From: "Philip Matthews" <philip_matthews@qmail.Newbridge.COM>
Date: 23 Nov 1994 17:31:28 -0500
Subject: Re: mapping from single aer
Message-Id: <9411232233.AA02685@Newbridge.COM>
RE>>mapping from single aerial ph
A number of people have recently noted the problems of making a
quick-and-dirty map
by making a basemap directly from a single aerial photo (without doing the
necessary
corrections).
What I would like to know is: how accurate can one get with multiple
photos?
Background: Where I live, it is pretty easy (and cheap) to get photos with
the 60% overlap required
to make a good photogrammetric basemap. Moreover, the terrain here (in the
Ottawa
area) is divided into two types: hilly with lots of water and rock
features, and flat with
a lots fewer features (no rock, for example). Needless to say, the former
tends
to be a lot more interesting orienteering-wise, but the latter is nearer to
most of the
population. So the Ottawa club has maps over both types.
In a perfect world, we would spend the $1000+ required for a base map even
when doing
maps of type 2. However, finances are always tight, so we tend to cheat.
- Philip Matthews, Ottawa Orienteering Club
--------------------------------------
Date:
To: Philip Matthews
From: Pat Dunlavey
OK, OK. I know several have written on this, but I can imagine that people
are still confused, so I'll take a stab at it.
First of all, it's not exactly true that the geometry of aerial photography
makes it impossible to have an aerial photo that registers accurately with
a map of the same area. IF the ground is perfectly flat AND the axis of
the camera is perfectly vertical, then you will indeed get perfect
registration (after scale, rotation and x/y position have been taken into
account, and ignoring the factors of geographic projection, earth curvature
and light refraction in air, which is safe at the map scales we're dealing
with). Of course, this is not normally the case! Though the camera used
for mapping usually tries to be vertical, and some modern aerial cameras
can actively compensate for aircraft tilt, it can never be perfect.
However, usually the much greater distortion occurs due to what is called
radial distortion. That is where the farther you are away from the center
of the photo (the principle point), the greater the effect elevation (Z)
will have on apparent horizontal position (X/Y). To illustrate this,
imagine yourself looking down from directly over a very tall building. The
top of the building would be lined up with the base of the building. Now
imagine that instead, the building is far away to the right. You would
still see the top, but it would no longer be lined up with the base of the
building, it would be displaced further to the right. This is very obvious
when you see buildings and trees this way, but it's not so obvious when
it's the undulations of terrain that introduce the change in Z.
This radial distortion, far from being an impediment, is actually necessary
to make measurements of elevation from aerial photos. Most mapping is done
from wide-angle photography, which has more radial distortion than
narrow-angle photography. However, it also means that simple tracing of
accurate information from aerial photos is not generally feasible. Some
GIS systems support digitizing of information from single aerial photos by
"rubber sheeting" (using the same technology as morphing) the aerial photo
to fit a number of points digitized from the map. Another way is to use
digital elevation models in combination with the photos, and run the math
backwards. I don't think either of these techniques are suitable for
orienteering mapping purposes due to our need to have very good relative
accuracy (the information taken from the photo needs to line up accurately
with the information taken from the map).
So, to sum it up. If you're mapping a very flat area, and you don't need
great accuracy, tracing from a single aerial photo can be useful.
Otherwise, don't bother.
-- Pat
Pat Dunlavey pdunlave@williams.edu
{dba Pat Dunlavey Cartographics, TopoMagic}
40 Oblong Road, Williamstown, MA 01267
voice (413) 458-9836 fax (413) 458-9273
------------------ RFC822 Header Follows ------------------
Received: by qmail.newbridge.com with SMTP;23 Nov 1994 13:11:42 -0500
Received: from nbkanata.Newbridge.COM (nbkanata-gw) by Newbridge.COM
(4.1/SMI-4.0)
id AA22038; Wed, 23 Nov 94 13:06:22 EST
Received: from ciocc.graphics.cornell.edu by nbkanata.Newbridge.COM
(4.1/SMI-4.1)
id AA03530; Wed, 23 Nov 94 13:06:19 EST
Received: by ciocc.graphics.cornell.edu (5.65/DEC-Ultrix/4.3)
id AA20835; Wed, 23 Nov 1994 10:42:02 -0500
Received: by merckx.graphics.cornell.edu (5.65/DEC-Ultrix/4.3)
id AA10631; Wed, 23 Nov 1994 10:42:02 -0500
Received: from hancock (hancock.cc.williams.edu) by rowe.williams.edu with
SMTP id AA26809
(5.65c/IDA-1.4.4 for <orienteering@graphics.cornell.edu>); Wed, 23 Nov
1994 10:41:45 -0500
Received: from DialupEudora (hancock) by hancock (4.1/client-1.3)
id AA07284; Wed, 23 Nov 94 10:41:26 EST
Date: Wed, 23 Nov 94 10:41:22 EST
Message-Id: <9411231541.AA07284@hancock>
X-Sender: pdunlave@popserver.williams.edu
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
To: orienteering@graphics.cornell.edu
From: Patrick.D.Dunlavey@williams.edu (Pat Dunlavey)
Subject: Re: mapping from single aerial photos
Sender: orienteering-owner@Graphics.Cornell.edu
Precedence: bulk
From: Patrick.D.Dunlavey@williams.edu (Pat Dunlavey)
Date: Wed, 23 Nov 94 10:41:22 EST
Subject: Re: mapping from single aerial photos
Message-Id: <9411231541.AA07284@hancock>
OK, OK. I know several have written on this, but I can imagine that people
are still confused, so I'll take a stab at it.
