Knowledge Base

Problem

EthoVision XT incorrectly detects stationary objects (e.g., plastic plants, decorations) or moving background elements (e.g., the experimenter) as the tracked subject, resulting in erroneous track data.

Details

EthoVision uses a differencing method to detect subjects by comparing the live image to a reference/background image. If the background reference image does not accurately represent all static elements in the arena (e.g., plants are not included, or a person walks through), EthoVision treats those differences as subject movement. This is a common issue in aquatic (fish) tracking with enriched environments and in open-field setups where experimenters enter the frame.

Solution

To reduce false detections from static objects (e.g., plants):
1. Recapture the background/reference image with all static arena elements (plants, decorations, substrate) in place, but with no subject present.
2. Enable 'Use Dynamic Reference Image' under Detection Reference Image options so the background model adapts over time to slow-changing elements.
3. In Detection Settings, adjust the detection threshold so that only the subject (higher contrast difference from background) is detected, filtering out low-contrast static objects.
4. Define zones to encompass only the areas of interest; detections outside the arena boundary can be excluded.

To reduce false detections from experimenter movement:
1. Use the Track Editor's Auto Select feature to identify and remove or interpolate aberrant samples — select by unexpected sample-to-sample distance to catch sudden large jumps caused by the tracker latching onto the experimenter.
2. After auto-selecting suspicious points, set them to 'missing' or use interpolation to fill gaps.
3. Where possible, remain stationary and outside the camera field of view during acquisition.
4. If experimenter position is predictable (e.g., always outside a certain X coordinate), use zone-based filtering to exclude detections in that region during post-processing.