How to remove false strip lines from UAV LiDAR data

Question

I have a multiple return UAV LiDAR point cloud and it has false strip lines 0.5-1 m below ground. I think two different flight lines are not overlapping perfectly causing these strip lines, but not sure. Strip lines consist of many points, therefore I was unable to remove with lasnoise.

And lasground ground classification algorithms are picking up these noise points as ground points, as a result, correct ground points classified as non-ground points. During normalization of point cloud, height of correct ground points are calculated based on the falsely calculated ground points and shifting upwards which is causing me a problem.

Ground points:

And Vegetation points:

And I created raster image from points below 1.3 m to show the problem clearly:

I have 100 1 ha plots and most of these plots have this issue. My question is how to get rid of these below ground strip lines?

Answer 1

I am working with lidar data however my data is in .csv files so easy to manipulate in QGIS. I am able to filter each file and extract any points meeting height criteria which is the same as your problem. It is easy to test with one file and there are 2 stack exchange answers with the information required to automate the process for any number of data files.

1. Open the layer properties and select source tab. Click on Query Builder and create a simple height related query.
   

2. If this works - to import a large number of .csv files use the following answer. https://stackoverflow.com/questions/53292233/errors-in-import-multiple-csv-files-into-qgis

3. To apply the query to all layers - follow the instructions in this answer - Note this will run on all active layers. I keep my layers in related groups to reduce clutter and enable easy manipulation.
   How to apply a filter to multiple vector layers in QGIS?

Answer 2

If you are using lastools to segment out the ground points, namely the lasground_new function, you can fine tune some of the parameters such that the algorithm will allow points within a larger vertical threshold to be classified as ground. Other software such as the R package lidR have similar functionality, but for lasground_new you can follow some guidance from the docs:

"The experienced user can fine-tune the algorithm further by specifing the threshold in meters at which spikes get removed. Setting ‘-spike 0.5’ will remove up-spikes above 50 centimeter and setting ‘-spike_down 2.5’ will remove down-spikes below 2.5 meters in the coarsest TIN. The maximal offset in meters up to which points above the current ground estimate get included can be set with ‘-offset 0.1’. You can also – if you want to try a lot of settings – play with ‘-bulge 1.5’ that – by default – is set to a tenth of the step size and then clamped into the range from 1.0 to 2.0."