Weird effect when scanning cylindrical highly reflective objects

[Kevin]

Hi all,

I have recently received the Livox mid-40 and am impressed with how easy it is to work with and it's high point density and good range.

I have however noticed something weird, which I hope someone here can fix and/or explain.

When I am scanning a highly reflective/shiny cylindrical object (e.g. an aluminum ventilation tube as shown below but also happened with shiny black plastic tubes) the returned data is not cylindral but deformed, usually with a dent in the center of the object (surface normal to the sensor). 

The shown pipe has a diameter of 16cm and was positioned at approx 3.5m from the Lidar sensor.

The resulting point cloud in the Livox Viewer is shown below (colormap reflectivity):

front Perspective left Top Bottom 

You can clearly see a sort of canyon over the center of the pipe, except for where we have put a white A4 piece of paper. Furthermore, you can also see clearly on the top and bottom views that the highly reflective points in the 'canyon' are actually measured as being inside the tube. In other words the range measured for these points is incorrect.

Is there something wrong with my device, the livox viewer data interpretation or is this something all Lidars have issues with?

Thanks!

Kevin

[Livox Dev]

Hi Kevin, 

Could you please record the data for this scene for a few seconds with the livox viewer, send it to d...@livoxtech.com or post it here, which will help us analyze your problem. We are also trying to reproduce your scene, but it may take more time.

Thanks.

[ano T]

front and back are reversed from what I can tell

[Kevin]

Hi Livox Dev,

Hereby some recorded data of the test set-up shown in the picture below: https://we.tl/t-nZvTp1D1qm

As you can see I have placed three tubes in front of the Livox sensor. One large diameter reflective tube with a much-less reflective A4 paper on it, a small diameter PVC tube inside the large tube and a medium diameter cardboard tube, with some reflective tape on it.

In the data you can see that the cardboard tube itself and the white A4 are scanned to be circular like you'd expect. The reflective tube and the reflective tape on the carboard tube distort the scan and return a non-circular shape.

Hope this is any use, please let me know if you need anything else!

Kevin

[Livox Dev]

Hi Kelvin, thanks for sharing with us the data. Our team is analyzing the data, and tried to reproduce your exact problems. This may take a while (e.g., a few days), but we will definitely update you ASAP.

[Kevin]

Hi Livox Team,

Any findings on this?

Thanks!

[Livox Technical Support]

Sorry for the late reply. We analyze the data, and reproduce your problem.

We find that your ‘weird effect’ is caused by the sharp contrast between the different reflectivities at the center of the pipe and the side of the pipe. For the center of the metal pipe, the laser is directly back-reflected (like retroreflector), while for the side of the pipe, the laser is scattered and a very weak laser pulse returns to the receiver of the Lidar. Unfortunately, the ranging precision of our lidar will deteriorate from the normal case for very weak reflection and very strong reflection (retroreflection). Such deterioration of precision still meets the specified precision (1σ = 2cm, 4σ = 8cm), but it seems to be weird to see the distortion when the very weak reflection and retroreflection appear simultaneously on the same material.

Now, we are working to improve the accuracy of our lidar. However, this may take a long time. If ranging precision is very critical to you (for example 1σ < 1cm), please let us know. 

[Kevin]

Dear Livox,

Thanks for you explanation, makes sense. The pipe was now quite close to the lidar as well, I will be testing if this effect is as severe for longer distance and bigger diameters as well. Depending on that I have to see if this accuracy is sufficient for my application, but higher accuracy is always better anyway!

Is there perhaps no way other lidar manufacturers are correcting for this?

[kk...@matterport.com]

Theory: It's ranging long due to multi-path reflection at the shiny parts of the pipe and floor.

Both the floor and the table look to be fairly reflective. I can see the reflection of the pipe in both images. If you are hitting the pipe and it reflects down and off the table back to the sensor, then it would rang longer. It would then make sense that the top of the pipe is not doing this to such a degree, but it's hard to see from your images if that's the case. Try the pipe on non-reflective floor surface (like a rug) and lower the lidar so its pointed parallel to the pipe and see if it gets better. Then, of course, let me know! I'd test it myself but can't find a pipe like yours.

[Kevin]

Thanks for thinking along! I have tried your suggestions, measuring on a rug at different distances, trying to cancel reflections except from the tube, but it doesn't seem to matter much. The reflective pipe keeps having this u-shape at the center. Probably Livox Tech is correct.

Placing the pipe right in the center of the laser pattern makes the u-shape look most severe, but then again it will get most hits by the Lidar in that position.

Angling the Lidar relative to the pipe orientation didn't seem to make much difference as well.

Looking at a (much less reflective) cardboard tube under the same circumstance returns a nice round shape. 

Did everyone ever encounter this with other high-res 3D Lidars as well (e.g. Velodyne equipment)?