W.I.P. Grand Prix de Trois-Rivières road layout

[maruto]

Let's try to put theory into practice.

Can you extract data from that Martin Brousseau onboard video that you showed us?

Can you isolate the biggest "peak" found in your data that would represent the "major bumps" found on the ideal line of the circuit?

IT would be nice to have a video that starts exactly at the starting grid (0) and ends exactly at the starting grid (1) so we could have compare the lenght of the video with the relative lenght of the track... and easily locate those spots....

this data could be then verified on site.

thanks, this is quite interesting

 

[physicsimulation]

All my discussion were related to a proof of concept concerning the possibility to obtain track roughness data from video stabilization of driver cam helmet or frame cam capture, aiming the horizon or the driver helmet.

The Martin Brousseau video is not so well for the stabilisation process.
The attached image shows which part of the video image is used to analyse motion.
- exclude the lower part (removing also the rolling shotter effect 8( )
- exclude the mirror content (which is AT the horizon 8( )
- red arrow shows to much mouvement to be included

With this in mind, I tried to stabilize the video, as best as possible.
I got something but the analyzer lost track of the moving pixels too many times to get interpretable result.

I got better result from "GUY MARTIN vs MICHAEL DUNLOP Isle of Man TT RACES", loosing track under tree shadows.

 

[maruto]

All my discussion were related to a proof of concept concerning the possibility to obtain track roughness data from video stabilization of driver cam helmet or frame cam capture, aiming the horizon or the driver helmet.

The Martin Brousseau video is not so well for the stabilisation process.
The attached image shows which part of the video image is used to analyse motion.
- exclude the lower part (removing also the rolling shotter effect 8( )
- exclude the mirror content (which is AT the horizon 8( )
- red arrow shows to much mouvement to be included

With this in mind, I tried to stabilize the video, as best as possible.
I got something but the analyzer lost track of the moving pixels too many times to get interpretable result.

I got better result from "GUY MARTIN vs MICHAEL DUNLOP Isle of Man TT RACES", loosing track under tree shadows.

Too bad, i tought the first split screen video you showed us was pretty stabilised. could we at least see the .txt file it produced to give us an idea?
@NiTZHD can you look for some other onboard camera you might have?

 

[physicsimulation]

Best capture :
0) Full HD, 60 fps
1 ) Camera on frame body (lower frequency vibration then helmet)
2) Seen the horizon (without car in front or with car but surrounding elements)
3) Seen driver helmet (if possible. vibration frequency is high and not usefull on dirt track)

Analysis:
phase 1 : frame body to horizon stabilization
phase 2 : horizon to driver helmet stabilization (optional)

Note : I never though to use image stabilization to get track roughness. I just though after Martin call me. So excuse my theoretical perspective but I am I am a strong thinker with a really good pratician objective, with time.

 

[Johnr777]

I don’t want to sound rude, and I appreciate new concepts that provide good data. But what’s the point of your posts if all you share is a theory?

you mentioned earlier this software you use provides an export option of camera movement data. Without sharing that detail, there’s no point in talking about how to capture video if we don’t know what to do next.

 

[physicsimulation]

I don’t want to sound rude, and I appreciate new concepts that provide good data. But what’s the point of your posts if all you share is a theory?

you mentioned earlier this software you use provides an export option of camera movement data. Without sharing that detail, there’s no point in talking about how to capture video if we don’t know what to do next.

Bye


Do you only eat fast food ?
Are you buying only builded house ?
Are you living with your parents ?

You seams to play ACM but before airplane fly, someone play with hypothesis and elaborate theories. The first serious simulator engine was build for airplane and aircraft using analog computer, yes, analog computer and now you play with your super computer, super display card using shaking seat to render track roughness.

What I understand, and try to figure out is how to get track roughness as simply as possible without going to the track, and sufficiently real to shake your seat and make it in your pants.

 

[luchian]

I'm guessing there's some language barrier at play.

Your initial message reasonably led to believe that there's going to be a solution of some sort to go through the process. It seems that in fact it was meant as a (constructive) discussion around an idea/a theory - which is fine by the way.

The invitation for a "group effort" in order to figure out how to do it was not that obvious. And probably created false expectations .

***
Au cas ou c'est mieux en Français: au début il semblait qu'on discutera d'une solution (c'est la plus part des cas sur le forum), et non pas d'une théorie. C'est cela qui a créé des fausses attentes, je pense. Merci pour la contribution.

 

[maruto]

When I try something for the first time, I don’t expect it to work perfectly , an experiment going well the first time needs to be tested again. Gotta try and fail and try and fail.

I don’t expect anything from that experimentation, if it could work it would be really interesting and could simplify the process of implementing bumps in trackmaking. Maybe not. Gotta try!

This discussion got me thinking how to implement the results into a script to generate the final physical mesh. those results( bumps location) can either come from our experiment, or (and) different sources.

There is no hurry, nitzh wanted to do a 360 GoPro on-board lap anyway, maybe he can try to follow your instruction when he does.

Can we give it a second chance?


