Notifications
Clear all

Scan results not correct.


Hamishi
Posts: 31
Topic starter
(@hamishi)
Eminent Member
Joined: 2 years ago

Hi,

I'm running V0.8.2 and I have problems with scanning objects that have surfaces that are beyond the centre of the table in whdat I would call a negative position.

I first recognized the problem with objects that basically are rotation symmetric but a bit skewed. Handcrafted chess figures. 

I placed them with the based centered as much as possible, but then the top of the object is not on the Z-Axis. The resulting scan looks normal for the parts where the surfacepoints have a positive position with respect to the origin of the table. The scan looks like a hole for the parts where the surface-points are in a negative position from the origin of the table.

In order to investigate this more I put the object centered on one of the outer rings of the table. only part of the surface looks normal (about 90 degrees of the total scan) the rest does not look like the surface at all.

Any hits what can cause this? I think my laser, calibration and camera alignment are OK.

I also have sometimes a disturbance with the software. It happens that during the second phase of the scan the object is displayed with a horizontal axis in stead of with the vertical axis. When I load the file the orientation is OK, but when I do a new scan the axis problem is back again. The only way I can revert to normal operation is to reboot the PI.

Overheating of the PI also occurs often, even in a place where it is only 15 degrees Celsius. I'm thinking of installing a fan or heatsinks on the PI.

Regards,

Hamishi

Reply
17 Replies
mario
Posts: 566
Admin
(@mario)
Member
Joined: 4 years ago
Posted by: @hamishi

I first recognized the problem with objects that basically are rotation symmetric but a bit skewed. Handcrafted chess figures. 

I placed them with the based centered as much as possible, but then the top of the object is not on the Z-Axis. The resulting scan looks normal for the parts where the surfacepoints have a positive position with respect to the origin of the table. The scan looks like a hole for the parts where the surface-points are in a negative position from the origin of the table.

Some pictures would help me to understand how the object should look and how the scanned result looks.

Posted by: @hamishi

I also have sometimes a disturbance with the software. It happens that during the second phase of the scan the object is displayed with a horizontal axis in stead of with the vertical axis. When I load the file the orientation is OK, but when I do a new scan the axis problem is back again. The only way I can revert to normal operation is to reboot the PI.

Can you also provide some screenshots of this issue?

Posted by: @hamishi

Overheating of the PI also occurs often, even in a place where it is only 15 degrees Celsius. I'm thinking of installing a fan or heatsinks on the PI.

Yes a fan is recommended. I am doing some further investigations on temperature problems at the moment.

Reply
Hamishi
Posts: 31
Topic starter
(@hamishi)
Eminent Member
Joined: 2 years ago

The requested screenshots.

The rotation axis problem for a new scan seems to emerge after loading a saved scan as I did to produce this screenshots.

Reply
Hamishi
Posts: 31
Topic starter
(@hamishi)
Eminent Member
Joined: 2 years ago

 

Capture2
Capture3
Capture1
Capture4
Capture5
Capture6

The requested screenshots.

The rotation axis problem for a new scan seems to emerge after loading a saved scan as I did to produce this screenshots.

Reply
mario
Admin
(@mario)
Joined: 4 years ago

Member
Posts: 566
Posted by: @hamishi

 

Capture2
Capture3
Capture1
Capture4
Capture5
Capture6

The requested screenshots.

The rotation axis problem for a new scan seems to emerge after loading a saved scan as I did to produce this screenshots.

This is a known bug in the UI. I think i will fix that within the new UI. It looks like your calibration values are bad. Can you post the default.config.json file?

What is the black background? It looks very shiny. This can cause calibration errors. A mat black would be better. Black cardboard or something else what does't reflect light that much. ( as a general suggestion, maybe that this background causes bad calibration but i am not sure)

Which pi cam are you using? Have you changed the resolution?

Reply
Hamishi
Posts: 31
Topic starter
(@hamishi)
Eminent Member
Joined: 2 years ago

The black background is a piece of an anti-slip foil used in car boots to avoid things slipping around in the boot. I just placed it behind the table to see if this would improve the scan results. The calibration was done before this black surface was put in the scanner. I re-calibrated after that I had changed the step value to 6400 and pulling the pads for MS1 and MS2 to +5V.

 

 

Reply
Hamishi
Posts: 31
Topic starter
(@hamishi)
Eminent Member
Joined: 2 years ago

I use a PiCAM V2 (green circuit board) and did not change any resolution.

Reply
mario
Admin
(@mario)
Joined: 4 years ago

Member
Posts: 566

@hamishi

Posted by: @hamishi

and did not change any resolution.

