Different spatial angle between strongest echo and last echo

[OuYaozhong]

Hi there,

I am a user of a dataset whose point cloud is collected by Velodyne, with dual echoes are provided.

And there is an strange result occur at that dataset.

In brief, that is the spatial angles (include azimuth and the pitch) are different between strongest and last echo.

Even that the data is raw (before calibrated by the ego motion), this problem is still existed.

The author confirmed that problem but did not know the cause. ref

If tried to match them in a way that minimizes the angle difference between the points in space, with the limitation (same beam index and the norm of last should be >= the one of strongest), it seems that though make the result more practical, but the result still strange, since in many frames of data, the angle difference can up to 1.0 ~2.0 degrees after matching by the way above.

Noted that I also use np.isclose to consider the norm that is < in float but close each other, to avoid the error caused by the floating point.

Normally, the azimuth resolution (e.g., 10 Hz) can be around 0.17 degrees depending on the manual, and the number of point that angle difference > 0. can be up to 10k / 50k.

The angle difference is up to 10 times of the one the resolution.

It seems strange and confusing.

I have gone through the manual of VLP-16 / 32, HDL-32 / 64, and the the previous issues of this GitHub repo, and found nothing about this.

And in the manual, it seems that when the driver sent back the data from sensor to the host, each data block can contain all (e.g., 32 beams) data with shared azimuth angle in the beginning of the data block, seems this ensured that the dual echo and the beams will have the same azimuth angle, if they are captured in the same time (<= 0.17 deg).

Could you provide some insight about the problem ?

Best Regards