I worked with a student in computer sciences who tried to do something similar for me once. He was using MatLab code to do it, and had the user prompt the tracker by identifying the target at the start of recording, and also select an area of the screen where the light intensity was liable to change a lot during recording so that could be excluded from analysis. It worked passably given the moderately awful video I was getting from the setup (and the fact that the insect sometimes went and sat motionless for a monute or more). Unfortunately the experimental arena set up needed so it could all be captured on one camera caused the insects to behave so oddly I ended up having to go down an alternative route of a BASIC event recorder to manually record behaviours and movements between sections.

I guess the key is to work out how to stabilise the background and having options to specify part of the display as the arena and ignore other parts. Personally, I always struggled with dynamic backgrounds for EthoVision - too much noise - so where possible, a static background that changes very little and a good contrast between animal and background help a lot. On the software front I suppose that means you have to be careful what you're detecting and how (contrast/colour/movement/change/darker than background/etc.).

I don't know if that's helpful or already things your person is doing!

Thanks for the answers, in terms of the study design it is a classic open field so no dynamic background. Unfortunately the freeware animal tracking software is not exactly user-friendly for people who are not literate in programming languages. I will keep you updated on my progress. On face value it certainly seems much cheaper to have an image tracking program coded and customised to your project by a computer programmer than to pay the $5,000.00 or more price tag of the tracking software like ethovision.

Choosing openCV for your project seems to be a good choice for your setting. If you have already an experienced programmer doing the coding you are lucky. Nonetheless the overall costs of doing the customization for a single project are likely to exceed the 5 Tsd $ pricetag for the programmer, if you need to hire someone in Germany (50-100 hrs work). If you don't want to get to expensive with the software you might also think about continuos manual tracking using an video overlay or timelapse images for movement control. Here is one approach based on the ethovision software Brendele J., Annas E., Weirich C., Hoy S. (2009): Results of computer supported measurements of distances moved by VideoMotionTracker. Proceedings of the 41. Internationale Tagung Angewandte Ethologie, Freiburg, Germany, but the same effort can be coded with reasonable workload without using ethovison for the backgroundlayer and tracking.

Please keep us updated, how openCV managed the job.