First of all, it's not exactly true that the geometry of aerial photography
makes it impossible to have an aerial photo that registers accurately with
a map of the same area. IF the ground is perfectly flat AND the axis of
the camera is perfectly vertical, then you will indeed get perfect
registration (after scale, rotation and x/y position have been taken into
account, and ignoring the factors of geographic projection, earth curvature
and light refraction in air, which is safe at the map scales we're dealing
with). Of course, this is not normally the case! Though the camera used
for mapping usually tries to be vertical, and some modern aerial cameras
can actively compensate for aircraft tilt, it can never be perfect.
However, usually the much greater distortion occurs due to what is called
radial distortion. That is where the farther you are away from the center
of the photo (the principle point), the greater the effect elevation (Z)
will have on apparent horizontal position (X/Y). To illustrate this,
imagine yourself looking down from directly over a very tall building. The
top of the building would be lined up with the base of the building. Now
imagine that instead, the building is far away to the right. You would
still see the top, but it would no longer be lined up with the base of the
building, it would be displaced further to the right. This is very obvious
when you see buildings and trees this way, but it's not so obvious when
it's the undulations of terrain that introduce the change in Z.
This radial distortion, far from being an impediment, is actually necessary
to make measurements of elevation from aerial photos. Most mapping is done
from wide-angle photography, which has more radial distortion than
narrow-angle photography. However, it also means that simple tracing of
accurate information from aerial photos is not generally feasible. Some
GIS systems support digitizing of information from single aerial photos by
"rubber sheeting" (using the same technology as morphing) the aerial photo
to fit a number of points digitized from the map. Another way is to use
digital elevation models in combination with the photos, and run the math
backwards. I don't think either of these techniques are suitable for
orienteering mapping purposes due to our need to have very good relative
accuracy (the information taken from the photo needs to line up accurately
with the information taken from the map).
So, to sum it up. If you're mapping a very flat area, and you don't need
great accuracy, tracing from a single aerial photo can be useful.
Otherwise, don't bother.
-- Pat
Pat Dunlavey pdunlave@williams.edu
{dba Pat Dunlavey Cartographics, TopoMagic}
40 Oblong Road, Williamstown, MA 01267
voice (413) 458-9836 fax (413) 458-9273
From: soft-forlaget@rf.se
Date: Wed, 23 Nov 1994 14:36 +0100
Subject: Swedish team 1995
Message-Id: <199411231337.OAA19439@mail.swip.net>
Swedish team for 1995 has been selected.
Men
Jimmy Birklin
Per Ek
H}kan Eriksson
Lars Holmqvist
Joakim Ingelsson
Johan Ivarsson
Martin Johansson
J|rgen M}rtensson
Andreas Rangert
Arto Rautiainen
Women
Anna Bogren
Katarina Borg
Elisabeth Drotz
Maria Gustafsson
Marlena Jansson
Anette Nilsson
Gunilla Sv{rd
Annika Zell
OL-regards
Lasse
From: "P.M.Todd" <P.M.Todd@sheffield.ac.uk>
Date: Wed, 23 Nov 1994 13:53:31
Subject: JK95 dates
Message-Id: <108783E6058@Rivelin.shef.ac.uk>
The dates given in the Digest#169 for JK 95 were wrong.
Apparrently the entry form was proof-checked by 10 people and still
the mistake wasn't spotted! The dates are (Good) Friday 14th April
to (Easter) Monday 17th April with the two individual days on the
Saturday and the Sunday.
Hope to see you there
Paul Todd+------------------------------+----------------------+
| Name : Paul Todd | |
| Email: P.M.Todd@shef.ac.uk | Witty comments |
| Snail: 1D16, Tapton Hall, | to be added here |
| Crookes Rd, Sheffield | |
| S10 2AZ | |
+------------------------------+----------------------+
From: Larsg.orientering@rf.se
Date: Wed, 23 Nov 1994 13:35 +0100
Subject: 10-mila
Message-Id: <199411231236.NAA05341@mail.swip.net>
Somebody asked about 10-mila next year.
