smooth noisy multi-laser surfaces

[samz...@gmail.com]

Hi Team,

We are looking at automated solutions for reducing the noise in noisy pointcloud data.  The noisy pointcloud data is typically from multi-laser scanning devices such as the velodyne systems.  Noise reduction would improve the outputs from these devices.

There are a few workflows that seem suitable in other software, for example in Terrascan.  We are interested to hear suggestions of existing automated tools for this task.

We think a simple yet powerful and automated method could be achieved by subsampling excessively dense pointclouds with Lasvoxel (or possibly adding a voxel based thinning option within lasthin).  We can get excessively dense clouds with low capture speeds such as handheld scanning and some UAV scanning. Currently lasvoxel provides us statistics for each voxel, plus it can ‘compute mean xyz’.

We think there are a few ways lasvoxel could be developed to output a smooth subsampled pointcloud;

Suggestion 1: Lasvoxel to compute mean xyz for each voxel plus also;

      1. Populate intensity field of the output point with mean intensity

         of all input points in the voxel

      2. Populate RGB field of the output with mean RGB of all input

         points in the voxel

      3. Switch the synthetic flag to 1 to flag the point as a

         synthetically generated point

      4. Populate the number of contributing points into the user field

         to keep a record for further filtering later

      5. GPS time could be a mean of the input points if the range of the

         input points is insignificant (i.e. all within 0.1 sec or so to

         signify all input points come from the same ‘flightline’). If

         the time of the input points vary significantly then the GPS

         time could be set to 0.

      6. Point_source_ID could be set to zero.

 

Suggestion 2: find the median point of each voxel and flag it.  Perhaps this method works best within lasthin, as it is a cubic subsampling of the input pointcloud;

  1. Lasvoxel flags the central most point of each voxel.  The

     ‘key-point’ flag could be set to 1.

  2. Lasvoxel populates the point count into the user field for all

     output points (to help with filtering later)

  3. Lasvoxel could output either;

      1. All input points

      2. Flagged key-points only

Separate control of the xy step size and the z step size may prove useful for either option.

The scenarios where we would use these are any time that our raw point cloud density is several times more than necessary, and the surface noise is similar to the desired density.  For example a pointlcoud surface with 20-30mm surface noise and typically <10mm point spacing.

We can provide sample scan data to use as test data seperately.


Kind Regards

Sam

[Jung-kuan Liu]

Hi Sam,

I am interested in this noisy point clouds smoothing topic. My question is how can we detect that the point clouds are noisy? Will it rely on visually checking or we can run applicable indexes to figure it out?

Thanks,

--
Download LAStools at
https://rapidlasso.de
Manage your settings at
https://groups.google.com/g/lastools/membership
---
You received this message because you are subscribed to the Google Groups "LAStools - efficient tools for LiDAR processing" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lastools+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/lastools/c362949f-1082-4646-9e79-fb6db7f94327n%40googlegroups.com.

--

Ernie

[Nelson Mattie]

Dear Sam,

 
Based on my experience with Velodyne multi-beam laser head-based systems, controlling certain aspects is crucial for minimizing noise production. These include:

• GNSS signal
• External calibration parameters such as boresight and lever arm
• Processing interval and time range
• Selection of channels during exporting
• Implementation of laser angle filters
• Management of returns.

As you know, the Velodyne multi-beam scanner employs a rotating mechanism to emit laser beams from different angles, ranging from 16 to 128 beams. To identify which beam corresponds to which return, the "user data" field and the "point_source ID" field of each point are used by the exporting software. However, the LAS format does not have a dedicated field for this information, and the "user data" field is not a reliable place to store important information as it can be overwritten by temporary data. To address this, Martin recommended to use LAS format's "extra bytes" concept to store laser beam IDs.

Rapidlasso has implemented this feature in their LAStools software, Martin recommended using two las2las commands to copy the beam index from the "user data" field to an unsigned 8-bit integer "extra byte". This enables the selection of specific laser beams for further processing, reducing potential noise and density.

las2las ^
-i project\*.laz ^
-add_attribute 1 "laser beam ID" "which beam ranged this return" ^
-odir project\00_raw_temp -olaz

las2las ^
-i project\00_raw_temp\*.laz ^
-copy_user_data_into_attribute 0 ^
-set_user_data 0 ^
-set_point_source 0 ^
-odir project\00_raw_ready -olaz


In addition to beam selection, correction (boresight re-calibration and drifts) using Bayesmap is recommended. To do this, channels must be provided as separate files, and las2las can be used for channel splitting.
Overall, controlling the aforementioned factors can significantly smooth the point cloud and improve the accuracy and quality of data collected using Velodyne multi-beam laser head-based systems. 

I hope this can help!, regards. 

Nelson Mattie, PhD© . Remote Sensing 

LiDAR Latinoamerica, LLC | LiDAR everywhere

P: +1 5879361278

E: nel...@lidarlatinoamerica.com

W: https://www.lidarlatinoamerica.com/

L: https://www.linkedin.com/in/nelson-mattie-485052177/

T: https://t.me/nmattie