Register | Login
Attackpoint - performance and training tools for orienteering athletes

Discussion: Wow.

in: Swampfox; Swampfox > 2019-04-24

Apr 26, 2019 10:19 AM # 
Just, wow. Good call shifting your plans.

Sorry we'll miss you this weekend. Much epic-ness to be had. Just think, you could have taken Jordan on for the TT Long course at Turkey Mtn! What fun.

And although I hear D/Hist will be a few training areas over, shooting their historical weapons historically, they are usually pretty good about not aiming them out of the parking lot at 293...
Apr 26, 2019 4:10 PM # 
I am going to try to simulate being at West Point in the barberry this weekend by low crawling around a course in the sage at Telegraph Hill.

It won't really be the same of course, especially when I hit snow drifts, but it will still offer a challenge.

Hopefully the weekend at West Point will be a great success! The last I heard is that you folks were able to talk security out of requiring thumbprint biometric information from all the participants at all the controls on all the days, post-run wanding at the finishes, and a mandatory blood sample collected to prove you weren't a robot. That would have been a bear.
Apr 26, 2019 5:01 PM # 
But we might be able to find a bear or two. 'Tis the season for wandering yearlings of the non-yuc variety. Usually hanging around dumpsters at Round Pond or in the cadet area. But I think they got the bulletin that Anton's not competing, so perhaps they'll leave us alone.
May 4, 2019 2:54 AM # 
It was my first time at West Point or Harriman. I noticed your name on the map, and wondered what the original basemap was like. I pulled the lidar and processed it. This is probably 3 hours of me at the computer and at least 16-20 hours of the computer working while I was wasn't there.

The download is about 412 MB. Images labeled "veg" use basically all returns, so the trees really obscure details. They're of limited usefulness, but they're easy to create and occasionally are useful. The "alr" images are all last returns, which sometimes help see trails and vegetation under the canopy. The really beautiful and interesting ones (IMO) are the slope_gnd and relief_gnd.

I've been working on hydrology algorithms that try to show the stream network and help to identify marshy areas. For lakes and marshy areas where the filled DEM "floods", it shows straight lines, though.

I'm posting this here because I figure several people who might be interested will find it here. The OCAD files that aren't labeled "ocad8" are OCAD 11.

The lidar is from 2013. I didn't look at the post spacing, but it's quite good. If I can find 4-band aerials, I'll try the NVDI algorithm and see if I can get bare rock to pop out.

Anyway, I'm interested in any comments.

{edit} I grabbed a single aerial. The NVDI processing didn't find the rocks very well. It did find the evergreens and swamp green areas. Marginally useful, I'd say. The aerial is pretty good all by itself. If anyone is interested, I can upload them.
May 4, 2019 4:31 AM # 
Any interested eyes are much more likely to find this stuff if you post it under "Orienteering Topics" instead of here.
May 4, 2019 11:21 AM # 
Incidentally, the Academy tends to err on the side of caution by putting into the credits anyone who was ever involved in an earlier version of a map, even if there's no trace of their work left.
May 4, 2019 1:31 PM # 
Cedarcreek, You should see Jon and Jordan’s emails flying back and forth. Jordan will be back to teach soon and we are long - term on the east coast...I think I heard one of them mention they have processed the entire reservation? That’s a lot of lidar...

Passes your comments on to Jon, he’ll have a look. Thanks for sharing. :-)
May 4, 2019 2:45 PM # 
Thanks! It's possibly a misunderstanding regarding the map credit.

I am still interested in how lidar compares to a stereoplotter basemap of this area.

I'm still trying to figure out Terje Matheson's scripts. My understanding is that he prefers the very small contour interval for mapping areas not unlike West Point (complicated, rocky, glacial). I'm guessing the contours I sent, from Karttapullautin, may be be too simplified. Here in our Ohio-Kentucky-Indiana terrain, the KP contours are almost finished except for adding ditches and redrawing to make them smoother. If anyone is interested, I set the flag in KP to make basemap contours at 1.25m if anyone wants to see that. It's a few minutes work to import them into OCAD.

And I just realized I forgot the KP images (9MB zip):

I was really trying to reach Mikell, Victoria, Jon, and J-J (and Spike), so I figured that worked out okay.
May 4, 2019 3:55 PM # 
Thanks for putting the files out there.

Are you processing the NDVI in QGIS?

What's your source for the air photo? I've looked at:

I've been playing around with remote mapping of an area along the Taconic Parkway:
May 4, 2019 4:37 PM # 
Well, in that case cedarcreek, it looks like you have come to the right place after all (looking at the responses above.) I'm in the middle of listening to an AGM that is going to go on for a few more hours, but maybe tonight I can write something more, with the advantage of doing it then being that possibly the interim period will give me a chance to think of something useful to write. ; )
May 4, 2019 7:05 PM # 
Spike---I looked at TNM and found nothing, but EarthExplorer had 1 foot resolution aerials from 2013.

