, but screen does not refresh anymore and after 30-40 minutes the stepping motor is still doing the same.

1551896737-calbord.jpg
ReplyQuote
Posted : 06/03/2019 6:25 pm
mario
(@mario)
Member Admin

Hi Erik, 

sorry but i think the crosshair is misunderstood. The important part is that the vertical line hits the center of the turntable. The horizontal line does not matter at all. I think i have to remove it in the next release. It only marks the horizontal center of the image. For the horizontal alignment it is more important to measure the origin point from the table to the top of the first square line on the calibration sheet (like it is described in the docs).

The camera should be aligned in a way that you can see the table at the bottom of the image and as much of the object as possible. And of course vertical center alignment as described above.

In the following video you can see how my cam is aligned. ( This video is older. The UI can differ from the current one.)

https://www.youtube.com/watch?v=OFSEnqRDmIE

- Mario

This post was modified 3 months ago 2 times by mario
ReplyQuote
Posted : 06/03/2019 6:32 pm
Share: