Issues with aligning scans

Something doesn't work and you don't know why? Ask here!
Post Reply
Rashiga
Posts: 18
Joined: 13 Oct 2019, 14:45

Issues with aligning scans

Post by Rashiga »

Hi, I've been trying to do some scans of simple objects but I'm failing to make the scans aligned.

My setup is a custom setup.
LED printer + 5 MP vison camera + 12mm c mount lens

The calibration panel is also custom made. The angle between the calibration panels are very close 90 degrees

From calibration, I get around 0.2 RMS error.

Any guidance would be highly appreciated!
Attachments
aligning issue.PNG
KakaoTalk_20191128_131034170.jpg
20191127_214558.jpg
User avatar
OBNRacerMan
Posts: 372
Joined: 12 Aug 2019, 09:46
Location: Obninsk, Russia

Re: Issues with aligning scans

Post by OBNRacerMan »

Rashiga wrote: 28 Nov 2019, 07:42 From calibration, I get around 0.2 RMS error.
I could never get an RMS value below 0.65 (usually 0.85-1.15) - but there weren’t any such serious “inconsistencies”.
P.S. I don’t know yet which way to go to fix your problem.
P.P.S. Usually, after calibration, I scan the calibration table and see how it looks - sometimes, for unknown reasons, it happens that its flat surfaces on the scan turn out to be curved - in this case, repeat the calibration.
Soft: RangeVision DIY, Calibration panel/Rotary tables - RangeVision, Prj: LG PF1500G, Cam: Daheng Mercury MER2-630-60U3M-L (USB3.0, monochrome) x2, Obj: ZLKC FM12036MP5 (F2.8/5Mp/12-36mm) x2. And a handheld scanner Creality Ferret
User avatar
Micr0
Posts: 586
Joined: 15 Nov 2016, 15:20
Location: New York City

Re: Issues with aligning scans

Post by Micr0 »

Rashiga wrote: 28 Nov 2019, 07:42 Hi, I've been trying to do some scans of simple objects but I'm failing to make the scans aligned.

My setup is a custom setup.
LED printer + 5 MP vison camera + 12mm c mount lens

The calibration panel is also custom made. The angle between the calibration panels are very close 90 degrees

From calibration, I get around 0.2 RMS error.

Any guidance would be highly appreciated!
Problems you are having are usually in calibration. Can you screen cap you calibration RMS figures? I was surprised at how much even tiny errors in the calibration panels matter in terms of the accuracy of scans. The panels I milled on a VMC with <.01mm accuracy were miles better than the same sized panels I printed on a laser printer or other method.
µ
Rashiga
Posts: 18
Joined: 13 Oct 2019, 14:45

Re: Issues with aligning scans

Post by Rashiga »

OBNRacerMan wrote: 28 Nov 2019, 08:04
Rashiga wrote: 28 Nov 2019, 07:42 From calibration, I get around 0.2 RMS error.
I could never get an RMS value below 0.65 (usually 0.85-1.15) - but there weren’t any such serious “inconsistencies”.
P.S. I don’t know yet which way to go to fix your problem.
P.P.S. Usually, after calibration, I scan the calibration table and see how it looks - sometimes, for unknown reasons, it happens that its flat surfaces on the scan turn out to be curved - in this case, repeat the calibration.
Thank you OBNRacerman. After your suggestion I did a scan of the calibration panel and found that it was way off from 90 degrees.
Micr0 wrote: 28 Nov 2019, 18:52
Rashiga wrote: 28 Nov 2019, 07:42 Hi, I've been trying to do some scans of simple objects but I'm failing to make the scans aligned.

My setup is a custom setup.
LED printer + 5 MP vison camera + 12mm c mount lens

The calibration panel is also custom made. The angle between the calibration panels are very close 90 degrees

From calibration, I get around 0.2 RMS error.

