Hi everyone. Recently I did one work that makes the rgb data calibrate the skeleton data in 3d coordinations. For example, when I hold one small-size object inside my right palm, intuitively speaking, the physical distance between some joints (e.g., HAND_RIGHT, HANDTIP_RIGHT, THUMB_RIGHT) and the object should be zero (at least a very small value). I selected some points from the rgb image and got their 3d coorinations (x, y, z) by combining with the depth information from the depth image. After calibration, I found there is error between these points' coordinations and the aforementioned joints data. The error is near to 30 centimeters. Previously, I thought the skeleton data is extracted by the depth sensor, now I doubted that... Can anyone else give me some tips and clues? Thank you very much.
qm13
commented
2 years ago
Hi everyone. Recently I did one work that makes the rgb data calibrate the skeleton data in 3d coordinations. For example, when I hold one small-size object inside my right palm, intuitively speaking, the physical distance between some joints (e.g., HAND_RIGHT, HANDTIP_RIGHT, THUMB_RIGHT) and the object should be zero (at least a very small value). I selected some points from the rgb image and got their 3d coorinations (x, y, z) by combining with the depth information from the depth image. After calibration, I found there is error between these points' coordinations and the aforementioned joints data. The error is near to 30 centimeters. Previously, I thought the skeleton data is extracted by the depth sensor, now I doubted that... Can anyone else give me some tips and clues? Thank you very much.