I was just confused because the first image with the background shows only a part of the table. That was the reason why i thought that the resolution was changed.

Reply
Hamishi
Posts: 31
Topic starter
(@hamishi)
Eminent Member
Joined: 2 years ago
Capture7

My camera alignment.

Reply
mario
Posts: 566
Admin
(@mario)
Member
Joined: 4 years ago
Posted by: @hamishi

My camera alignment.

Your config shows me that you are scanning with just one laser, in this case the laser/camera alignment is not that critical as using multiple lasers. I think your calibration values look good. But can you do another calibration and send this values too? I need to compare multiple values.

Reply
Hamishi
(@hamishi)
Joined: 2 years ago

Eminent Member
Posts: 31

@mario

From that picture the table seems not to be exactly in the middle. I tried to align the centre of the table with the vertical green line as in the documentation. 

I will repeat everything again tomorrow an share the results.

Reply
Hamishi
(@hamishi)
Joined: 2 years ago

Eminent Member
Posts: 31

@Mario

Capture8

 

As promised yesterday. 

The scan results have improved but are still not perfect.

What did I do:

1. Realign the camera to cross the green line more accurate over the center of the scan table;

2. Realign the camera to get a more accurate vertical alignment by using the calibration sheet.

3. Cut-off the excess parts of the calibration sheet, it was hitting the wall of the scanner;

4. Calibrate the scanner again;

5. Perform some scans of the object;

6. Download the default.config.json file and save it as default.config1.txt.

I had to skew the camera PCB with another 1 mm to align it with the center.

Questions:

1.Does it matter if the camera is not perfectly vertical aligned? And if this is the case where should the crossing of the two lines point to?

2. Does it matter during calibration that I get the overheating/throttle warning? It now looks like that the lines are not always with the same angle between them.

Regards,

Hamishi

 

 

Reply
Danilo0702
Posts: 393
(@danilo0702)
Reputable Member
Joined: 3 years ago

@hamishi your config

 

""""""""......

"folders": {
"www": "/usr/share/fabscanpi/",
"scans": "/home/pi/scans/"
},
"online_lookup_ip": "8.8.8.8",
"laser": {
"color": "R (RGB)",
"numbers": 1,    -------------------------------------((when you hav 2 lasers-change numbers to 2))
"interleaved": "False"
},
"scanner_type": "laserscanner",
"calibration": {
"weight_matrix": [],
"dist_camera_matrix": [],
"pattern": {
"square_size": 10,
"rows": 6,
"columns": 8,
"origin_distance": 30.2 ----------------------((change to 35.0 mm and look to the scan))
},

.............."""""

Reply
Hamishi
(@hamishi)
Joined: 2 years ago

Eminent Member
Posts: 31

@danilo0702

30.2 is the distance I measured from the table surface to the top of the first black square on the bottom row. This is the value according to the instructions for the filling in the values in the default.config.json. My calibration sheet is sitting right on the table without anything beneath it. So I don't see the reason why to change it to 35 which is a value like the default 37.2 when using the printable stand for the calibration sheet.

Reply
mcinerney
(@mcinerney)
Joined: 3 years ago

Trusted Member
Posts: 60

@hamishi

Posted by: @danilo0702

"square_size": 10, <-------??
"rows": 6,
"columns": 8,

Is your calibration sheet 10mm or is it 11mm?

Reply
Hamishi
Posts: 31
Topic starter
(@hamishi)
Eminent Member
Joined: 2 years ago

@mario,

I still cannot get decent results unfortunately. Could you have another look at the calibration results? I'm suspecting that the calibration process does not work correct with only 1 laser in the system and the V 0.8.2 version for 2 lasers set up for 1 laser.

I looked how the calibration process is going on. In the last part I see a few times that the laser comes on for the right side and after that it continues with what looks like calibrating the left side but of course there is no laser line present then.

Why I assume something in the calibration is going wrong? Parts of the scanned surfaces look very nice, however there is a kind of ghost image at an angle going through the scan, it is not 90 degrees but more like the angle that the lasers are supposed to aim at the object. I don't have a rotation of the object that is more then 360 degrees (400 step motor 6400 steps TMC2208 problem)

Thanks in advance.

Reply
Hamishi
Posts: 31
Topic starter
(@hamishi)
Eminent Member
Joined: 2 years ago

actually it is 10.555 mm for 9 squares I measure 9.5 cm

Reply
Hamishi
Posts: 31
Topic starter
(@hamishi)
Eminent Member
Joined: 2 years ago

I changed it to 11, rebooted and calibrated again. The results improved a lot but still it looks that some parts of the surface are placed in the point cloud with a vertical and angular offset. That looks like a sync problem in the scan

Reply
Share: