Knowledge Base

Problem

EthoVision XT struggles to reliably detect and track a small animal (e.g., an insect) when it moves onto a natural or patterned background element (e.g., a green leaf) that is visually similar to the subject. Tracking is generally acceptable in open areas but fails when the subject is on the patterned surface.

Details

When the subject's color or brightness closely matches the background element (e.g., a green insect on a green leaf, or a pale insect on light substrate), EthoVision's pixel-difference or gray-value based detection cannot reliably distinguish the subject. This is a fundamental contrast limitation. Partial improvement can be achieved through detection setting optimization, but complete reliable detection may not be possible in very low-contrast situations. Affects all EthoVision XT versions using standard detection.

Solution

1. Provide Noldus Support with sample video clips showing the tracking failure — support can review the video and suggest optimized detection settings specific to your setup.
2. In EthoVision XT, experiment with the following detection settings:
a. Adjust the detection threshold (lower or raise depending on whether the subject is lighter or darker than the background).
b. Switch between 'Lighter than background,' 'Darker than background,' and 'Different from background' detection modes to find which provides the best subject isolation.
c. Adjust the minimum and maximum subject size settings to reduce false detections from the leaf texture.
3. Consider modifying the physical setup to improve contrast:
a. Use colored lighting (e.g., a colored filter that makes the subject stand out from the leaf color).
b. Add a visual marker (e.g., a small colored dot) to the subject if the experimental protocol allows.
c. Replace natural leaf material with an artificial substrate that has lower visual similarity to the subject.
4. Accept that in cases of very low contrast between subject and background element, some tracking loss when the subject is on that element may be unavoidable with standard EthoVision detection. Consider supplementing with manual scoring for those periods.