Detection settings

The Ethotrack detection module has several algorithms in order to allow a good detection of the animal whatever the conditions of the experiment. Automatic mode (default mode) does not require any adjustment and gives good results in most cases . If the detection conditions are more difficult (low contrast between the animal and the background of the image, variation in brightness, etc.) and if the automatic mode does not provide a sufficient detection rate, you can use the one of the following 5 manual modes in order to find the best settings adapted to the situation:

Grayscale : In this mode the original image is converted to grayscale. All pixels whose value falls within a given interval are detected.
This mode gives very good results when the image background is uniform and the contrast between the animal and the image background is high.

Image Subtraction : This mode uses a black and white reference image of the experimental area recorded in the absence of any animals.
During the test, the current image is transformed into black and white and compared to the reference image. The differences between the 2 images indicate the position of the animal.
This method is effective when the recording conditions (and in particular the light conditions) do not vary during the experiment.

Image Subtraction (advanced) : This mode uses a color reference image of the experiment area recorded in the absence of any animals.
During the test, the comparison of the current image with the reference image can be made on a single color layer (red, green or blue) or on the combination of the 3 color layers.
This method is effective when the arena background is in color and there is low contrast between the background and the animal.

Color (simplified) : In this mode the color information of the image is analyzed. This mode is called: "simplified" because it only requires setting a tolerance value around the reference color.

Color (advanced) : In this mode the color information of the image is analyzed. The representation of the colors of the image uses the model TSV ( Hue Saturation Value ), also called HSV ( Hue Saturation Value ) or HSB ( Hue Saturation Brightness ) in English. In this mode it is possible to fine tune the detection range for each color component. This mode allows finer adjustment than the "Color (simplified)" mode for a user who is familiar with the HSV color management model.

In detection parameter settings mode, each detected area of the image is clearly marked with a color and a bounding box. By default, detected areas are indicated as follows :


This area is detected, but cannot correspond to an animal because of its shape or size.	This area could correspond to an animal, but it's not the only one. There is another area in the arena that could also correspond to an animal.	This area can correspond to an animal and it is the only one.

Note : the appearance of detected areas can be configures using the display panel.

In order to allow a precise adjustment, each modification of a parameter is immediately taken into account and its effects visible on the current image of the video :

Several zones are detected, and at least two meet the size criteria : detection is poor.
To improve detection, you can modify the detection threshold or reduce the size filter range.

Several zones are detected, but only one matches the size criterion. The animal's tail is detected : the detection is correct, but can be improved.
To improve detection, it is recommended to adjust the detection threshold.

Only one area is detected and it meets the size criteria. The animal is fully detected : the detection is good.

The detection parameters must be adjusted so that the detected area corresponds as closely as possible to the surface area of the animal :

Incorrect setting.

Threshold too low : the animal is not fully detected.

Incorrect setting.

Threshold too high : the shadow around the animal is detected.

Correct setting.

When adjusting detection settings, it's best to choose an image where the animal isn't on the edge of the arena. The edge can cast unwanted shadows that make detection more difficult. The animal might also climb onto the edge and appear smaller.

Once the detection settings seem correct, it's necessary to check them on several images, and then on the entire video.

Regardless of the detection mode chosen, filtering according to the size of the animal improves detection.

Description of the "Detection Settings" side panel :

Mode :

Selection of the algorithm used for detection:

Automatic

Grayscale

Image Subtraction

Image subtraction (advanced)

Color (simplified )

Color ( Full )

	Mode settings :
	Central part which varies according to the choice of the detection algorithm.

	Filter by animal size :
	Choice of filtering parameters according to the size of the animal. For more information, see the section: Filtering according to the size of the animal.

Analyze :

When the adjustment looks good on an image, click this button to run a full analysis and check that the animal detection is correct on the whole video.

Once the analysis is finished, you can choose to keep these results, or to return to the previous analysis.

Click "Yes" to save the detection settings and analysis results.

Click "No" to keep the results of the previous analysis.

	Import :
	Click this button if you want to import detection settings from another test.

	Ok / Cancel :
	Use one of these buttons to close the detection settings modification panel. "Ok" to save the new detection settings or "Cancel" to return to the previous configuration.

To open this panel:

Click the button in the "Detection Settings" area of the test information side panel