Thanks guys.

 

[Johnr777]

Yeah, maybe a language barrier. But he doesnt need to post video settings and video "analysis" every time we ask for the software name or for this supposed TXT file this software can output. If he wants to keep this a secret, them by all means, go right ahead. But to mislead everyone with numerous posts that seem to become a dead end, then what's the point?

@maruto , in an effort to help you with actual accurate data, here is the LiDAR from Foret ouverte in FBX format. Its high poly, so beware when importing.

https://drive.google.com/file/d/17FOmYfHDy_DE1_PFTp85uvyyWNyLLQs5/view?usp=sharing

 

[maruto]

@maruto , in an effort to help you with actual accurate data, here is the LiDAR from Foret ouverte in FBX format. Its high poly, so beware when importing.

https://drive.google.com/file/d/17FOmYfHDy_DE1_PFTp85uvyyWNyLLQs5/view?usp=sharing

View attachment 5674

Wow... thank you so much.
The .fbx model opens fine in rhino, it is quite lite i can work easily with it.
just to compare i brought the old contour and projected it on the lidar data, some corners are quite off as you said...

Do you think there is a way of importing the aerial photo from foret ouverte to be georeferenced exactly on top of that mesh?

That Lidar data is perfect. I need to learn how to extract that too.
I downloaded and installed Qgis, an open source program that process .tiff format, it seems quite powerful, has tons of plugins (LAStools), but it needs some time to do some reading on how to extract data, will do that later for a future project... But it might ba a good alternative to global mapper.

Thanks again for your help on that matter, i'll come back to you for questions later

 

[Johnr777]

Wow... thank you so much.
The .fbx model opens fine in rhino, it is quite lite i can work easily with it.
just to compare i brought the old contour and projected it on the lidar data, some corners are quite off as you said...

View attachment 5675

Do you think there is a way of importing the aerial photo from foret ouverte to be georeferenced exactly on top of that mesh?

That Lidar data is perfect. I need to learn how to extract that too.
I downloaded and installed Qgis, an open source program that process .tiff format, it seems quite powerful, has tons of plugins (LAStools), but it needs some time to do some reading on how to extract data, will do that later for a future project... But it might ba a good alternative to global mapper.

Thanks again for your help on that matter, i'll come back to you for questions later

View attachment 5676

Glad you are able to see and use that data!

The original track doesn’t seem that far off, a better match than a lot of tracks I’ve compared with LiDAR in the past.

I didn’t see a download for the ortho image though... you might need to do things the crappy way, which is creating your own ortho map from google most likely, then unwrap it on to this mesh. Adjusting the map until it fits.

And no worries, will do my best to answer your questions

 

[maruto]

Hello guys, keeping you up to date:

The Lidar mesh is nice, i Made a aerial picture from Foret Ouverte and unwrapped it onto the mesh ( not easy had to fight a little to fit it right , not 100% perfect, but more than ok)



I started drawing the contour of the track, positioning manholes, concrete patches...



@Martin Dupuis , @NiTZHD , I would need someone to go on site to get me some information : the diameter of the small and big metal plates found on the streets ( in cm please) and some picture of them from above so i will start doing texture atlas for the track (three types small around 20 cm, big full, around 100 cm, and crate, there are some square crate in turn 2, the long straight too, would be nice to incude those too... While beeing there it would be nice to have the exact height of the concrete walls surrounding the track, i noticed in street view some are higher in the pit area... I'll use the height from Gilles's model for now, but i would like to be sure.

oh yeah, one more thing, very important: The exact width of porte duplessis, when i measure in the model i get around 7.20m



Thanks

 

[NiTZHD]

We do have more onboard videos but the Gopro is located at the same place everytime. Does the gopro telemetry can help for the bumps?

I'll be uploading the 360 video tonight and will get the dimensions from what's needed tomorrow night.
Keep on the good work!

 

[maruto]

yo.

I finally got some time today to work on the track, time is gold nowadays...

I shrinked the track contour to the lidar mesh, only using much fewer control point to have a smooth limit, correcting where the mesh was messy and jaggered. I extracted points from the lidar mesh inside those limits and removed some (those in red in the picture).

Most of the points were fine except some part..

But the Lidar is really precise, you can see the camber!

So i filled holes with corrected point (green)

 

[maruto]

Here's a first drivable mesh, for analysis.
98% lidar points, no texture.
bumpy but not as much as i expected.
Biggest bumps in the long straight, might have to smooth a little.
@Johnr777 , when you got some time, can you have a look at it and tell me what you think? if the density is enough, if too bumpy, etc...
@NiTZHD , same thing, if M. Brosseau can try it to see what he thinks of it...

Thanks.

edit: (driving it at night with lights on allows you to see the road irregularities and bumps...)

 

[Johnr777]

Here's a first drivable mesh, for analysis.
98% lidar points, no texture.
bumpy but not as much as i expected.
Biggest bumps in the long straight, might have to smooth a little.
@Johnr777 , when you got some time, can you have a look at it and tell me what you think? if the density is enough, if too bumpy, etc...
@NiTZHD , same thing, if M. Brosseau can try it to see what he thinks of it...