I do NVDI in the QGIS raster calculator, but to get more than the raw grayscale image, I have to assign three colors to the values and play with them to bring out details. My way is iterative. It would be nice to find a software tool with knobs to turn to bring out various details. NVDI does highlight water pretty well, but the main purpose is to differentiate photosynthesis ("green") from inorganic. So moss-covered rocks are mostly green.
May 5, 2019 5:28 AM # 
Unfortunately, the passage of a few hours of time was apparently not enough to fill me with amounts of brilliance and insights not formerly possessed. I was hoping this time would be different.

As I remember, the original photogrammetry was done from off the shelf photography of good quality, and the basemap produced was quite good. It certainly didn't in any way limit what was possible in terms of the quality of the finished map.

I have no idea of what re-workings have been done to Turkey Mtn in more recent years, or who might have been involved, but it appears to me that the map used last weekend (assuming that's what was used for RG) can be traced back to that original photogrammetic plot.

Comparing the those KP contours to the current map, they look fine. If you were preparing a basemap to take to the field, you could import them directly into OCAD as a first step. Of course you would be making edits from the fieldwork, but the vast majority of what was imported would make it unaltered to the final map--saving a *lot* of drawing time. There would be a good many lower height details you could--and would want to--add, by bringing in an image of the intermediate contours (1 1/4m or 1 m or whatever was generated), but still, these basic contours are fine--you could get around on them with no great problem.

Of course, when you're looking the the existing finished map, some of the differences with the KP map are because of the surveying. But other differences will be there purely from a difference in whatever reference--or starting--contour height was used, if there was a difference. If you start the first contour at 100m above sea level versus, say, 103m above sea level (I'm just making the numbers up), mostly things will look the same, but some finer details will look markedly different.

And where the basemap will struggle the most--any basemap, whether a photoplot or one from lidar--will be in a flatter, detailed area where the details are lower height with more gradual changes in shape. As an example, you can see how the KP contours don't look much at all like the contours in a flattish section between Turkey Hill Pond (SW of the pond) and the parkway. The photoplot probably missed a lot of those shapes as well--just guessing. Anyway, that's the kind of area where as the surveyor you would expect to have to use more time working with the contours, whatever the source.

If by "too simplified" you're referring to the amount of smoothing KP did, then again I'd say to me the contours look quite good in that respect--"lagom". At some point, microdetail (at the level that might get smoothed away) loses or has lost informational content and becomes noise--at least for a racing orienteer.
May 5, 2019 5:18 PM # 
Jon updated a square km (last loop for most courses) into the finish including lidar and georeferencing. There’s probably some magic to how that works with the rest of the map...
May 10, 2019 6:49 PM # 
An unexpectedly rainy Boston afternoon has put a damper on my gardening plans, so I'll jump into this.

Matt: Thanks--downloaded all files. I haven't looked at them yet, but certainly will.

The vast majority of the Turkey Mountain map was unchanged for April's event, so it is the original photogrammetric basemap and Mikell's 90s field checks. Just west of Camp Shea and the bridge there was a fully re-field-checked and drafted section. This section uses new LiDAR contours. If you look closely, you can find remnants of the stitch work. I jumped into that section because it had a high density of controls and given the bridge, it's a high-use area.

Here's sort of my approach to making LiDAR and field check (area) updates to the USMA maps, which are fundamentally sound from photogrammetry and Mikell's original field work and drafting. It applies to the on-post (ski slope map) which I updated when there as faculty in 06-09 and the areas used the last few years for the A-meet. Because the overall mapped area is so big (> 50 sq km, if I recall correctly), and there is so much detail (I think there were 40,000 individual stony ground dots just on the Turkey Mtn map), I only change an area if I can verify the changes with a field-check. And, I'll only field check with the suite of tools the LiDAR and aerial imagery puts at my finger-tips (no azimuths and distances for me). I haven't found an efficient way to update great swaths at once and not do more harm than good; for example, just pulling in LiDAR contours by themselves breaks more than it fixes. A boulder that was on the nose of a spur ends up on the E shoulder of the spur, stony ground down through a reentrant ends up on a hillside, etc.

Armed with the "old" map and the LiDAR and geo-referenced/orthorectified imagery, I will go section by section, getting the high-speed linear features (roads, trails, streams, stone walls, etc) right, then the linear blocking features (long clifflines, lakes, fences, etc). Then, I "find" as many of the old maps features as I can on the LiDAR; >95% of the old cliffs are easy to find, 50% of the boulders are hiding in the LiDAR noise (it's much easier to find something that you know is there than to go the other way), and another 25% of the boulders pop out on the imagery...more if I use a couple of different sets of photos, with different shadows, camera angle, etc.

The contours are about the last thing I update, since even the latest smoothing functions still beg for more hand-smoothing, and I find the overall legibility improves greatly if the brown contours and the black cliffs interplay well. I'll look at the old map for formlines--what was trying to be shown, and how does it look in the LiDAR. And, I'll do the same for other "wiggles" in the old contours. The hillside "shelves" tend to be highly visible in the LiDAR, but the LiDAR fails spectacularly in the flats (as Mikell mentions above).

