[point_types] Feature request: missing point type XYZIT

[rokusottervanger]

Problem
There is no point type including per-point time data.

Context
Many lidars nowadays provide per-point timestamp data. This is relevant for compensation of ego-motion etc. There is currently no PCL point type for such points.

Expected behavior
A point type PointXYZIT (and optionally PointXYZT) is available.

Current Behavior
Such a point type does not exist.

Describe the solution you'd like
I'd happily provide a PR, but I'd need some pointers on where to start.

Describe alternatives you've considered
I thought about keeping track of per-point timestamps in a separate container, but keeping the two in sync is inefficient and error prone.

[mvieth]

Hello, thanks for the suggestion. The procedure is described here, files point_types.h and point_types.hpp need to be adapted. However, PCL already has a lot of different point types, and we need to make sure that such a new point type is needed by several people. If, let's say, only you need this point type, there would be no reason to add it to PCL (you could simply define it locally in your project). I would suggest that we leave this issue open, and everyone who needs such a point type with intensity and time (or just time) can comment here. If enough people express their interest, we can further discuss adding the new type.
Can you also describe, which types intensity and time should have in your opinion? float/double/int32_t/uint32_t/...? Intensity would probably be float?

[rokusottervanger]

Thanks for the quick response and elaborate explanation.

I'm fine with leaving this issue open for a while to let it collect votes.

Since the per-point timestamps are mostly useful for correction of ego motion of a sensor, and objects moving in the scene during recording, the most relevant part of the timestamp is in the order of magnitude between the scanning period and the lidar beam firing interval. This is typically in the order of 0.01s to 0.001s for most lidars. Some lidars output these timestamps in absolute time since Epoch, meaning they would already need the full 4 bytes of a float just to represent the seconds in full precision. That's why, for the timestamp, I would suggest a double.

Intensity is already implemented in point types like XYZI, so I would stick with the same datatype for that.