Thanks.

edit: (driving it at night with lights on allows you to see the road irregularities and bumps...)

View attachment 5749

Very bumpy in my opinion, which is typical of using aerial lidar without smoothing it. The irregularities and bumps from your photo above is unfortunately lidar noise, which should be removed.

I get its a street track, and they are rougher sometimes, but you introduce that later via microbumps to the physical mesh. Aerial lidar is really just to provide elevation and corner accuracy, surface noise is added later.

But good start!

 

[maruto]

Ok I understand.

But i feel it could be possible to create a nice mesh directly from the lidar points, including camber and all that information that is include in the data... maybe i'm wrong. but i think it is worth some trial and error...

In Rhino there are a few methods to create mesh, one is directly using a point cloud. with a sample parameter, one can control the size of the cell, Rhino then takes the average distance between all those points within that aera with an algorithm and this will smooth the transition between points. That's how i did the first mesh, using a very low sample rate to catch as much info from the point cloud as possible.

Tonight i tried to smooth thing's up by taking less points into account when creating the mesh. with bigger cells.

you can see there are some points not showing because they are under the mesh and the one we see are above. so in a way it cancel's the noize.

I tried two different sample rate, i think it it better from yesterday , but still too bumpy i guess. Turn 1 and two are still pretty harsh

Next time i'll try a different method.

(I won't upload all of my everyday trials but this one is for M. Brousseau, and for those interested
i completed with a quick ai file so that the laps time works, and started a section.ini files but not well synchronized yet... )

this link will last a week...

https://we.tl/t-AZ4BO0HG14

(@NiTZHD sent me today some really nice 360 videos i hope he will make them available to you guys too.)

 

[Martin Dupuis]

Hello guys, keeping you up to date:

The Lidar mesh is nice, i Made a aerial picture from Foret Ouverte and unwrapped it onto the mesh ( not easy had to fight a little to fit it right , not 100% perfect, but more than ok)

View attachment 5713 View attachment 5719

I started drawing the contour of the track, positioning manholes, concrete patches...

View attachment 5717 View attachment 5716 View attachment 5715

@Martin Dupuis , @NiTZHD , I would need someone to go on site to get me some information : the diameter of the small and big metal plates found on the streets ( in cm please) and some picture of them from above so i will start doing texture atlas for the track (three types small around 20 cm, big full, around 100 cm, and crate, there are some square crate in turn 2, the long straight too, would be nice to incude those too... While beeing there it would be nice to have the exact height of the concrete walls surrounding the track, i noticed in street view some are higher in the pit area... I'll use the height from Gilles's model for now, but i would like to be sure.

oh yeah, one more thing, very important: The exact width of porte duplessis, when i measure in the model i get around 7.20m

View attachment 5718

Thanks

HI, would be hard for me to go there, I live 1h30 away and don't go there as often as I used to unfortunately.
But I found this PDF and I thnik is has everything you need to know about every type of manhole possible in Canada

Sorry to be of no further help. I'll try the track now

 

[maruto]

HI, would be hard for me to go there, I live 1h30 away and don't go there as often as I used to unfortunately.
But I found this PDF and I thnik is has everything you need to know about every type of manhole possible in Canada

Sorry to be of no further help. I'll try the track now

thanks for the info on the iron casting plates, i'll use these dimension ; i wasn't too far....

Don't be sorry @Martin Dupuis , i'm grateful you took interest in the project and i'm pretty sure you will testdrive this track as the project evolve and provide me with feedback later on ... right now is still pretty raw , i didn'nt have much time to work on the track this week, I found a way to smooth the mesh, but not yet satisfied...

Open Optimisation question: when working with circular object in meshes, it creates a lot of triangles, how can we optimize that? for the visual mesh i was thinking of having a simple square quad mesh with the circular iron cast plate and simply use transparency. but for the physical mesh i would have liked to put the metal surface in a separate layer so that maybe we can feel when we drive over it...

It is too much going into details? when racing in real life, do driver have to be careful with those aera? is it more slippery? does it produce a different sound? I only tough about it because there is one right in the middle of Ryan corner.

Same thing for the concrete reinforcement in the corner, how does it compare in term of adherence with asphalt? Concrete must be harder so my guess it is has less adherence?

 

[physicsimulation]

http://code.eng.buffalo.edu/dat/sites/tire/tire.html
https://en.wikipedia.org/wiki/Friction#Coefficient_of_friction
https://www.google.com/search?sxsrf=ALeKk02xgtp-708l5_TaY719m07FO-hhog:1600082540222&source=univ&tbm=isch&q=coefficient+of+friction+table+for+vehicles&safe=active&sa=X&ved=2ahUKEwjZgK_7w-jrAhUuvlkKHXj2BeQQsAR6BAgKEAE&biw=1920&bih=937#imgrc=POGaQO5w-Cqx-M&imgdii=M_bZlbiCE-VABM
Sorry, it's theoritical.