Hi Omid,

actually, you don't need to calibrate anything, because it already is a calibrated 3d-scanning device. But in order to use it you need commercial software like this (see on youtube):

" Kinect Fusion Lets You Build 3D Models of Anything"

https://www.youtube.com/watch?v=of6d7C_ZWwc

" 3D Scan with Xbox Kinect and K-Scan: Beginners Tutorial"

https://www.youtube.com/watch?v=yXZRvrnQ51w

But I assume, you do not have such software. In this case you have to conduct your own calibration procedure. They rely on open source software like OpenNI, PCL, and others. I did not use a Kinect so far, so I can not tell you about technical details nor problems using it. But I can give some hints where to find more details (see following).

A good starting point for the use of a Kinect scanning device is given by Miika Santala (see respective PDF).

https://www.theseus.fi/bitstream/handle/10024/41621/Santala_Miika.pdf

A sophisticated description of the calibration procedure is given by Jan Smisek, Michal Jancosek and Tomas Pajdla from the Czech Technical University (see "3D with Kinect" on researchgate):

Chapter 3D with Kinect

Other helpful weblinks:

https://en.wikipedia.org/wiki/Kinect

https://structure.io/openni

http://www.willowgarage.com/papers/3d-here-point-cloud-library-pcl

http://wiki.ros.org/kinect_calibration/technical

http://pointclouds.org/

Regards

Olaf

Thanks Mr Peters. your answer was very useful for me.

Thanks Mr Borstell

wich kind of kinect is better to 3-D modeling ?

Hi again,

@Hagen: Yes, that's true. Kinect v1 sensor uses structured light, whereas Kinect v2 uses TOF. I mistaken them, I thought it's the other way around.

However, I did some research on the internet, in order to find some comparisons between both sensors. As far as I understand, one can not clearly say that the one is better than the other. It depends on the task, in particular on the working distance (to the target), indoor or outdoor usage, ad-hoc usage or not, and others. And it seems also to be important to take care of all influences that affect the image quality, what holds for both sensors.

Several papers show no big difference between Kinect v1 and v2 for short distances like ~1 m. Both sensors show a precision of 1-3 mm here, if carefully treated.

If the distance to the target increases both sensors show a decrease of depth precision. But Kinect v2 performs much better in comparison to Kinect v1. E.g. in [Pagliari] (see Figure 9) an accuracy assessment is shown, where at a distance of 3 m Kinect v2 has an depth error of 2 cm and Kinect v1 has an depth error of 10 cm.

Kinect v2 suffers from a warm up effect. Several papers show that one should wait 20-30 min to warm up the device before using it. In contrast the Kinect v1 sensor does not show this effect [Lachat et.al. 1], [Wasenmueller].

[Wasenmueller] also reveals a color dependency of Kinect v2 for depth measurement, but Kinect v1 does not. The same effect is shown in [Lachat et. al. 1].

[Lachat et.al. 2] show the possible outdoor usage of Kinect v2 but mention many problems with incompleteness and flying pixels. But they also state, that Kinect v1 is not suitable for outdoor applications.

As one can see, there are good reasons for the Kinect v2 but some against it.

Regards

Olaf

References:

[Lachat et.al. 1]

Conference Paper First experiences with kinect V2 sensor for close range 3D modelling

[Lachat et.al. 2]

Article Assessment and Calibration of a RGB-D Camera (Kinect v2 Sens...

[Pagliari]

Article Calibration of Kinect for Xbox One and Comparison between th...

[Wasenmueller]

Conference Paper Comparison of Kinect V1 and V2 Depth Images in Terms of Accu...

Thanks. Mr Borstell,

Your guide was helpful