What are retroreflective markers?
NextStage works by tracking circular markers made out of retroreflective material.
Retroreflective material works differently from normal reflective or shiny materials. Normally light bounces off of a surface when it’s reflected, away from the light source. But retroreflective materials reflect light back in the direction it came. The most common way it does this is by using tiny glass beads embedded within fabric.
The Microsoft Kinect has a built in infrared camera, and an infrared emitter. What happens when the Kinect is facing a retroreflective marker, is that the light from the infrared emitter bounces off the marker back in the direction of the Kinect, and right into the infrared camera.
While the infrared emitter does illuminate the entire scene, these markers appear as exceptionally bright points, making them very easy for NextStage to identify.
Getting and making markers.
There are a number of ways to get or create these markers. The best solution I have found is using Reflecmedia’s chromatte tape. Reflecmedia makes a really cool alternative to green screens using retroreflective cloth. They also sell a 2” thick roll of retroreflective tape. All you need to do is cut the tape into 1” circles, peel off the back and place the markers.
Now if you go this route you will probably be buying the most expensive roll of tape you will ever purchase. A single roll is about $50 for 6’. But there are a lot of benefits to going this “premium” route. This material is most reflective I have found, it works from as far back as the Kinect can go, and has a nearly 180 degree viewing angle. It’s also the most professional and discreet looking. All you see are these small grey dots which should be very easy to remove in post if needed.
The second best option is to use the retroreflective cloth found in safety equipment. A lot of safety clothing and roadway equipment uses retroreflective material since it’s very easy for drivers at night to spot the light reflected back from their headlights.
Now unlike the chromatte tape this likely will not come with an adhesive material. One option is to take the material, cut it into circles, and use double sided tape on the back. Another option is to take a strip of it and a piece of thick paper tape. Cut a circle into the paper tape that has a diameter the same as the width of the retroreflective strip. Place the strip on the adhesive side of the tape, so that the retroreflective side is visible through the hole on the other side of the tape. This way the tape shapes the material into a circle, and can easily be placed and replaced. This is not the prettiest solution, but it is easy to work with and quickly set up.
Finally you can purchase retroreflective stickers that are pre-cut into circles and other shapes. These are actually very cheap, and are even available in different colors in case you want to use them on a green screen.
The problem is that these are nowhere near as reflective as the chromatte tape or safety equipment. The viewing angle is a lot narrower so NextStage will have trouble tracking some of these. Also the green colored ones are very dark, much darker than traditional green screen cloth. So in a green screen setup you would probably end up rotoscoping them out anyway.
Ideally you want your markers to be 1” or more in diameter, placed randomly on a flat wall with 6”-12” inches between each marker.
When placing markers it is essential that they are placed on a flat, diffuse, non-reflective surface with nothing else in a 3”-5” radius around it. NextStage is not able to retrieve depth information from the markers themselves, so it needs to reconstruct the depth from the surrounding pixels. If the surface around the marker is uneven, or too reflective it will return incorrect depth data and throw off the track.
Otherwise you can place markers wherever you want as long as they do not move and you avoid repeating patterns. The more randomly placed the markers are, the better the track will be.
NextStage uses the different patterns made by the markers to identify them. If patterns repeat or are too similar, NextStage may lose the track. If everything was placed in an even grid NextStage would still be able to track the markers, but it would have a hard time telling which markers are which.
While NextStage can accurately track position and rotation with just three markers, adding more markers is the easiest way to increase the precision and stability of the track.
With 3 markers it is essentially only 1 “sample” and it’s not the most precise result. With four markers there are 3 triads within the markers and NextStage is able to track using 3 “samples”, and with a dozen markers NextStage would have 78 samples for each frame. This is a huge increase in precision compared to just three markers.
You also want to have multiple markers spread out in a scene so you can move the camera around. While NextStage can track with three markers, it needs to see those markers every frame. If one marker goes out of frame, or an actor temporarily obscures one of them, the tracking will freeze until it sees at least three markers again. The more markers you have, the less likely it is that tracking will be interrupted.
Spreading out the markers within the Kinect’s field of view also helps increase precision and reduce jitter. Having just four markers, one at each corner of the Kinect’s infrared field of view will produced a better result than if they were all bundled together in one corner of the frame.
The final method of increasing precision is to place markers on different planes. While you could place all of your markers on a single flat wall, this would essentially be tracking the two axes along the wall, up and down, left and right.
This is fine if you just want to create a small virtual room or place a 3D object onto live action footage. But virtual sets and objects far past the wall will jitter, even if the Kinect is completely still.
While NextStage can extrapolate the third Axis (the one running back and forward perpendicular to the wall) the more information you can give NextStage the better. Placing even one marker a foot or more away from the wall is crucial for tracking large virtual sets and faraway objects. This gives NextStage enough information to precisely track all three axes.