At the end of all this, all I've really done is shifted around features from the old map, although the locational accuracy has gone up. For the remaining features that I couldn't find, I'll either move them "relationally" to other stuff, or mark them in another color (purple, typically) as a reminder for my field-checking. I will also purple-mark things that LiDAR or imagery suggests might be there but weren't on the previous map.

Typically for the field check map, I hide all the vegetation (maybe leaving some of the medium green, which tends to be relatively stable patches of mountain laurel) and use faint grey 1m contours. I find all this pre-work to be worth the time/effort (vs. just taking out the "old map")...I know what I can "trust" and what may be distorted. I tend to visit/look at the whole area while field checking: making decisions on the items in purple, deciding on the iffy items (boulders which are 1.2m from the low side and 0.8m on the high side, cliff vs. bare rock vs. stony ground, etc), drawing in the vegetation, etc. I also tend to carry an unmodified copy of the old map with me when field checking. I guess you could say my philosophy is it stays the same (as the old map) unless I have clear evidence (LiDAR or imagery and/or field check observation) to make a change.

The original Mylar photogrammetric basemap still exists (I saw it in ~08 when there as faculty)--in some dusty corner of the Geography department. The original maps were all done in discreet sections (the Lake Popolopen map, Bull Pond, Black Rock Forest, etc), but in the OCAD era, they were all stretched/transformed/rubbersheeted and stitched together into a giant map, which became unwieldy and is now four maps. That great stitching introduced some errors and also gave a false sense of continuity--hillsides that were at the edge of maps (and not field checked) are suddenly in the middle of things, beckoning to course setters who don't sense the historical fallacy of "it's a white hillside; it must be good." It also allowed for some great new uses across what were historical "map boundaries."
May 10, 2019 7:03 PM # 
Oh, I forgot to mention the contour stitching. Basically, I follow LiDAR contours along until they and the previous photogrammetric contours are at the same place/orientation. Snip, snip, adjust an end, and voila. Be careful not to inadvertently add a non-existent reentrant or spur in the process.

In the OCAD file, there is plenty of evidence...two contour symbols (one original, one LiDAR)...maybe one of them is lots of point-to-point straight lines and the other is Bezier curves, etc, but printed at scale, you miss it (at least at a can certainly find them with a magnifier).

In the end, my two guiding principles are:
1. Better than it was
2. Good enough

I also pull out old copies of the 5-color "original" maps (pre-OCAD, or at least pre-cadet OCAD)...there have been many hands in the USMAOC OCAD files over the years, often solving one problem and maybe inadvertently creating another, through many versions of OCAD (and symbol changes), a couple of official versions of ISOM, etc. A surprising number of inanimate objects have appeared, disappeared, or marched across the map :)
May 10, 2019 7:13 PM # 
Jon, do you know if the whole Long Mountain map was redrafted from scratch after lidar became available? I originally merged the old Lake Stilwell and Turkey Mountain maps into 0CAD, but I couldn't get everything to line up, I think it all boiled down to an area at the north end of the ridge where the map was stretched, so the paper showed more distance E-W than there was on the ground. Seemed like no matter what you did, getting that fixed was going to be a bunch of work.
May 11, 2019 11:55 AM # 
It was not, but there's an effort to do so (see Jordan comment above...probably not coming to fruition until he's there as faculty in the GIS lab).

Your comment makes sense--I found an area in the far south (near the civilian (?) shooting range marked off in purple overprint) where the old, stitched-up map bore no resemblance to reality. The one identifiable hill was about 150m "off" and there were vague, single contour squiggles that made no sense.

I suspected (and your comment is further supporting evidence, in my mind) that this was an area of "underlap" when the individual maps were rubbersheeted/transformed and stitched together. Because it was on the edge (from the big picture) and marginally useful, I can totally see the logic in faking a few contours and providing "closure" rather than having a empty "Here be Dragons" section of the map.
May 11, 2019 4:31 PM # 
In mapping, the use of the term of "faked" is generally frowned upon and considered unprofessional. The accepted terminology is "under-surveyed", "light pass area", "inferred", or even "interpolated from high quality alternative data" (the latter more often seen in engineering survey applications or the current White House.) Though sometimes you will see refreshing self honesty: "didn't get within a kilometer of that hillside". This may be lidar's biggest advance: the ability for a mapper to now fool all of the people some of the time.
May 11, 2019 5:14 PM # 
I fondly remember sitting with SF somewhere in the northeast many years ago as he filled in the missing corners on the fieldwork of the Florissant Fossil Beds map with contours that were pulled straight out of hammerspace.

But the green blobs in the north part of Lake Frederick? That was just a phone call. "They're putting controls where? That's not good. Umm... just fill it all up with blobs of medium green and I'll deal with it when I get there." Oh Yuck indeed.

Please login to add a message.