One problem: when the ball is right on top of the "100" sign, its colour is clearly more red than the sign itself. Which means the distance detected at the ball is much closer than 100, which means that there is a time walk or another source of ToF error. Plus this highlights how overlapping the depth map on top of the intensity map gives false "great results". An actual 3D map made solely by the detected distance would have shown this error instead of covering it up with "a good picture"
i cant decide if the image series gives depth or z-length in the color encoding. depth would be a spherical shape in the true 3D space whilst z-length is a parallel plane to the imager plane. further as in any lens system there might be lens effects like pincusion but there might as well be undistortion operations applied - cant decide either.
i am wondering a bit why the first and the second setup created different depth images. the rainbow patterns in the second look somewhat imbalanced or tilted. further there is a dark blue spot on the rear wall for the second case - roughly in the area where the visible light halogen lamp hits it.
the feasibility is demonstrated but the limitations as well - and they come without explanation from the vendor that would make us understanding these limitations.
One problem: when the ball is right on top of the "100" sign, its colour is clearly more red than the sign itself. Which means the distance detected at the ball is much closer than 100, which means that there is a time walk or another source of ToF error.
ReplyDeletePlus this highlights how overlapping the depth map on top of the intensity map gives false "great results". An actual 3D map made solely by the detected distance would have shown this error instead of covering it up with "a good picture"
i cant decide if the image series gives depth or z-length in the color encoding. depth would be a spherical shape in the true 3D space whilst z-length is a parallel plane to the imager plane. further as in any lens system there might be lens effects like pincusion but there might as well be undistortion operations applied - cant decide either.
ReplyDeletei am wondering a bit why the first and the second setup created different depth images. the rainbow patterns in the second look somewhat imbalanced or tilted. further there is a dark blue spot on the rear wall for the second case - roughly in the area where the visible light halogen lamp hits it.
the feasibility is demonstrated but the limitations as well - and they come without explanation from the vendor that would make us understanding these limitations.