Displaying a digital "Down Marker" on a live video feed

I was watching the Superbowl and I was wondering how the digital "Down Marker" was being displayed on the live video feed. Below is a screen show of the digital down marker which is a red line.

Some behavior that I noticed is that the line is always displayed on the "field", so if a player is standing on the line, you can see his feet as if he is standing on top of this line. If the camera moves, the line stays in a static position on the field. All cameras recording the event see the same line, even though they are in different vantage points.

How would someone go about implementing this? What se开发者_运维技巧nsors would you use? How would you correlate the data? What libraries or development platforms are ideal for this problem?

_{(source: sbnation.com)}

---

Sportvision uses a computer per-camera, and another master computer to achieve this.

The name for their product is "1st and Ten"

They use positional information from angle encoders on the camera mounts, along with inferred perspective info from markings on the field captured in the image to allow "Match Moving"

They have a prebuilt 3d model of the field that they line up to the backward-solved camera positions. That model, with input from the officials, is where the line is drawn... and then it's overlayed.

From a programming perspective, this is an excercise in:

• Video processing to extract geometry details

• Photogrammetry to turn those details into some concept of camera position in the model.

• Geometry computations to integrate the data from the camera position encoders and photogrammetric information

• Realtime User input to control the system

• 3D modeling and rendering to render the overlays

• More video processing to actually overlay the video in sync with the broadcast feed.

I guess this kind of thing is the answer to all those question "Do I really need math as a computer scientist???"....