Low-accuraty results

Hello.

I am a new user and I’m checking the possibilities of Webodm. After a few sets of data I have questions and problems. I would be grateful for your help.

The project area is relatively difficult - an open-pit mine. There are big differences in hights, high greenery etc. 5 RTK-GCP were used for the project.

  1. Dense cloud points created by WEbodm has very high noise up to 1 meter.

The same data set processed in Pix4d has a noise of about 0.05m.

Check-points are on the asphalt road. I need high-accurate data for terrain measurements.

Sample Pictures

  1. Is it possible to use the rerun function using WebODM Lightning? Can the rerun function be used for a completed task?

  2. What options should I use to generate a good point cloud and good orthophotomap?

Using Windows 10, Chrome, manual install of WebODM,

Console output: Data temporarily unavailable on Webodm Ligthning

Processing parameters:

{“uuid”:“8e1dcca4-73a9-4b33-aece-20eb7f8712e6”,“name”:“Poludnie6”,“dateCreated”:1584783506940,“processingTime”:96358439,“status”:{“code”:40},“options”:[{“name”:“pc-classify”,“value”:true},{“name”:“pc-sample”,“value”:0.01},{“name”:“mve-confidence”,“value”:0.8},{“name”:“orthophoto-resolution”,“value”:5},{“name”:“orthophoto-no-tiled”,“value”:true},{“name”:“dem-resolution”,“value”:2},{“name”:“opensfm-depthmap-min-consistent-views”,“value”:4},{“name”:“orthophoto-cutline”,“value”:true},{“name”:“pc-las”,“value”:true},{“name”:“dsm”,“value”:true},{“name”:“depthmap-resolution”,“value”:1000},{“name”:“opensfm-depthmap-method”,“value”:“BRUTE_FORCE”}],“imagesCount”:447,“progress”:100,“output”:[]}

Thanks for the help.

Unfortunately no, this would require keeping intermediate results in storage, which currently the network doesn’t keep (for scaling and cost reasons).

This might be a good starting point https://docs.opendronemap.org/tutorials.html#creating-high-quality-orthophotos but if noise is a concern, it might be worth trying to use --use-opensfm-dense and tweaking --opensfm-depthmap-min-consistent-views.

3 Likes