In the last issue of Skogssport (9/94) there is an article about 10-mila
-95.
There are some changes in the legs length etc. The start is at 22.00 hours.
The first is leg unforked!
Men
leg 1 12,7 km night
2-3 12,1-12,2 km night
4 9,0-9,1 km night
5 16,4-16,5 km night
6-7 6,5-6,6 km dawn/day
8-9 11,6-11,7 km day
10 13,9 km day
First and 10th legs are unforked.
Women
Start at 15.00 hours. All legs in daylight.
leg 1 6,5 km
2-3 6,7-6,8 km
4 3,8 km
5 8,1 km
Only leg two and three are forked.
Next year the 10-mila celebrate 50 years anniversary.
Lars
From: roberts@doublon.unice.fr (Tony Roberts)
Date: Wed, 23 Nov 94 12:13:42 +0100
Subject: mapping from single aerial photos
Message-Id: <9411231113.AA06388@vala.unice.fr>
>
>> .... Then I scanned in the aerial photo to get buildings, veg.
>> boundaries, and the few visible trails. Unfortunately, despite
>having the
>> scale right, the two sources did not match up well. I had to
>re-zero the
>> aerial photo at different areas of the map to trace features.
>Lining up one
>> feature would not allow lining up another feature on another
>portion of the
>
>The geometry of an aerial photo is such that a simple approach such
>as this will never give a perfect fit. The photo has a central
>perspective (scene viewed from a single point) not the orthogonal
>perspective (scene always viewed from above) of a map. Also any
>relief in the ground will result in local scale distortions in the
>photo. Those factors are some of the reason why for "proper" base
>maps a stereoscopic pair of photos are used.
>
More distortion factors that are taken into account are typical atmospheric
distortions in light-ray paths, and known defects in the camera lens that
takes the photograph.
Tony :-{)}
------------------------------------------------------------------------------
Professor A.J. Roberts
Currently visiting: Institut de Mecanique Statistic de la Turbulence
and: Institut Non Lineaire de Nice
Dept of Mathematics & Computing E-mail: aroberts@usq.edu.au
University of Southern Queensland Phone: (076) 312943
Toowoomba, Queensland 4350 Fax: (076) 312721
AUSTRALIA WWW: ftp://ftp.usq.edu.au/pub/aroberts
------------------------------------------------------------------------------
From: einar.teslo@nlh10.nlh.no (Einar Teslo)
Date: Wed, 23 Nov 1994 11:23:29 --100
Subject: Training in New Zealand
Message-Id: <9411231023.AA07488@nlh10.nlh.no>
I shall be in New Zealand from 5. December to 8. March. I shall be
stationed at Lincoln University, Christchurch, and I am going to spend much
of my time in that region. But there will also be some travelling in New
Zealand.
Possibilities of training with maps and of participating in competitions,
and meeting O-people in New Zealand, are of great interest. I would be very
grateful if someone could give me information and/or addresses.
With O-regards
Einar Teslo
(E-mail address: itfete@itf.nlh.no)
Einar
From: "Hugh Buchanan" <Hugh.Buchanan@newcastle.ac.uk>
Date: Wed, 23 Nov 1994 09:02:18 BST
Subject: mapping from single aerial photos
Message-Id: <MAILQUEUE-101.941123090218.256@BULLEN>
> .... Then I scanned in the aerial photo to get buildings, veg.
> boundaries, and the few visible trails. Unfortunately, despite
having the
> scale right, the two sources did not match up well. I had to
re-zero the
> aerial photo at different areas of the map to trace features.
Lining up one
> feature would not allow lining up another feature on another
portion of the
The geometry of an aerial photo is such that a simple approach such
as this will never give a perfect fit. The photo has a central
perspective (scene viewed from a single point) not the orthogonal
perspective (scene always viewed from above) of a map. Also any
relief in the ground will result in local scale distortions in the
photo. Those factors are some of the reason why for "proper" base
maps a stereoscopic pair of photos are used.
It is sometimes possible to get orthophotos, which are single photos
which have had these distortions removed. These are a processed form
of the original photographic image, and are therefore more expensive.
Hope this helps to clarify.
*-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-*
Hugh Buchanan - Department of Surveying Phone: +44(0)191-222-6351
University of Newcastle upon Tyne Fax: +44(0)191-222-8691
Newcastle upon Tyne, NE1 7RU, UK Email: hugh.buchanan@newcastle.ac.uk
*-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-* *-*-*-*
From: Scott Donald <ua145@ciao.trail.bc.ca>
Date: Wed, 23 Nov 1994 00:51:02 -0800 (PST)
Subject: Re: O-net Digest V3 #169
Message-Id: <Pine.3.89.9411230011.A28498-0100000@ciao>
Photo distortion -
Visualize a rectangle of flat ground with the camera lens directly above the
middle of the rectangle. Place a measuring stick on the ground, then
trace to light rays from the ends of the stick to the center point of the
lens when the stick is directly below the lens, and when the stick is at
one edge of the patch of ground. In both positions, keep the axis of the
stick on a line which passes through the center of the patch. You now
have two triangles (from the center point of the lens to one end of the
stick, along the length of the stick, then back to the center point of
the lens along the other light ray.
