Strathmore Motorsports Park Sim Track Project
Posted: Sun Apr 17, 2022 1:19 am
Hello everyone!
The few I've met and chatted with at the Rookie School know me as Eric, I'm looking forward to meeting everyone face to face at the track, come say hi and give me a free pointer I'll need it (just look for the clueless guy )
I'm new to the club this year, "discovered" CKRC looking for AutoX events and realized where the REAL seat time was at.
However learning about CKRC while snow was still on the ground was both a blessing and a curse, lots of time for reading and doing my homework, but also time for Alberta to do it's "thing" and bait us one week and kick us the next so I was a little disappointed when I learned there was no Sim track available as a tool to study the layout, know where the marshal stations are, the pixel lights and get a general feel for the track prior to showing up as a green horn.
So I'd like to attempt to create that
Some of you familiar with iRacing might also be with LiDAR or laser scanning tracks down to the millimeter, unfortunately I personally don't have access to any of this equipment so the technique I'll attempt for the scope of this project to start is entirely within Photogrammetry (the use of photography in surveying and mapping to measure distances between objects). The biggest drawback to photogrammetry is the road surface will not be matched to the mm on the Z axis (depth), unfortunately only to around ~3-4+/- cm and road feel though FFB (force feedback) in the Sim would have to be manually tuned in but eventually adequate though community feedback, but this will be exponentially more accurate than any free available topography mappings. The good news is if LiDAR data is collected at a future date it can be combined with the photogrammetry for best of both worlds, mm accurate roads and accurate visuals.
My first goal is to get out to the track with a drone and a fist full of memory cards to take purposeful drone aerial's and terrestrial photos walking the track in both directions with all curb angles. (The shots from Michael Decnodder's video for proof of concept below were useful but not before a lot of post processing the keyframes to mask out the sky, drone shots for this type of recreation need high overlap and a more direct angle to the ground, no sky seen, the output is very dependent on the source)
I do want to preface that although I'm a "jack of all trades" IT professional day in day out I mainly deal with hardware and 3D modeling is purely from a hobbyist standpoint, no one has paid me to do any of this kind of work... nor should they That being said if you do have some skills you'd like to contribute, I would love to collaborate with all who would like to help!
I have left a couple proof of concepts below if you're interested or want to help with development and I look forward to hearing any and all feedback!
Eric M
Started with just slapping a repeating cross section surface down on a grainy satellite image imported the sparse google elevation data in the area. Road width, camber, grade are all wrong but functional in Assetto Corsa and sorta lap-able, I didn't want to waste time optimizing anything so you'll find the last 4-5 corners really slapped together, no sense going further with this workflow/development, more just to ensure I can modify material properties and collision and actually get something to load.
Video - https://imgur.com/a/BjZO6gA - (Assetto Corsa Download link below) https://drive.google.com/drive/folders/ ... sp=sharing
This model (below) was projected using these 3 images all shot from the same perspective moving forward a few feet between (with a cellphone camera at night). Total capture to render time ~10 min
This one was "extracted" from ~150 keyframes from the first 18s of Michael DeCnodder video https://www.youtube.com/watch?v=zZebYGEokQE and required lots of masking to the source images. The holes in the first pictures mesh is due to only having one flight direction and not purposefully taken photos with a sharper downward angle, the camera looks across the horizon and in recreation this also leads to skewing of the alignment making holes.
After I masked out the sky and horizon portions of the source images I was able to reduce the skew from the single flight angle substantially
The few I've met and chatted with at the Rookie School know me as Eric, I'm looking forward to meeting everyone face to face at the track, come say hi and give me a free pointer I'll need it (just look for the clueless guy )
I'm new to the club this year, "discovered" CKRC looking for AutoX events and realized where the REAL seat time was at.
However learning about CKRC while snow was still on the ground was both a blessing and a curse, lots of time for reading and doing my homework, but also time for Alberta to do it's "thing" and bait us one week and kick us the next so I was a little disappointed when I learned there was no Sim track available as a tool to study the layout, know where the marshal stations are, the pixel lights and get a general feel for the track prior to showing up as a green horn.
So I'd like to attempt to create that
Some of you familiar with iRacing might also be with LiDAR or laser scanning tracks down to the millimeter, unfortunately I personally don't have access to any of this equipment so the technique I'll attempt for the scope of this project to start is entirely within Photogrammetry (the use of photography in surveying and mapping to measure distances between objects). The biggest drawback to photogrammetry is the road surface will not be matched to the mm on the Z axis (depth), unfortunately only to around ~3-4+/- cm and road feel though FFB (force feedback) in the Sim would have to be manually tuned in but eventually adequate though community feedback, but this will be exponentially more accurate than any free available topography mappings. The good news is if LiDAR data is collected at a future date it can be combined with the photogrammetry for best of both worlds, mm accurate roads and accurate visuals.
My first goal is to get out to the track with a drone and a fist full of memory cards to take purposeful drone aerial's and terrestrial photos walking the track in both directions with all curb angles. (The shots from Michael Decnodder's video for proof of concept below were useful but not before a lot of post processing the keyframes to mask out the sky, drone shots for this type of recreation need high overlap and a more direct angle to the ground, no sky seen, the output is very dependent on the source)
I do want to preface that although I'm a "jack of all trades" IT professional day in day out I mainly deal with hardware and 3D modeling is purely from a hobbyist standpoint, no one has paid me to do any of this kind of work... nor should they That being said if you do have some skills you'd like to contribute, I would love to collaborate with all who would like to help!
I have left a couple proof of concepts below if you're interested or want to help with development and I look forward to hearing any and all feedback!
Eric M
Started with just slapping a repeating cross section surface down on a grainy satellite image imported the sparse google elevation data in the area. Road width, camber, grade are all wrong but functional in Assetto Corsa and sorta lap-able, I didn't want to waste time optimizing anything so you'll find the last 4-5 corners really slapped together, no sense going further with this workflow/development, more just to ensure I can modify material properties and collision and actually get something to load.
Video - https://imgur.com/a/BjZO6gA - (Assetto Corsa Download link below) https://drive.google.com/drive/folders/ ... sp=sharing
This model (below) was projected using these 3 images all shot from the same perspective moving forward a few feet between (with a cellphone camera at night). Total capture to render time ~10 min
This one was "extracted" from ~150 keyframes from the first 18s of Michael DeCnodder video https://www.youtube.com/watch?v=zZebYGEokQE and required lots of masking to the source images. The holes in the first pictures mesh is due to only having one flight direction and not purposefully taken photos with a sharper downward angle, the camera looks across the horizon and in recreation this also leads to skewing of the alignment making holes.
After I masked out the sky and horizon portions of the source images I was able to reduce the skew from the single flight angle substantially