PDA

View Full Version : Calibrating aerial photos for GPS



Drop Bear
17th October 2005, 05:03 PM
Ive been playing around with google earth which is really good (i prefer it over the nasa job as its better detail). Ive made myself a map of a favourite wheelin area of mine by printing screen, cropping, and then pasting into one large map. I have attempted to calibrate this map into Oziexplorer with limited success. Ive tried to use 3 calibration points entering in the long and lat of road intersections etc based upon other maps i have of the area that work and are spot on when i use my GPS with them but its still off track by about 100metres south and west. Anyway to cut a long story short has anyone had any success converting aerial photos to a usable map in Ozi?

mitchofsutho
18th October 2005, 08:04 AM
Ive been playing around with google earth which is really good (i prefer it over the nasa job as its better detail). Ive made myself a map of a favourite wheelin area of mine by printing screen, cropping, and then pasting into one large map. I have attempted to calibrate this map into Oziexplorer with limited success. Ive tried to use 3 calibration points entering in the long and lat of road intersections etc based upon other maps i have of the area that work and are spot on when i use my GPS with them but its still off track by about 100metres south and west. Anyway to cut a long story short has anyone had any success converting aerial photos to a usable map in Ozi?


Drop Bear,

It is possible that the track coordinates you have are based on AGD66 Datum and map coordinates are set at WGS84 (GDA94) Datum. When we went to GDA94 (near enough to WGS84), the equivilent AGD66 coordinates moved to a North Westerly direction.

To convert GDA94 to AGD66 you need to decrease Northings by 179m and Eastings by 122m.

OziExplorer should automatically carryout the conversions if you have configured it correctly.

Mitch

Drop Bear
18th October 2005, 11:04 AM
Thanks Mitch, im checking the datum now in oziexplorer, i have it set to Australian Geodetic 1966 (AGD66 i presume?) Originally i had it set on WGS84 which also wasnt any good. Now ive set it to GDA94 and is still off by about 100 metres. I also have the map projection set at UTM.

Ill explain a little of how i calibrated the map just to be sure im doing it correctly. Using the topoview maps i have, i have written down the long and lat (im not using eastings and northings) of visable landmarks (road intersections etc) and then when i calibrate the aerial photo i simply locate these landmarks and enter in the co-ordinates. I did this with 3 points scattered across the photo. When i check the co-ordinates of the points i marked they are spot on but as soon as i check another landmark the co-ordinates are off by 100 metres or so. What am i doing wrong?

Cheers, Ben.

festy
18th October 2005, 02:09 PM
Try turning on the map grid. If the grid isn't pretty much square, then your calibration points are probably not quite right.

Moo73
18th October 2005, 03:57 PM
I've been having the same trouble, thanks Mitch

Drop Bear
18th October 2005, 04:53 PM
Festy, i double checked the co-ordinates and they appear to be correct, the "other" grid showing metres etc is not straight. Does this mean then that the map isnt positioned north? do i need to rotate the image file to fix things up?

The positioning is off by about 50m south and 100m west.

Thanks for the input people!

jindydiver
19th October 2005, 08:35 AM
Your problem isn't datum, it is projection. IIRC Clark is the projection you need for aerial photos, and it may work better for you with the sat maps.

Are the maps you are trying to calibrate big. Is the distance between the position at the top left corner and the position at the bottom right corner measured in 10's of k's or 100's, or 1000's of k's. The bigger the area the bigger the distortions towards the edges of the map.

mitchofsutho
19th October 2005, 11:04 AM
Drop Bear,

I was just about to suggest that the projection is probably out, due to the distortion of the photograph, but Jindydiver beat me to it.

Drop Bear
20th October 2005, 06:20 AM
Thanks guys, ill try this tonight. The aerial shot is approx 10k's from top left corner to bottom right.

jindydiver
20th October 2005, 01:08 PM
You might find it will work better if you can get more calibration points into it and use the polynomial calibration option.
I have done plenty of aerial photos, and the more points you can get in there the better it gets.
good luck with it.

