The purpose of this page is to document experiments with the kinect in aquatic environments. I am especially interested to see whether it is possible to record 3D images underwater. It is a known fact that infrared lasers are not the preferred tool for underwater vision due to the absorption of water. Refraction provides an additional problem, especially if the boundary surface of the medium is constantly changing (water is waving).
However, in a controlled environment (swimming pool, water tank) it might be possible to record useful data; in this page I am showing my results of experiments realized with the kinect in such environments.
Observing flow patterns with the use of particles
Understanding flow patterns is one of the first steps for designing objects for underwater locomotion. In research labs expensive systems are used for tracking the movement of the 3D particles: Digital Particle Image Velocimetry (DPIV) is measurement technique with which the velocity of seeded particles can be measured. I was interested whether a low-cost, low-resolution DPIV can be achieved by using the kinect; what is the smallest particle size the 3D camera might see.
The image above shows how the granulates are used to observe flow pattern around underwater vehicles. For example, by Ariel, the wake has a reverse rotational direction which helps in thrust generation, similarly to biological fish.
Experimental set-up
I used my bathroom tube filled with water for scanning an espresso mokka under water. The recorded image was quiet promising: the shape of the mokka was quiet reasonable. Please note, that the water is shallow (30 cm high).
Afterwards, I was interested whether pepper could you be used as particles for tracking the 3 dimensional flow pattern. Unfortunately (and not surprisingly), kinect failed on being able to “see” the small sized spice.
In the following experiment I used retro reflective material in different size to understand what is the smallest piece that is beyond the kinect’s precision. I glued the material squares on a clear sheet of plastic. First, I placed it on the surface of the water and let it flow.
The retroreflective patches are not seen by the kinect, which also provides a useful information. However, when I pushed the the sheet inside the water, the patches immediately became visible for the kinect, but soon (3-5 cms below the surface) they were not outstanding anymore from the environment.
In the followings I show how a person’s body looks like in a shallow water.
As it can be seen, the shape of the body remains similar and recognizable under water, in a bath tube. This also raises the question: what is the maximum depth the kinect can see in?
Scanning in the swimming pool
The target swimming pool was the diver’s pool at NYU Coles’s gym. We did not have much freedom to play around in the space as we were quiet constrained to the electricity as well as the dry part of the area. Therefore, we only recorded at the side of the pool. As there was a lot of movement in the water, the surface of the medium was disturbed. According to the results, in this circumstances the kinect works approximately one foot deep, below that the laser beams are absorbed. The following video shows the recorded image:
My results show that with the given set-up the kinect barely can see under water, however, I believe the depth of visible landscape could be increased with different arrangement. Consider a clear, waterproof box in which the kinect is tightly fitted; placing that in the water intuitively would work better.
Even if my results were not satisfying for the purpose of underwater vision, I find the recordings of water as medium aesthetically very interesting.
A diver also volunteered to be recorded in 3D while he is jumping.