Hello everyone. I am just curious about how the points down sampling is done in the Ouster OS1 LiDAR, like when an object is being so close to the LiDAR. What type of down sampling method is used to decrease the density? Is it possible to be changed, or to enable or disable it? Thank you.
It is really up the user to decide how they want to downsample, you can use utilize methods from know pointcloud libraries like Open3D to handle this such as:
- https://www.open3d.org/docs/0.6.0/python_api/open3d.geometry.voxel_down_sample.htm
- Downsampling a PointCloud using a VoxelGrid filter — Point Cloud Library 0.0 documentation
The OusterSDK has some builtin method to but it is still under development and the interface of this method may change ouster.sdk.core — Ouster Sensor SDK 0.16.0 documentation (you can search the code to find how this method is applied and used).
Hope this helps!
The python interface of the OusterSDK have this public method ouster.sdk.core.scan_ops module — Ouster Sensor SDK 0.16.0 documentation which you may find interesting too (this acts differently than downsampling by reducing the number of beams rather than having uniform voxel grid like the other filters).
@Samahu Thank you for your fast response! I will take a look at it.
I was using Ouster Studio to get the point cloud from the OS1 lidar. Based on the data I collected long time back, what I noticed was that the horizontal resolution gets lower on some neighboring points on a closer object to the lidar. I just wanna confirm, is it true that there was a down sampling or filtering happened? If yes, was that done by the SDK, or something else? Will the same thing happen on the side of ROS pipeline? How to regulate this sampling?
I was thinking this is so normal as on a closer object, the distance between one point to its neighboring point will be lower, so it results a higher density. A higher density can create a less meaningful point, and there will be a need of points downsampling. I wanna confirm that this is true and done naturally (by default? I don’t know what’s the correct term XD) in the ouster perception pipeline.
Thank you in advance!
There is no downsampling performed on the lidar. Feel free to share screenshots or an Ouster Studio link if you want us to diagnose what you are seeing.
@yggdrasil Thank you for you confirmation! So currently I am working in a project where we need to identify the characteristics of lidar returns from a glass object. In the case of a glass plane, what I understand physically is that on the same glass object which is morelikely to have the same characteristics of reflection on all of its surface, the number of points we can get from that glass object will be same on all different distance due to its specular characteristics. So the angle of incidence of the lidar beam to be able to returned by the glass will likely be the same over all the surface. Followings is the illustration
But what I got is the closer the lidar to this glass plane, the lesser the number of points that I got, as shown in the below figure (you can just focus on the number of the points). This is then makes me conclude that there could be something like downsampling happened due to the points density that gets higher as the lidar gets closer to an object. Is this true or is it actually some kind of physical phenomenon?
To me, what was actually happended is seems to be like this, so that is why th enumber of points is less due to the down sample or something else
Thank you!
Oh btw, I am sorry that this is actually from a VLP-16 lidar. Hopefully it is still relevant
I can only send one media per reply so I made it to be in multiple reply
No problem. This is an unusual use case for any lidar for two reasons - you are measuring in the very short range (~1m) and you are measuring glass. Unfortunately I don’t think we can give you an expectation for how this might perform for your application and you will have to test and interpret the findings yourself.
Oh yeah, no problem. At least for now, the confirmation that this is not a down sampling is enough. Thank you so much for your reponse, @Samahu @yggdrasil !
Then, I will just close this.