Any guidance would be highly appreciated!
Problems you are having are usually in calibration. Can you screen cap you calibration RMS figures? I was surprised at how much even tiny errors in the calibration panels matter in terms of the accuracy of scans. The panels I milled on a VMC with <.01mm accuracy were miles better than the same sized panels I printed on a laser printer or other method.
Thank you, Micro as always for your valuable input. Below I have attached the screengrab of the diagnostic panel. Seems like the angle between the calibration panel is way off. How can I fix this?
Attachments
diag 2.PNG
diag 1.PNG
User avatar
OBNRacerMan
Posts: 372
Joined: 12 Aug 2019, 09:46
Location: Obninsk, Russia

Re: Issues with aligning scans

Post by OBNRacerMan »

This is my calibration result :D :
My typical calibration result...
My typical calibration result...
And so it looks after scanning:
Scan result - front
Scan result - front
Scan result - top
Scan result - top
rms_03.jpg (10.32 KiB) Viewed 15397 times
Soft: RangeVision DIY, Calibration panel/Rotary tables - RangeVision, Prj: LG PF1500G, Cam: Daheng Mercury MER2-630-60U3M-L (USB3.0, monochrome) x2, Obj: ZLKC FM12036MP5 (F2.8/5Mp/12-36mm) x2. And a handheld scanner Creality Ferret
User avatar
Micr0
Posts: 586
Joined: 15 Nov 2016, 15:20
Location: New York City

Re: Issues with aligning scans

Post by Micr0 »

Rashiga wrote: 28 Nov 2019, 19:28
OBNRacerMan wrote: 28 Nov 2019, 08:04
Rashiga wrote: 28 Nov 2019, 07:42 From calibration, I get around 0.2 RMS error.
I could never get an RMS value below 0.65 (usually 0.85-1.15) - but there weren’t any such serious “inconsistencies”.
P.S. I don’t know yet which way to go to fix your problem.
P.P.S. Usually, after calibration, I scan the calibration table and see how it looks - sometimes, for unknown reasons, it happens that its flat surfaces on the scan turn out to be curved - in this case, repeat the calibration.
Thank you OBNRacerman. After your suggestion I did a scan of the calibration panel and found that it was way off from 90 degrees.
Micr0 wrote: 28 Nov 2019, 18:52
Rashiga wrote: 28 Nov 2019, 07:42 Hi, I've been trying to do some scans of simple objects but I'm failing to make the scans aligned.

My setup is a custom setup.
LED printer + 5 MP vison camera + 12mm c mount lens

The calibration panel is also custom made. The angle between the calibration panels are very close 90 degrees

From calibration, I get around 0.2 RMS error.

Any guidance would be highly appreciated!
Problems you are having are usually in calibration. Can you screen cap you calibration RMS figures? I was surprised at how much even tiny errors in the calibration panels matter in terms of the accuracy of scans. The panels I milled on a VMC with <.01mm accuracy were miles better than the same sized panels I printed on a laser printer or other method.
Thank you, Micro as always for your valuable input. Below I have attached the screengrab of the diagnostic panel. Seems like the angle between the calibration panel is way off. How can I fix this?
72° is way off. Those very small RMS numbers are not numbers we are interested. There is an aggregate RMS value which, if my understanding is correct, is the value of the volumetric calibration accuracy. David can compensate to some degree for panels that are not exactly 90° to each other and can compensate for panels that ate rotated slightly relative to each other. however it is better not to have any error then to try to compensate around it. the part of the panels that matters most for good calibration is the positioning of the dots. The grid of centroids needs to be as square and as positionally accurate as possible. Flatness of the panels is also very important. The best panels I've made have been milled out of Melamine, which is surprisingly flat.

Similarly, milled brackets to hold them in place are far superior to 3D printed ones. Post machining 3D printed brackets works well too.
I also mill alignment marks for a quick way if insuring that the panels are assembled properly.

