Calibration problems

[Francis Leach]

Scanning Friends. 

I am unable to calibrate the kinect. I follow the guidance of the "usage file" implicitly, when asked to grab a screen  by pressing 'n' , the left hand screen freezes for a second or two but nothing appears in the right hand screen. Several Grabs are taken with the same result and when the window is closed, the console reports 0 Grabs.   

The realtime scanning system with the transmitter uncovered appears to work fine but I am not confident with the accuracy since the calibration appears to not function.  Any advice?

[MagWeb]

Hmm, I can not reproduce your problem. Here calibration works fine. Could you give us a screenshot of the (left) IR-sensor image?

Generally only successfully detected grabs are reported. Only a successfull grab is shown on the right hand screen with the colored distortion pattern.

Maybe:

- You did not cover the IR emitter?

- The IR sensor's contrast is to low/high? (Try to add/reduce ambient light sources emitting also wave lengths in the IR spectrum - as standard light bulbs)?

- The IR sensor does not see the whole pattern (your cam is to close)?

- The pattern is to small in the image?

[Francis Leach]

MagWeb

You good name goes before you.  Thanks for helping.

• I do have the emitter covered. 
  
• I suspect the contrast is to low.
  
• The board is illuminated only with fluorescent lamps
  
• Be assured, all the attempted calibration images of the board, did include the whole board, filling the view as best I could.
  

I note that the gain can be increased using the --Alpha arg parameter, the default is at 1. 

What is the correct syntax of the command ?. ( --alpha arg (=2)  ) and once changed does it stay set, as I could not see any note of it in the default config.  Would it have to be reset to 2 everytime for realtime scanning?

Can the kinect gain be increased within the sensor?

regards

[MagWeb]

Syntax to use alpha with value 2 for calibration:

reconstructme.exe --calibrate --square_size 33.875 --alpha 2

I would try to use another light source to illuminate . I´m using a single 100W filament lamp

[Francis Leach]

MagWeb

Thanks for the consist reply and no abuse which is nice.

Alas the alpha change made matters worse and simply washed out the image completely.

I now suspect the kinect sensor. see attached infra red image. It lacks any depth of Black. I have done further test in a completely blacked out space with only a 60w lamp and the results were as bad.

Moral of the story, buy nothing from ebay!!

Thanks

[Christoph Heindl]

Could you post a screenshot pointing at the chessboard pattern? The left images doesn't look that bad.

[Francis Leach]

Christoph

Firstly may I thank you Guys for a superb system

Please see attached the calibration chart viewed at several Alpha values :- 1, 2, 3, 6 & 9. There is very little contrast in any of the screen images.

I get a lot of noise on the screen. I am using a extended cable or is that just another problem.

Regards

[Christoph Heindl]

Am Montag, 30. April 2012 15:41:57 UTC+2 schrieb Francis Leach:

Christoph

Firstly may I thank you Guys for a superb system

Glad you like it.

 

Please see attached the calibration chart viewed at several Alpha values :- 1, 2, 3, 6 & 9. There is very little contrast in any of the screen images.

True.

 

I get a lot of noise on the screen. I am using a extended cable or is that just another problem.

We don't have good experiences with extension cables (just try without), but I think your problem stems from a different source. Is there a direct sunlight influence or similar?

[Francis Leach]

Christoph

I have tried the calibration many times now. Originally under fluorescent at night without the possibility of natural light and it make little difference. I have tried a calibration in complete darkness but then had to use the 60W light to see what I was doing.

Please tell me how do you guys do the calibration, tungsten only, zero natural light, what alpha?

By the way the extension cable I am using is that fitted to the poeer supply, nothing longer that 500mm.

[Christoph Heindl]

Am Montag, 30. April 2012 16:32:14 UTC+2 schrieb Francis Leach:

Christoph

I have tried the calibration many times now. Originally under fluorescent at night without the possibility of natural light and it make little difference. I have tried a calibration in complete darkness but then had to use the 60W light to see what I was doing.

Please tell me how do you guys do the calibration, tungsten only, zero natural light, what alpha?

Strange, I use the calibration without special alpha tuning under normal daylight conditions without problems. What sensor are you using btw?

[Francis Leach]

Hi

It is the Kinect for xbox 360. I get a feeling the sensor is at fault but need to confirm it before I send it back to the ebay vendor

[Christoph Heindl]

Am Montag, 30. April 2012 17:30:59 UTC+2 schrieb Francis Leach:

Hi

It is the Kinect for xbox 360. I get a feeling the sensor is at fault but need to confirm it before I send it back to the ebay vendor

You could try some other tools. For example, try RGBDemo.

[Francis Leach]

Christoph

In a light hearted sort of way may I say:- 

There appears to be have occurred a serious case of mistaken identity. In offering me the opportunity of using the RGBDemo you must have assumed I know what I am doing, That is clearly someone else.

The RGBdemo could have been written in Arabic or ancient Nordic, as it is surely not for mere mortals or generally just  inquisitive parties. Please be advised and do not be offended, but program/systemems development engineers have an unfortunate view of life in that they assume everyone else understands them.

I am generally  task managed.

I can read, I have a degree in mechanical engineering.

I can follow a simple list of instructions.

Anything else and I lose the will to live.

You tried very hard with me but I have failed 

Please tell me why my Kinect xbox 360 sensor has poor contrast when using the calibration system.  If you do not know, I believe I will have no chance.

My last attempt using the calibration system just now, was under natural light, without any sunshine, without fluorescent light, without tungsten light and still the system would not grab an image.

Regards

[Francis Leach]

Christoph

Hold the presses , breaking news.

Check out the attached file. It is a screen grab of the calibration chart in 60W tungsten, with my mobile phone in shot. Note the contrast with alpha setting 1, much better on black cased phone.  I am lead to believe the problem is the chart itself in that it was inkjet printed and maybe the ink pigment in not opaque to infra red.  Are you charts printed on laser printers?   I will try a laser printed A3 chart.  It may be some thing you can verify from your companies end.

Regards

Francis

[Francis Leach]

Christoph

Eureka!

The calibration charts is best printed on a laser printer with pigments opaque to infra red.   

Also charts printed by offset lithography would have to be checked before using any.

see attached screen grab with Inkjet printed chart, phone and laser printed chart in same the lighting conditions, 60W tungsten bulb only. the A3 inkjet printer was a Brother, the A4 laser printer a Samsung. 

My apologises to the anonymous ebay vendor who sold me the Kinect x box 360, I was convinced was a W**ker. He's not. 

Goodnight all

[MagWeb]

Argh, I should have considered that.

Visible-light-black does not necessarily means IR-black as well.

(I found out something similar using sunglasses to dim down the IR beams to improve the nearby result. Some gasses do the job, others fail.)

I´m pretty sure you are on the right track.

My printouts came from a laser printer. Seems we need a compatibility martix for printers as well ;-)

Thanks for the insight.

[Christoph Heindl]

Am Montag, 30. April 2012 22:26:25 UTC+2 schrieb Francis Leach:

Christoph

Eureka!

The calibration charts is best printed on a laser printer with pigments opaque to infra red.   

This is great news! I haven't considered the ink to be the problem. I've updated the docs and included a reference to this thread. Thanks for your patience and investigation!

Best,

Christoph