5 ms·

One of the big changes happening in the industry is increasing collection of road line quality data. Previously we'd use handheld meters to spot check pavement

by mastax 5y ago

One of the big changes happening in the industry is increasing collection of road line quality data. Previously we'd use handheld meters to spot check pavement markers. The new state of the art is to use continuous laser scanners attached to vehicles to measure all of the pavement markers on the road. This data is used to identify what areas need to be re-striped, to see how different materials fare in different conditions over time, and to submit warranty claims when things aren't up to spec.

toomuchtodo 5y ago
I'm surprised Tesla (or similar data collection fleets, like Google Street View) isn't measuring this and offering it to road authorities as a dataset. They have 1 million+ vehicles traversing roads daily, and they're already collecting labeled data (10 second video clips) for Dojo. Lane markings directly impact quality of driving assistance experience, so vehicle manufacturers have a vested interested in high quality road/lane markings.
- Dangeranger 5y ago
  Transit authorities care most about the quality of the data and the chain of control over the data from collection to QA and finally to delivery. Most commercial LiDAR trucks are accurate down to millimeters in relative terms and centimeters in absolute position terms. The equipment has been certified by approved surveying bodies, and there are official records as to that specific piece of collection equipment’s accuracy and precision. If any autonomous car company could deliver data to that standard then it would be valuable. But to my knowledge none of them can, because the accuracy of measurement systems for surveying roads at motion are so specialized and expensive.
  - toomuchtodo 5y ago
    Do you have any links on the topic you could share? Very interested in learning more! I am a bit of a precision positioning and fleet data collection wonk.
    - dpelaez10 5y ago
      Would also be interested in some more info! @toomuchtodo I'm just breaking into the fleet data collection space - any tips on where to go to learn more?
gertlex 5y ago
State of the art, aye, but definitely funny hearing about the job I was at 7 years ago (where things were fully operational, not experimental) being referred to as so new! When I was there, it was a combination of (1) downwards facing line lasers measuring distance + reflectivity, (which was viewed as a greyscale image, around 4k pixels == width of a lane iirc) and (2) forward facing cameras for tagging assets like signs, plus (3) velodyne lidars for much coarser 3D point cloud (including measuring bridge heights). The scanning worked up to a bit under 70 mph, which was impressive to me, particularly that first part of the sensor package.
- mastax 5y ago
  Well, yes. The 'art' now is convincing state DoTs that the tech is worth investing in ;)

by mastax 5y ago

toomuchtodo 5y ago

I'm surprised Tesla (or similar data collection fleets, like Google Street View) isn't measuring this and offering it to road authorities as a dataset. They have 1 million+ vehicles traversing roads daily, and they're already collecting labeled data (10 second video clips) for Dojo. Lane markings directly impact quality of driving assistance experience, so vehicle manufacturers have a vested interested in high quality road/lane markings.

Dangeranger 5y ago
Transit authorities care most about the quality of the data and the chain of control over the data from collection to QA and finally to delivery. Most commercial LiDAR trucks are accurate down to millimeters in relative terms and centimeters in absolute position terms. The equipment has been certified by approved surveying bodies, and there are official records as to that specific piece of collection equipment’s accuracy and precision. If any autonomous car company could deliver data to that standard then it would be valuable. But to my knowledge none of them can, because the accuracy of measurement systems for surveying roads at motion are so specialized and expensive.
- toomuchtodo 5y ago
  Do you have any links on the topic you could share? Very interested in learning more! I am a bit of a precision positioning and fleet data collection wonk.
  - dpelaez10 5y ago
    Would also be interested in some more info! @toomuchtodo I'm just breaking into the fleet data collection space - any tips on where to go to learn more?

gertlex 5y ago

State of the art, aye, but definitely funny hearing about the job I was at 7 years ago (where things were fully operational, not experimental) being referred to as so new! When I was there, it was a combination of (1) downwards facing line lasers measuring distance + reflectivity, (which was viewed as a greyscale image, around 4k pixels == width of a lane iirc) and (2) forward facing cameras for tagging assets like signs, plus (3) velodyne lidars for much coarser 3D point cloud (including measuring bridge heights). The scanning worked up to a bit under 70 mph, which was impressive to me, particularly that first part of the sensor package.

mastax 5y ago
Well, yes. The 'art' now is convincing state DoTs that the tech is worth investing in ;)