Tweet
Hello,
In our current lab set-up we are using 8 Osprey cameras set on a 180 degree space around the front of the subject at various heights in order to track 2 mm markers placed on the phalanxes of index and middle fingers, hand, and wrist. The hand lays supine on a table and the subject performs two tasks: a hand open (MCP, PIP, DIP extension) and a pinch (MCP flexion; PIP, DIP extension).
After data collection we proceed to analysis and find that two issues happen very often:
1) ghost markers appear very close to real markers and cause the tracking to shift the position of the marker from the real one to the ghost.
2) sometimes two markers that are close to each other start shaking.
Our objective is to track angular motion, so you can imagine how 'pretty' it looks the graph later when the start/end points of those vectors are dancing all over the place. The usual manual tracking of the data helps, but I am much more concerned that the data is overall affected negatively by this because of the amount of linear/cubic joining that needs to be done.
I've tried to adjust the configurations on Cortex to compensate for marker size (2 mm) and prediction error (20 mm/frame). The whole process happens over the span of 10 seconds, during which the markers move for 5 seconds. The largest motion (of the distal phalanx marker) is in the order of 80-100 mm... to no avail. Data looked exactly the same after the adjustments.
I have thought about a few of the reasons why this could be happening:
1) Too many cameras. Cameras are creating doubles of the markers when seeing them from both sides.
2) Camera resolution (640x480) is too low. The camera that is farthest away from the markers sits at a high point, about 4 ft from the hand. The cameras are seeing two markers on the same pixel.
3) Markers are too close when the joints achieve extension. There is a separation of ~5 mm between joint markers from opposing phalanxes (e.g. MP distal/DP proximal). The cameras are seeing both markers as one.
These seem to me like sound hypotheses, but there are a few things that do not seem to add up. For example, markers that get too close to each other sometimes shake and have a ghost marker too. This would only make sense if 1, 2 and/or 3 are happening at the same time.
I am considering changing the camera set-up to have 4 cameras positioned each at a 45 deg. angle from the sagittal plane, two higher up, two higher down, so as to create a pyramidal focus on the hand and fingers. I believe this could eliminate the ghost markers, but I do not have extensive experience with this equipment yet (Motion Analysis).
Any thoughts on this?
Thanks!
In our current lab set-up we are using 8 Osprey cameras set on a 180 degree space around the front of the subject at various heights in order to track 2 mm markers placed on the phalanxes of index and middle fingers, hand, and wrist. The hand lays supine on a table and the subject performs two tasks: a hand open (MCP, PIP, DIP extension) and a pinch (MCP flexion; PIP, DIP extension).
After data collection we proceed to analysis and find that two issues happen very often:
1) ghost markers appear very close to real markers and cause the tracking to shift the position of the marker from the real one to the ghost.
2) sometimes two markers that are close to each other start shaking.
Our objective is to track angular motion, so you can imagine how 'pretty' it looks the graph later when the start/end points of those vectors are dancing all over the place. The usual manual tracking of the data helps, but I am much more concerned that the data is overall affected negatively by this because of the amount of linear/cubic joining that needs to be done.
I've tried to adjust the configurations on Cortex to compensate for marker size (2 mm) and prediction error (20 mm/frame). The whole process happens over the span of 10 seconds, during which the markers move for 5 seconds. The largest motion (of the distal phalanx marker) is in the order of 80-100 mm... to no avail. Data looked exactly the same after the adjustments.
I have thought about a few of the reasons why this could be happening:
1) Too many cameras. Cameras are creating doubles of the markers when seeing them from both sides.
2) Camera resolution (640x480) is too low. The camera that is farthest away from the markers sits at a high point, about 4 ft from the hand. The cameras are seeing two markers on the same pixel.
3) Markers are too close when the joints achieve extension. There is a separation of ~5 mm between joint markers from opposing phalanxes (e.g. MP distal/DP proximal). The cameras are seeing both markers as one.
These seem to me like sound hypotheses, but there are a few things that do not seem to add up. For example, markers that get too close to each other sometimes shake and have a ghost marker too. This would only make sense if 1, 2 and/or 3 are happening at the same time.
I am considering changing the camera set-up to have 4 cameras positioned each at a 45 deg. angle from the sagittal plane, two higher up, two higher down, so as to create a pyramidal focus on the hand and fingers. I believe this could eliminate the ghost markers, but I do not have extensive experience with this equipment yet (Motion Analysis).
Any thoughts on this?
Thanks!
Comment