[HOT] [info-hotosm] Reference Project #1030 Nepal Earthquake

Kevin Bullock kbullock at digitalglobe.com
Thu May 7 20:28:22 UTC 2015


This is a great thread and I wanted to provide some additional information on behalf of DigitalGlobe:


1)      Our goal in collecting imagery and making it available via open license is to provide as much data as possible given the humanitarian nature of this event. This means we use our satellites in a manner not typically seen. Charlie did a great job summarizing this in his recent blog [1]

2)      Given this, we are actually swiveling our satellites to “point” at Nepal even when our orbits are not directly over Nepal. As an example, a satellite may be vertically overhead Bangladesh, yet, the satellite is looking back at Nepal. This is actually quantified and measured in the image metadata by referencing the Off Nadir Angle[2] and Target Azimuth. In typical circumstances, best accuracy is achieved when Off Nadir angle is less than 20 degrees. In these cases, the ground RMSE is within a few meters [3] . However, the events in Nepal are not considered to by typical circumstances, and in some cases, we are pushing Off Nadir Angles above 40 degrees.

3)      I wanted to confirm that all imagery is indeed ortho-rectified and geo-corrected to the best of our ability considering timeliness and the fact many many people and organizations are waiting for imagery and heavily dependent on its availability. In our orthorectification process, we are leveraging a variety of elevation models. Important to note that most elevation models have linear error that can range from 5-15m. [4] As the off nadir angle increases, these inaccuracies in the elevation model propagate into horizontal displacements in the imagery. This is why we are seeing large offsets.

4)      The tradeoff here is timely, massive amounts of post event imagery acquired under less than ideal circumstances containing horizontal error, or, very limited imagery only collected under ideal circumstances with minimal horizontal error. As noted below, typically, the former is preferred.

Hope this helps, Kevin

[1] - https://www.mapbox.com/blog/nepal-imagery-collection/
[2] - http://www.landinfo.com/buying-optical-satellite-imagery-2.html
[3] - https://www.digitalglobe.com/sites/default/files/WorldView_Geolocation_Accuracy.pdf
[4] - http://www.satimagingcorp.com/services/orthorectification/

From: Steve Bower [mailto:sbower at gmavt.net]
Sent: Thursday, May 07, 2015 10:49 AM
To: Milo van der Linden
Cc: Heather Leson; info at hotosm.org; Ross Taylor; HOT at OSM (Humanitarian OpenStreetMap Team)
Subject: Re: [HOT] [info-hotosm] Reference Project #1030 Nepal Earthquake

Springfield,
You raise important points, and are not "raining on a parade". The resulting data will not be suitable for all purposes, but it can be very useful for this crisis response.

I do think there is significant risk that some mappers will map directly from un-rectified imagery, and introduce problematic location errors. That needs to be minimized, e.g., through clear instructions and good validation. I think there's room for improvement on the instructions, e.g., it would be good to have a wiki page on mapping from un-rectified imagery in combination with rectified imagery, for crisis response.

Thanks

On Thu, May 7, 2015 at 5:14 AM, Milo van der Linden <milo at dogodigi.net<mailto:milo at dogodigi.net>> wrote:
Hello Springfield Harrison,
As a 20 year GIS veteran I understand what you say. I do agree that in communication with first responders it is important to have them clearly understand that the accuracy of features can be off ~100m. But for them having maps that give a good indication is way better then having no maps at all. In the end, and that is what I hope for, it can save lives.
I have a long running discussion with y'olde GIS community on "how can a map created by amateurs be better then what we professionals do?". It is my opinion that it can be. I believe that "the many are smarter than the few" (quote by James Surowiecki). And the HOT tasks have all the ingredients to succeed:
1. There is diversity of opinion
2. People involved in the mapping process have opinions not influenced by those around them
3. People operate decentralized
The only thing that might need more attention (and this is where geospatial experts can take their role) is that HOT and openstreetmap as a whole could use more mechanisms to turn all these little "private judgements" into collective quality. This process could involve analysing quantity and different representations of the same feature through time. In that way, you could see the mapping activity (in dense area's) as GPS. There are faults, influenced by methodology, opinion and conditions. And as a GPS professional, you know that it is _knowing the error_ that automagically creates accuracy. I would love the GIS/GPS community to think about how to know the error in community mapping.
I love this new way of mapping. It creates new opportunities. It involves new ways of thinking. It is not influenced by what GIS people say GIS should be like.
Kind regards, with respect,
Milo


https://en.wikipedia.org/wiki/The_Wisdom_of_Crowds

2015-05-07 10:21 GMT+02:00 Springfield Harrison <stellargps at gmail.com<mailto:stellargps at gmail.com>>:
Hello Steve,

        Sorry to rain on the parade yet again but I find this matter of image alignment to be puzzling and concerning.

        One of the first things I learned when embarking upon GIS/GPS mapping was that accurate georeferencing of all layers, but especially the base layers (imagery in this case) was sacrosanct.  If things are not in their correct point in space, what use is that to the end user?  Especially in rugged terrain, with difficult access and rapidly changing stream flows, it is important to know where a trail or road really is.  Why try to cross a raging torrent when you don't need to?

        Having untrained users realign the imagery willy-nilly is amazing to me.  What faith can anyone have in the new tracings if the earth is literally moving every time a new user opens up the file?  Accurate map datums and projections were created for a reason.

