Discussion: SGM vs LiDAR

Hi all,

Does anyone have experience with data derived from the SGM ('semi-global matching') process, as opposed to LiDAR? One of the local aerial survey companies waved it at me as a significantly lower-cost alternative to flying LiDAR but I'm curious as to the quality of its output, especially in evergreen forest.

I was trying to get examples of the DEM output from SGM from the survey company, but the only ones they gave me were of open-cut mines with no vegetation in sight. That data does look very impressive though.

Any or all input would be gratefully received. :)

cheers,
Paul

No experience with SGM, but a quick read of this article, written by a group promoting the method, indicates that SGM processing of stereo aerial imagery will do best in open areas. The authors indicate that LiDAR >> SGM wherever there is significant vegetation.

Heh...the survey company sent me that article too. :)

The profile image on page 5 suggests that both of them are pretty useless under tree cover, since both methods return <5 points under the canopy of the tree on the right. Most of the LiDAR discussion on AP (and OllesMaps etc) seems to be from countries where leaf-off data is a possibility, but our forest (while relatively open) is everbrown. It'd be evergreen if it ever rained.

My interest is primarily cost-driven, since the survey mob said they won't even bother getting in the air for less than a $30k LiDAR contract, but they quoted me <$15k for 100sq km of SGM. (giving me orthophotography *and* DEM)

Lidar seems to be working well over here despite being everbrown (taken at end of decade drought).

Are you flying your own, Neil? Or is one of your layers of government remarkably more competent at technology than anyone in WA? :)

Eucalyptus trees tend to hang their leaves vertically reducing the amount of sunlight they catch and minimising evaporation. deciduous trees tend to spread their leaves horizontally to maximise capture of sunlight. Ever noticed how Australian forests are generally less dark than European forests. So in general the leaf cover in eucalyptus forests is quite sparse and just as this meant it was easier for a photogrammetrist to see ground (and so produce good basemaps) it should also mean that LiDAR will also be pretty good. Of course this is a generalisation, may not apply to say the temperate rainforests of Eastern Victoria, but it is certainly relevant to the sorts of forests found around Perth and Bendigo.

I am no expert on LiDAR but the experience i have of using LiDAR data in Japanese cedar forests (evergreen 'needle' trees that form a pretty dense uniform canopy) is that it was surprisingly good. In similar terrain I have used professional orienteering photogrammetry and that was not good at all, but the contours from LiDAR are amazingly accurate and they pick up even very small features. So maybe 'needle' trees are more penetrable to LiDAR??

The density of the LiDAR points makes a difference - perhaps you can compensate for dense leaf cover with higher density LiDAR - which would cost more?? I am just guessing about that stuff.

I wonder how much LiDAR (ie how many km2 and how many different areas) you get for $30k. If you got all the local clubs together and planned 5-10 yrs in advance you might be able to make it worthwhile. We used to do that in Victoria with low altitude aerial photos for photogrammetry.

That sort of long term 'infrastructure' project might also be the sort of thing that you could apply for a govt grant for.

Everything in WA is run through the state association, not individual clubs, so nothing to coordinate on that front, but nobody to cost share with either. (Different to the rest of the country). Don't know how many future areas we've got identified as worthy of mapping...

A consortium of govt agencies (catchment managers, local govt, state govt) coordinated a flight across a number of sites in the north of the state. It was mainly concentrated in flood plains which aren't good orienteering country. But some local govt bodies did a more extensive coverage, including Bendigo. It so happens that somehow Bendigo collected a staff member who is truly good in this area. So they did a lot of processing themselves. As an indication of their progress, they are now testing 3-D lidar models of their infrastructure using ground based observation and tracked vehicles under bridges etc. But I digress. The next step is that we gained access to some of this data processed data unofficially (If it was the US it would all be publicly available). Its a 1 metre grid. Elsewhere i have read that 75cm is ideal. But 1 metre seems pretty reasonable. As Rob says, the quality of contours is impressive, leaving older style photogram in its wake. I read in a presentation posted elsewhere on AP that the advantage of Lidar is that you only need one fix on the ground rather than a stereo pair.
A summary of our first ground truthing is here.
https://picasaweb.google.com/114573865667830672126...#
Yes, we were impressed. But we couldn't afford to pay for a flight ourselves. If we have an area we want lidared (and there are several) it will be a case of keeping our ears to the ground to hear of a consortium we can join. Not much money in govt at the moment though.

Something we should get Council to get a grant for Juffy? Given the crapness of Landgate contours for use on basemaps it would be worth investigating.

From Department of Planning website ... Department of Water might have what you want ...

The Federal Government is coordinating a national coastal vulnerability assessment, with the aim of identifying coastal environments, communities and infrastructure at risk from the impacts of climate change. The Department of Planning and Department of Transport's LiDAR survey and resulting datasets will add significant high-resolution detail supporting these aims.

The project also complements a project being undertaken by the WA Department of Water. In 2008 the Department of Water commissioned Fugro Spatial Solutions to undertake a land-based LiDAR survey of 7000 km² of the Swan Coastal Plain. The survey was conducted by in February 2008 at 1metre x 1metre resolution. The Digital Terrain Model which resulted is being used to determine the patterns of flooding, groundwater/surface water interaction and ecological systems.

The data is now available for distribution to government agencies and external stakeholders.

For further information or data orders please contact Peter Muirden at the Department of Water on telephone (08) 6364 6925 or email peter.muirden@water.wa.gov.au.

Heh... No, I've already talked to them - the key words there are 'swan coastal plain'. Their data is limited to all the flat boring stuff. :(

I thought they had more lidar data for other catchments including the Peel-Harvey and Lechenaulte(?) which looked like it included the hills.

Organisations having a problem with flash floods is your best bet.

You make me laugh!! Flash floods - in Perth haahhaahahahahaha. We'd be lucky to have a flash trickle.

Any Australian mappers wishing to have a go at mapping (i.e. fieldchecking) some US terrains - perhaps even using publicly available LiDAR for basemapping - may be able to take advantage of the E-3 visa program, with Orienteering USA as the sponsoring employer on behalf of one or more local clubs. Contact me if interested.