Now, draw a straight line across the triangle, near to the stick end, in
such a fashion that the line just drawn crosses the sides of the triangle
at the same angle at both sides. I think I have said it correctly - where
are the pictures, when you need a thousand words! :-)
When the center of the stick is under the center of the camera lens, the
line you have just drawn will be parallel to the ground, and the camera
"sees" a picture of limited distortion.
At the edge of the patch, the line will be at a substantial angle to the
stick, and will be much shorter. A the line is the "picture" which the
camera sees.
To further complicate matters, take the stick at the edge of the patch
and rotate it so the axis of the stick is exactly 90 degrees to a line
joining the center of the stick the the center of the patch and repeat
the process of ray and line drawing. NO DISTORTION!
Mind you, not everything I have said above is quite right, but it is a
good approximation, and give you a feel for the types of distortion you
can expect when trying to transcribe air photos without the optical
equipment designed to correct the distortions.
I hope this helps.
Scott
From: "Sundquist, Janet" <JSundqui@ubmedd.buffalo.edu>
Date: Tue, 22 Nov 94 23:19:00 PST
Subject: O-net Digest V3 #169
Message-Id: <2ED2ED91@ubmed.buffalo.edu>
>Thanks to Pat, Scott, and Forest for the info. on file acquisition. Don't
>know where I stand now but will get on with something this winter. I have
>two areas I want to map for next year. Here is what I will do. Please
>comment. I will order an aerial photo and have a print made to the size
>map that I want. I will have a picture made of the 1:24,000 (for contours)
>blown up to match the photo. I will trace the two on the copy table and
>hope this will give me a base map that can go on OCAD so field checking can
>start. Sounds a little old fashioned but that is how I have made the five
>other maps we now use. Contours are usable sometime and need a lot of
>adjusting from field check. Photos are 1960 models and come from the
>Agriculture department. Some major features are current but little else.
>I place all trails by field check and they are, at best, close. If all
>fails, I will copy the maps we now use. They work and will look real
>pretty in OCAD. Sure wish the trails were in the correct place though.
>For B and C meets, I guess it will have to do. Again, thanks for the help
>and remember, "when in doubt, do the right thing". Jim <H:)
>
This may not work; at least not with the map I'm working on now. I'm making
a map for the 1995 New York State Ski Racing Assoc. (NYSSRA) Ski-O champs
this February. As I was asked to run this only this fall, I decided to make
the map using USGS quads and an aerial photo, plus lots of field checking.
I'm using OCAD and first I scanned in the USGS quad and traced the contours,
roads, etc. Then I scanned in the aerial photo to get buildings, veg.
boundaries, and the few visible trails. Unfortunately, despite having the
scale right, the two sources did not match up well. I had to re-zero the
aerial photo at different areas of the map to trace features. Lining up one
feature would not allow lining up another feature on another portion of the
map. Again, this was not a scale issue, I checked that. The aerial photo
was from the USDA Soil Conservation service, 8 inches to a mile (if I
remember correctly). A beautiful photo (taken, as Pat Dunlavey pointed out,
with everything in full bloom), but evidently there was some distortion or
something that prevented it from being perfectly orthogonal. Of course, I'm
assuming the USGS quad was correct, and it's scan was non-distorted, which is
not necessarily true, but I had to settle on on some datum, so I chose to go
with the USGS (hey, I already traced all those contours!)
Another comment while making this map: A couple of months ago, there was a
post describing how well drawn grid lines were digitized into OCAD with an
error of a couple of percent (i.e., instead of showing up as 10 mm, they were
10.1 mm, for example). A response blamed this on the digitizing tablet.
However, with my new map, I added 10 mm grid lines to my original base map
(the scanned quad) using the image editing program that came with the
scanner. I believe thsi makes the marks as accurate as you can get them.
However, when I imported this as a template, I had to set the OCAD grid lines
(i.e the true map grid lines) at 24.2 mm rather than the expected 24 mm (the
1:24,000 quad is used to make a 1:10,000 map, thus the expected 24 mm). This
seems to be a similar problem as was noted before, and it obviously isn't a
digitizer problem in this case. It isn't a scanner problem either,
sinve the lines were drawn after scanning, and the scales were set directly
by the "scales" menu item. However, if this is the worst problem with OCAD,
I can live with it. Overall, I think it is a powerful and well designed
program.
Jon Sundquist