Live data from Hacker News

LiDAR scans are finding hidden Roman roads and cutting crime

wired.co.uk

61–70 of 79 posts

Re: LiDAR scans are finding hidden Roman roads and cutting crime

#62
post #41
post #6

Has anyone seen any nice examples of work done with the open LiDAR data?

I downloaded all the 1m dsm composite data from the Environment Agency and Natural Resources Wales and used C#, System.Drawing.Graphics and hefty server from AWS EC2 to render 2 million png files, which I then moved onto a Scaleway C1 bare-metal ARM server, to produce this slippy lidar map of England and Wales: https://houseprices.io/lab/lidar/map

Incredible and you can even see ripples on the sea. I was wondering (as I live by the coast) what the deal was with the sea but I see that because they did multiple passes over the area, the tide is at different heights.

Re: LiDAR scans are finding hidden Roman roads and cutting crime

#63
post #5
post #3

I have a question about self-driving cars (SDCs). Some time ago it was mentioned that SDCs can't use deep-learning because it is not sufficiently reliable, and that instead techniques like LiDAR should be used to recognize objects. However, this made me wonder: how does the car recognize road-signs and traffic lights?

The car sends a photo of the sign back to google and a user decodes it in a CAPTCHA. (I'm mostly joking, but the answer does seem to involve big databases of what the signs and roads are expected to be. I'm not convinced that most of the current "self-driving" solutions can cope with roadworks, traffic cones, diversions etc)

Relevant xkcd: https://xkcd.com/1897/

Re: LiDAR scans are finding hidden Roman roads and cutting crime

#65
post #23

Oops, Vespasian wasn't Emperor in 43AD, that would have been Claudius. Vespasian was most likely present for the Roman invasion of Britain. He would become Emperor in AD 69.

Cerialis was appointed by Vespasian and fought the Brigantes, so it's more likely the year 43 was wrong and should have been ~71. Vespasian invaded Britannia in 43, probably confused the two.

Re: LiDAR scans are finding hidden Roman roads and cutting crime

#66
post #49
post #4

Earlier quoted context omitted.

I think you're confusing things. Deep Learning is a method of machine learning. LiDAR is a tool that collects depth data. Using Deep Learning on flat image data to try and detect depth is unreliable, hence SDCs use LiDAR to detect depth. They also use image data from the visible spectrum and beyond to read road signs. etc.

Ok, but what if the deep learning software of the SDC confuses a "no entry" sign with a 100mph sign? Or is DL classification at 100% accuracy?

I don't think 100mph signs exist, for a start. But the speed limits for sections of road are already known anywhere that a SDC is going to be used. It's an offline database, no sign-reading on the fly required.

Re: LiDAR scans are finding hidden Roman roads and cutting crime

#67
post #6

Has anyone seen any nice examples of work done with the open LiDAR data?

If you're looking for just general examples and inspiration for things that can be done with LIDAR data, the LAStools blog [1] has a lot of em. Their mailing list is interesting as well. I've been using LIDAR data for doing canopy height modeling, and screening remote areas for possible exceptional/champion trees prior to checking it out in person. Haven't put anything online yet because I'm very new to GIS and this…

What software are you using? I have found PDAL[0] to be very simple to run with docker, though building from source is a bit consuming, especially as you reach out for more plugins you didn't realize you wanted initially.

PDAL is my current software crush.

[0]http://pdal.io

Re: LiDAR scans are finding hidden Roman roads and cutting crime

#68
post #46

Is there any repository of lidar data for US? Especially coastal states like Florida, Texas and California?

US lidar still remains fragmented across many different portals.

USGS's 3DEEP data for the National Map covers lots of federal and private land. Open Topography holds lots of research data funded by various federal organizations (NSF, NOAA, others) amongst other stuff. State-level Departments of Natural Resources very commonly host survey data.

Re: LiDAR scans are finding hidden Roman roads and cutting crime

#69

Anyone know about LIDAR hardware i can purchase that is great for high resolution at low range. I'm looking for 1-3mm accuracy at a range from 0.1m up to 2m. 2D or 3D. IP67 rated.

LIDARs are awfully expensive for personal use. Look for other 3D scanning methods, and if you still think it's required, rent one.

Re: LiDAR scans are finding hidden Roman roads and cutting crime

#70

Self driving cars have lowered some LIDAR costs a factor of 20, with a promise of far more. I wonder if this can translate over into archeology and geology. Imsaw one project at SIGGRAPH using a cheap LIDAR.

If you're surveying on a national scale, then the costs of chartering flights (or driving around with drones) is going to keep costs high in the near future even if the surveying equipment itself plummets in cost

Eventually we'll get drones to do this job. But first LIDAR needs to get small and cheap enough and large drones must be developed to be reliable enough to fly over populated areas.
Post reply on HN