When you have examined your sequence you can now concentrate on the strategy of placing your key points to achieve the best results. It is important to understand the difference between the different keys and the effect they have on the tracking process before you begin placing them.
When selecting tracked points, you should choose points that:
All of the 3D calculation depends on the correct choice when placing points.
Scatter - Place the key points over the widest possible area trying to cover the 3D volume. Concentrate on areas where you want to put a 3D object.
Balance - As you move through the sequence some points may leave the frame and other points may enter the frame. Therefore try to keep a balance of the number of points within each frame and avoid a lot of points leaving or entering at the same time.
Depth - Place key points in the background and the foreground of your sequence to enable depth calculations.
Image masking - You want to track a point from image 1 to image 20. You know that in images 12 to 15 there is a partial masking of the point.
You place a Begin key in image 1 and an end key in image 20. For the images 12 to 15, you may be able to place single keys. The tracker ignores these keys and this part of the sequence. Therefore the tracker tracks from image 1 to image 12 then ‘jumps’ to image 16. However, the 3D tracker uses the single key points in the same way as other points.
Intermediate keys - You want to track a point from image 1 to 20. However, the point in image 1 and in image 20 is very blurred so you cannot place a begin key or end key. In image 10 the point is clear so you place an Intermediate key and launch a forward track to image 20 and a backward track to image 1. By doing this you succeed in tracking the point throughout the entire sequence despite the blur.
Variable Zoom - If you track a feature with an image size that changes drastically (large motion or zoom), it is better to place a key in a frame where the resolution is high and start tracking from this key towards frames where the feature appears smaller.
