Calculation method

Introduction :

The raw position data collected by Ethotrack (as by all video tracking systems) suffers from noise problems and artefacts inherent to the acquisition method. This noise and artifacts are due to several reasons, including:

A video camera is made up of a finite number of pixels. The position of the animal is therefore of a discrete nature. Two distinct successive positions cannot be less than one pixel wide apart.

The data calculated from the animal's position (distance travelled, speed, etc.) are calculated from the animal's center of gravity. A modification of the animal's posture can result in a displacement of the center of gravity of a few pixels, without this corresponding to a real displacement of the animal.

Even if the error between 2 images can be considered low, or even negligible, the sum of all these errors quickly becomes significant when calculating certain data, such as the total distance traveled. This is why it is necessary to "smooth" the position data before using it for calculations.

Ethotrack offers several configurable smoothing methods in the experiment options dialog . These methods are applied in order:

First step - Minimum distance taken into account:

This step consists in deleting the points considered too close to the previous point.

Consider the following example which represents the position of the animal (raw data) on frames 1 to 7:

Image 1: X = 2, Y = 1
Image 2: X = 4, Y = 2 Image 3: X = 4, Y = 4
Image 4: X = 2, Y = 1 Image 5: X = 4, Y = 2 Image 6: X = 4, Y = 4Picture 7: X = 2, Y = 1

In our example, if the minimum distance taken into account is 3 pixels, points 2, 5 and 6 will be "deleted" because they are deemed too close to the previous point. The trace obtained will be as follows:

Image 1 and 2: X = 2, Y = 1

Image 3: X = 4, Y = 4 Image 4, 5 and 6: X = 2, Y = 1

Image 7: X = 2, Y = 1

This smoothing step has priority and influences all the results of the experiment :

Immobility of the animal.

Position within an area.

Distance traveled, speed.

etc

Step Two - Smoothing Algorithm:

For this step, the animal's position data (hitherto in pixels) is first converted to actual distances (in meters) from the upper left corner of the image, before applying a smoothing algorithm .

Several algorithms are available:

No smoothing.

"Lowess" (strong) method.

"Lowess" method (light). ( recommended algorithm)

Weighted moving average.

Weighted Moving Average (Stavisky-Golay)

This smoothing step influences all the calculated physical quantities :

Distance traveled.

Speed.

etc