I was surprised at how much better my scanning got when I had carefully made panels.
IMG_20180212_210456.jpg
IMG_20180309_143026.jpg
µ
Rashiga
Posts: 18
Joined: 13 Oct 2019, 14:45

Re: Issues with aligning scans

Post by Rashiga »

OBNRacerMan wrote: 28 Nov 2019, 20:36 This is my calibration result :D :
rms_01.jpg
And so it looks after scanning:
rms_02.jpgrms_03.jpg
Thanks for sharing your calibration data OBNRM!
Micr0 wrote: 29 Nov 2019, 21:06
Rashiga wrote: 28 Nov 2019, 19:28
OBNRacerMan wrote: 28 Nov 2019, 08:04 I could never get an RMS value below 0.65 (usually 0.85-1.15) - but there weren’t any such serious “inconsistencies”.
P.S. I don’t know yet which way to go to fix your problem.
P.P.S. Usually, after calibration, I scan the calibration table and see how it looks - sometimes, for unknown reasons, it happens that its flat surfaces on the scan turn out to be curved - in this case, repeat the calibration.
Thank you OBNRacerman. After your suggestion I did a scan of the calibration panel and found that it was way off from 90 degrees.
Micr0 wrote: 28 Nov 2019, 18:52 Problems you are having are usually in calibration. Can you screen cap you calibration RMS figures? I was surprised at how much even tiny errors in the calibration panels matter in terms of the accuracy of scans. The panels I milled on a VMC with <.01mm accuracy were miles better than the same sized panels I printed on a laser printer or other method.
Thank you, Micro as always for your valuable input. Below I have attached the screengrab of the diagnostic panel. Seems like the angle between the calibration panel is way off. How can I fix this?
72° is way off. Those very small RMS numbers are not numbers we are interested. There is an aggregate RMS value which, if my understanding is correct, is the value of the volumetric calibration accuracy. David can compensate to some degree for panels that are not exactly 90° to each other and can compensate for panels that ate rotated slightly relative to each other. however it is better not to have any error then to try to compensate around it. the part of the panels that matters most for good calibration is the positioning of the dots. The grid of centroids needs to be as square and as positionally accurate as possible. Flatness of the panels is also very important. The best panels I've made have been milled out of Melamine, which is surprisingly flat.

Similarly, milled brackets to hold them in place are far superior to 3D printed ones. Post machining 3D printed brackets works well too.
I also mill alignment marks for a quick way if insuring that the panels are assembled properly.

I was surprised at how much better my scanning got when I had carefully made panels.
IMG_20180212_210456.jpg
IMG_20180309_143026.jpg
Thank you, Micro for the information.

Btw I found the culprit, I had turned on the "CompensateAngleBetweenPatterns" option in calibration settings ones I removed it I was able to get my scans aligned.

I will also have to invest time to get some proper calibration panels made. I'm thinking of giving a try using two glass panels with calibration pattern printed in a sticker paper, also I will look into the possibility of machining melamine sheets.

Yes, I agree with you regarding the brackets getting direct 3d printed, they have issues with tolerances. So planning to get aluminum brackets machined for the next iteration.

Btw cool setup you have got there. If I am not mistaken is that two DSLR lens fitted to your vision camera?
User avatar
Micr0
Posts: 586
Joined: 15 Nov 2016, 15:20
Location: New York City

Re: Issues with aligning scans

Post by Micr0 »

Rashiga wrote: 06 Dec 2019, 13:55
OBNRacerMan wrote: 28 Nov 2019, 20:36 This is my calibration result :D :
rms_01.jpg
And so it looks after scanning:
rms_02.jpgrms_03.jpg
Btw cool setup you have got there. If I am not mistaken is that two DSLR lens fitted to your vision camera?
Thanks.

Yes. Those are 2 Nikon 18mm prime lenses. Used manual focus, manual aperture lenses are cheap, and if you get Name brand are superior to CCTV lenses by far. the major difference is that the focal length is different as DSLR lenses are made to focus an image onto 35mm film/sensor. So if you are using a chip smaller than 35mm, you have to compensate for the apparent telephoto effect.
µ
Rashiga
Posts: 18
Joined: 13 Oct 2019, 14:45

Re: Issues with aligning scans

Post by Rashiga »

Did you made a custom flange for the lens? If so did you manually calculated the required flange length?
User avatar
Micr0
Posts: 586
Joined: 15 Nov 2016, 15:20
Location: New York City

Re: Issues with aligning scans

Post by Micr0 »

Rashiga wrote: 06 Dec 2019, 18:10 Did you made a custom flange for the lens? If so did you manually calculated the required flange length?
No. Cheap Nikon to C mount from eBay.
µ
Post Reply