L515 global depth and image scan time - global latency, hence error between different pixels for a moving object
Dear all
I am studying an application where I would have to scan objects of variable size moving at speed of 500mm/sec (about 20 inches/sec) on a belt scanned from above with a L515.
Please consider I will have to take an approximate measure of all three dimensions of the moving object.
As the single pixel is declared to be scanned in less than 100nsec but I am not finding the detailed info of the total amount of time needed to acquire the full depth (and possibly also RGB) image, so I am not sure how much jitter error I will have in the XY position of pixels of the moving bigger objects. Shall I assume as a pessimistic estimate 30 fps so about 33 msec for a full scan, hence 500mm/sec*33msec= 16.5 mm max error due to scan time?
Or is the acquisition time much lower, may be also because of interlacing, redundancy and other stuff?
Thank you in advance
Regards
-
Official comment
Hello Alessio,
For RGB: If the frame rate is 30fps then the maximum time of image acquisition is 33ms per frame, but could be much lower if the exposure happens very fast. If there is a lot of light and exposure is fast then acquisition happens quickly and there is blanking time before the next frame. For example, the image could be captured in 10ms and then there would be a 23ms wait time before the next frame is captured. Be aware that the color camera is rolling shutter so that can create artifacts for fast-moving objects.
For Depth: The depth image is built from a scan of the scene. Each depth point is captured very fast and the depth has no motion blur. Because this is a scanning system it can produce rolling shutter-like artifacts. The attached video shows the depth output of a fan in motion. The individual fan blades are clearly seen. Therefore, motion blur will not be an issue, but the rolling shutter-like effect may “stretch” or “squeeze” the object slightly. I believe your assessment is correct about a worst-case scenario: "a pessimistic estimate 30 fps so about 33 msec for a full scan, hence 500mm/sec*33msec= 16.5 mm max error due to scan time"
Sincerely,
Jesus G.
Intel Customer SupportComment actions
Please sign in to leave a comment.
Comments
2 comments