Loss of points/data packages (aka) Jumps in timestamps between frames
math-1307 opened this issue · comments
Loss of points/data packages during recordings
The recordings were performed in dual-return mode of a Livox Tele-15 LiDAR. From the device manual, the continuously rotating lasers capture point cloud from the scene. One could imagine how a few seconds of interruption in data recording could result in a sequence of missing points. This is happening between frames in the Livox Tele-15. This phenomenon is illustrated in the following images.
| Usual case | Example of missing points |
|---|---|
|
The points between consecutive frames is continuous. |
The points between consecutive frames is discontinuous. |
Naturally this is also reflected in the timestamp.
Each value in the above plot represents the difference in time (time-gap) between the last point of a specific frame and the first point of the subsequent frame. The plot displays such time-gaps between consecutive 100 frames. Three sets of time-gaps can be observed from the plot.
- The lowest set is the most frequent being the median with a value close to ~4 microseconds (4.35e-06 seconds). This is also the value between any two consecutive points in the point cloud. This means in this case no points are lost.
- The middle set is least frequent and is closer to a value of ~200 microseconds (0.0002 seconds): This means that a certain number of points is lost during this jump. Calculation of the average number of points lost correspond to 1 data package lost (from Livox-SDK-Communication Documentation, approximately 48 samples/points are lost)
- The highest set is least frequent and is closer to a value of ~400 microseconds (0.0004 seconds): This means that a more number of points is lost during this jump. Calculation of the average number of points lost correspond to 2 data packages lost (from Livox-SDK-Communication Documentation, approximately 96 samples/points are lost)
It is important to observe that the time-gaps become double during recording in single return mode.
What could be the cause for this problem? Is there a solution for not losing these points/data packages during the recording?