Tracker Parameters

Now that it is clear what a tracker is (a "box" that will follow the object in the image that you select), it is important to understand what the parameters mean.

The first three parameters are very simple:

Width

Height

Rotation

Resolution - Example - Impact

Width: This is how wide the tracker should be (in pixels). A good value is 40, which is about the size of the cross on the white object. You can experiment different values but to start you should use 40.

Height: How tall the tracker should be. Remember that the tracker is a rectangle. The coments above apply to this paramenter also.

This is a 40 by 40 window, whereas this is a 50 by 70 window.

Rotation: This switch determines wheather or not the tracker (rectangle) should rotate if the object it is following rotates (try it out with the switch off first and then with the switch on).

Resolution: This parameter requires a more detailed explanation. The tracker knows where the object that it is tracking goes, by examining the differences between frames in the image sequence.

If the size of the tracker is 40 x 40, then the area of the tracker is 160 pixels. That means in looking for the difference between two frames in the image sequence, the tracker will examine 160 pixels (certain features are extracted from the 40 x 40 area).

The resolution parameter divides the width/height of the tracker by a certain factor. If the tracker measures 40 x 40 and the resolution is 1, VVL will examine all 160 pixels.

If the resolution is 4, the tracker "size" will be reduced by a factor of 4, so the actual values will become 10 x 10.

The size of the tracker that appears on the screen remains the same, regardless of the resolution value. What happens is that the sub-image under the tracker is reduced before it is processed internally.

This is done by averaging groups of pixels.

Resolution's Example: Let's consider the following values for an image:

10 20 30 40

This image measures 4x4. If the resolution is 2, the image size (internally) will be reduced to 2 x 2. Pixels are grouped and averaged to obtain new values and a smaller image (tracker):

10 20 30 40 becomes 15 35. The same applies to the other rows so that the original tracker will hold the following image obtained from the original:

15 35

Notice that in this example we averaged only the rows. What is important here is to understand that the image under the tracker is reduced by a factor that depends on the resolution.

What impact does the resolution have on the tracking?

As mentioned above, the resolution causes values in the rectangle to be averaged. This means the image examined will contain less information (the lower the resolution value, the more information the rectangle will hold).

As a result of this, if we have a very low resolution value (such as 1), we will have more information but the tracker will be more sensitive (if the object moves too fast, it may not do a good job following it-the rectangle might be "jumpy" or completely loose the object). It will be more sensitive but also more precise-if it is able to keep up with the moving object, it's location will be more precise. A lower resolution value also means the tracking will be slower, since more computation is done.

If the resolution value is set too high, too much information will be lost due to averaging and the tracker will not do a good job. Speed will increase with a higher resolution value but at the same time precision will decrease.

It is best to find a balance in the resolution value, but you are encouraged to try out different values and try to determine what the best value is depending on the size of the tracker.

It is also important to point out that the image of the object being tracked also has an effect on the tracking (contrast, grayscale values, patterns, etc).

Up to Tracking Contents

Back to Introduction