Kinect Outdoor

Technology analysis at daylight

Why did I do this?

Check how the Kinect performs in outdoor settings at daylight.

Make the Kinect Ourdoor-ready

The stand alone Kinect comes with a 220V AC/DC adapter with an output of 12V (1.08A). To use it outdoor on a car's 12V socket I soldered a 2nd cable to the Kinect. Because I did not want to cut the Y-Cable, I opened up the Kinect and soldered the cable direct to the PCB to supply the power. The pins you might be looking for when doing this are Pin 5 for 12V and Pin 6 or Pin 4 for GND. On the other end of the lead I soldered a standard 12V cigarette lighter car plug for convenience. The cable I used is thin enough to fit trough Kinect's motorized foot and when you gently cut the existing hole for the cable a bit bigger, you end up with a clean solution.

Kinect with Processing

To use the Kinect with Processing you can either use freenect or the OpenNi Framework. Both use different drivers and you have to use one or the other to avoid strange conflicts. Remember - a USB-Device on Windows systems can only be driven by one device driver so you have to decide which one to use. I used OpenNi, because the Framework comes with some convenient functions that extend Kinect's hardware capability quite far.

The setting

The outdoor test was done in a semi-open parking block on a bright (typical Scottish) day with very strong winds, couds, rain and bright sun. You can actually see the camera shaking from the strong winds. The camera was mounted on the rear lower edge of the opened back door of a car. The distance from pillar 6 (see image) to the camera was measured with a distance meter at about 8,86 m. For the test no actual GUI interaction was done, but several OpenNI (.oni) videos were recorded.

Result

The test in itself does not look that much spectacular in the video above, but having in mind that the Kinect was not designed for such environments, it is pretty amazing to see that the overall system was able to work quite good on such a bright sunny day.
The user can be 'tracked' with OpenNI at a maximum distance of about 4,40m. If you compare this with indoor settings (max. distance 10m), this is much less, but still a good result. When measuring the distance from the video image within the RGB Demo, the Kinect delivered exact the same distance of approx. 8,8m that was delivered by the distance meter. Even so the measuring was a bit noisy, I was quite astonished that the Kinect can deliver such a precision. It is not centimeter perfect but good enough for useful applications. All in all the test showed that Kinect is a well engineered piece of hardware that also works outside of its intended space. However, for Urban HCI applications the interaction space that is created by the Kinect is still a bit too limited, especially when you have to consider spectators or multiple users in your application design.


 

Technologies used:
Kinect (12V modified)
Processing