Looking for the updated Free Gait Analysis Software post? Scroll down or click here
Attaching markers is a clever idea that started when we (the biomechanists) wanted to assess human gait but videos were too big for available computer power. We needed to track how people moved, so we put retro-reflective markers on people and tracked the markers using infra-red cameras. Now we only had to deal with a set of point co-ordinates, not big bulky videos, and the computers could cope. Tracking more than seven co-ordinates was a major technical challenge so we placed the markers directly over the joints of interest and kept the number of markers low.
Move forward twenty years and computers can cope, cameras can cope, and we have fire-wire cables. Not only can we use video but we can use high-speed and multi-camera video set ups. I used to use spherical (not flat) markers and opto-electronic cameras for a lot of film industry applications because we need data that’s fast and easy, backwards compatible and not hugely accurate. The picture above is from the set of King Arthur in 2004, and clearly I was still using markers then, but I use this pic a lot to remind me what a nightmare the markers were on that shoot! These days I’d rarely attach markers for film, research or investigation.
How can you track without markers?
Tracking without markers is a lot more accurate, as your video contains an image of the horse, not just a point. If you only track a marker that is attached to the skin of the horse, then you’re reliant on how accurately it’s placed, and you have to cope with the fact that it moves will respect to the horse’s skeleton due to skin and muscle movement. However accurately you think you can place a marker, you can’t put it over the joint’s centre of rotation, as this is a functional, changing point not a fixed physical part of the joint. The very basic models place markers over the “joint centres” and just join the dots so their outputs should be taken with a pinch of salt. Moving one marker by a single centimetre results in a massive change in joint angles, even before skin movement comes in. I don’t think anyone has used these for clinical human assessments for over a decade!
To track without markers the computational model simply fits a skeleton to the horse in the video, and works out functionally where the bones, joint centres, muscles and tendons must be. This is easy to do as we know that the bones don’t change length, that joints don’t dislocate and that skeleton moves in a predictable way. Combine these things and you get a system which can work out where a horse’s skeleton is accurately and reliably. If you are interested in the even more technical side you can have a browse through the publication list, and see joint angle accuracy of less than one degree, and even single millimetre accuracy on the joint centres in the really flash version!