Lidar support

I was curious if anyone knew if I can feed lidar data into webodm to stitch cloud point data?

2 Likes

Not directly and not with WebODM. Using ODM you could run the pipeline until the odm_filterpoints step (--end-with odm_filterpoints), then replace the point cloud in dataset/odm_filterpoints with one generated with LIDAR (it should already be geographically aligned and vertices need to be offset by odm_georeferencing/coords.txt (substract from the second line on the coords.txt file every point’s X/Y coordinate), then resume the pipeline with --rerun-from odm_meshing. The LIDAR point cloud will be used for orthorectification, DSM generation, etc.

It’s an interesting use case, this wouldn’t have been possible until a few weeks ago when we switched to a geographic CRS for the point cloud, we could look to add this as a fully automated feature (interested in contributing a PR?)

2 Likes

Also, do you have a LIDAR dataset you could share with associated aerial images? I’ve heard this feature request often, but never got any data to try it out myself.

3 Likes

Yes, I have a direct use case that I can use to contribute dataset. It is a .las file.

3 Likes

We have a data set for you :slight_smile:

Photos - Synology Drive

LAS - Synology Drive

The photos folder there are singles shot at both nadir and oblique as well as a zip file with them already stitched into a GEOTiff Ortho.

4 Likes

I’ve started looking into this dataset btw. Exactly what I was looking for :clap:

Can you share some insights on how the Lidar point cloud is georeferenced? The trickiest part of getting a hybrid mechanism to work is alignment between the photogrammetry results and those obtained from Lidar.

Edit: PDAL to the rescue: filters.cpd — pdal.io

2 Likes

Haha! I was about to suggest PDAL.

2 Likes

Also, wondering if you could share information about the methodology that was used to capture both images and the Lidar scan. The two seem to have been captured on different timeframes? Certain items such as cars are missing in the Lidar dataset.

E.g.

image

vs

image

2 Likes

So both the photos and LiDAR are pulling via ntrip for the corrections into each separately on the drone. Both should be WGS84. I would need to go back and look at the data but they may be from different days. That could be the issue.

3 Likes

Thanks for the details!

It would be super-cool to have a dataset that has been captured within a short time span, or even at the same time. Not having a consistent elevation model would lead to some artifacts (although minor ones) in the orthorectification process.

2 Likes

Should be flying some this weekend.

4 Likes

Looking forward to it!!!

2 Likes