        How is it that, "...the DigitalGlobe 2015-05-03 (DG) images have had minimal georectification.."  This is bizarre, this is not GIS, this is merely sketching.  Why is such imagery being offered and accepted?  I know that this is a major emergency but then all the more need for quality data.

        However, I am newly arrived, and it seems that most people are content with a world that can be up to 200 m out of whack.  I'm not sure if I can contribute much under the circumstances other than this gloomy criticism.  Sorry, will try not to dampen the enthusiasm further.

                 Thanks for your patience, Cheers . . . . . . . . Spring



At 06-05-2015 11:59 Wednesday, Steve Bower wrote:

Ross - If you haven't already, see the recent threads on "data alignment to satellite imagery" and "imagery alignment", in the archives for May:
https://lists.openstreetmap.org/pipermail/hot/2015-May/thread.htmlÂ

Note some links pointed out there by althio:
 http://wiki.openstreetmap.org/wiki/Using_ImageryÂ
 http://learnosm.org/en/editing/correcting-imagery-offset/Â

Because the DigitalGlobe 2015-05-03 (DG) images have had minimal georectification (needed mainly for elevation distortion), they may be offset by 100m or more. On one tile (5.5km wide) I saw offsets relative to Bing of 125m to the west and, elsewhere, 85m to the east. The offsets may vary considerable even in nearby areas, especially in steep terrain.Â

You should align your work with Bing imagery. Thus to digitize from the DG imagery you should first adjust the DG imagery to the Bing imagery, and re-adjust it as you move from place to place. As you noted, adjusting in one area makes it worse in others, so you have to keep re-adjusting as you go. You should be able to compare the Bing and DG imagery to confirm where a feature visible on DG is located on the Bing imagery (if Bing is clear enough). I try to adjust based on buildings, or road intersections/curves (keeping in mind that roads are sometimes relocated), or even less permanent features (rivers generally are not good, they move around to much). It's a time-consuming process, but needed to correctly locate features.

It's not essential that everything be within a few meters of its true location, but features should be mapped correctly relative to one-another.

The links above provide guidance on how to align imagery to correct locations. It's easy in JOSM with the Imagery Offset tool (on the toolbar).

Steve

On Wed, May 6, 2015 at 1:53 PM, Steve Bower <sbower at gmavt.net> wrote:
I don't think Chad's IDP guidance document (though very helpful) addresses the issue of spatial accuracy of the DG imagery, raised by Ross. I'm going to post that as a separate issue with more detail.
On Wed, May 6, 2015 at 4:35 AM, Heather Leson <heather.leson at hotosm.org > wrote:
HI Ross, sorry for my delayed response. It is best if you ask your questions on the main Hot at openstreetmap.org mailing list.
Chad provided this guidance document on IDPs http://hotosm.github.io/tracing-guides/guide/Nepal.html#IDP%20Collection%20Guidance
Hope this helps
Heather
On Tue, May 5, 2015 at 12:40 AM, Ross Taylor <ross at byrdtechnology.com > wrote:
Hi, I am seeing many more IDP sites using DigitlaGlobe imagery vs Bing. I can toggle between the two image sets, but they are significantly nonaligned. I created a landuse=brownfield tagged area which aligns with Bing, but if I mark and tag the individual IDP sites showing up in DigitalGlobe imagery, the brownfield and idp are not going to be aligned. I want to help out as much as possible and would like the data to be correct. Please advise, thanks!
Note: I tried to adjust alignment but it fixes one area and creates more offset in other areas.
-Ross
Sent from mobile

_______________________________________________
HOT mailing list
HOT at openstreetmap.org
https://lists.openstreetmap.org/listinfo/hot


_______________________________________________ HOT mailing list HOT at openstreetmap.org<mailto:HOT at openstreetmap.org> https://lists.openstreetmap.org/listinfo/hot

_______________________________________________
HOT mailing list
HOT at openstreetmap.org<mailto:HOT at openstreetmap.org>
https://lists.openstreetmap.org/listinfo/hot


--
[Image removed by sender. http://www.dogodigi.net]<http://www.dogodigi.net>

Milo van der Linden
web: dogodigi<http://www.dogodigi.net>
tel: +31-6-16598808<tel:%2B31-6-16598808>



This electronic communication and any attachments may contain confidential and proprietary information of DigitalGlobe, Inc. If you are not the intended recipient, or an agent or employee responsible for delivering this communication to the intended recipient, or if you have received this communication in error, please do not print, copy, retransmit, disseminate or otherwise use the information. Please indicate to the sender that you have received this communication in error, and delete the copy you received.

DigitalGlobe reserves the right to monitor any electronic communication sent or received by its employees, agents or representatives.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/hot/attachments/20150507/391e8bec/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ~WRD000.jpg
Type: image/jpeg
Size: 823 bytes
Desc: ~WRD000.jpg
URL: <http://lists.openstreetmap.org/pipermail/hot/attachments/20150507/391e8bec/attachment.jpg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.jpg
Type: image/jpeg
Size: 413 bytes
Desc: image001.jpg
URL: <http://lists.openstreetmap.org/pipermail/hot/attachments/20150507/391e8bec/attachment-0001.jpg>


More information about the HOT mailing list