Drop Bear
20th October 2005, 04:06 PM
There doesnt appear to be a "clark" projection in the drop down list, could it be something else? I will try using more points aswell. I have checked the calibration of the topoview maps all they have is 2 points, one point is the very top left corner of the image (0,0) and then the next point at 1000,1000 (image co-ordinates) using the UTM projection. It is essentially a scan of a paper map. How distorted can the aerial photo be? I gather that no one has tried doing this before (using google earth for the image) as ive searched google and havent found anything about doing this. Anyway ill try to include as many points as i can and see if that improves it.

bluezooky
20th October 2005, 05:06 PM
.

Ill explain a little of how i calibrated the map just to be sure im doing it correctly. Using the topoview maps i have, i have written down the long and lat (im not using eastings and northings) of visable landmarks (road intersections etc) and then when i calibrate the aerial photo i simply locate these landmarks and enter in the co-ordinates. I did this with 3 points scattered across the photo. When i check the co-ordinates of the points i marked they are spot on but as soon as i check another landmark the co-ordinates are off by 100 metres or so. What am i doing wrong?

Cheers, Ben.

Let me suggest a quicker way than manually entering co-ordinates:
Load you topomap, create the waypoints you want to use as calibration points and save the waypoint file.

Leaving the waypoints loaded open calibration window,select your calibration point and click on the Wp icon and click on the appropriate waypoint.

Save the map file.

Regards Charlie

jindydiver
20th October 2005, 07:07 PM
drop bear

When I first start trying to make the aerial pics fit I went out to Natmap and spoke to a couple of guys there about my maps. My problem was that the pics are of farm land that is now under lakes, and it is not posible to "stand" at a land mark and check the calibration.
These guys explained to me that the best way to get the pics calibrated was to make my own grid. It took some imagination but I managed to get all 9 points needed to get a good fix with Ozi.

When you put in your two wp's and save your map how do you know that an error exists elsewhere on the map. The answer i assume is you tested it by visiting the site with a gps and seeing that where the gps says you are on the map isn't where you are standing. Would this be right?
If it is then when you know where you are standing, and you can see it on the map, you should take a wp and use this in your calibration. Or if you are just comparing landmarks from your topo to your pic then just use those as calibration points too. And go back and check all your calibration waypoints to make sure you have entered them properly.
The method Bluezooky explained is the best way to avoid keystroke errors.

bushie
20th October 2005, 08:35 PM
I doubt that you will get a very satisfactory calibration with aerial photos, there is some pretty powerful maths needed to rectify them.

Scale on airphotos is not consistant across the photo and varies with the terrain height as well.

If you can imagine the front of the lens in the camera then the corners of the photo are the furthest away (and hence smallest scale) directly below the camera is closest (theoretically) and a different scale again and terrain/elevation throws another variable into the mix.

Couple this with yaw, roll and pitch on the aircraft it all combines to make calibration a nightmare.


Bushie

Drop Bear
21st October 2005, 09:00 AM
Thanks again for all the input! I have now used 9 calibration points and its much better than before, not perfect but better, im only out by about 10m off to the east now which is pretty good considering. I will go to all the landmarks i have used and get an actual reading from the gps to double check that they are right using the waypoint idea bluezooky suggested.

Also, arent google earth images taken from satellite photos not from aircraft?

bushie
21st October 2005, 05:32 PM
Google earth uses a mix of both satellite and aerial photos (at least thats what two separate GIS managers have told me).

Bushie

Drop Bear
22nd October 2005, 03:50 PM
I think you may be right. I have been accessing the photos via google maps, i just downloaded google earth as another area wasnt available on google maps in good detail, but it is with google earth, i cant match these shots up together like i did with the google maps images, id say these are aerial photos and google maps is actually satellite photos.

jindydiver
23rd October 2005, 05:56 AM
Does anybody know if there is some way to get the geo referencing info out of the google images. it must be in there somewhere to have co-ords under the curser?

MarkkyMarkk
25th October 2005, 04:01 PM
Just thought i'd make a few notes on this thread to muddy the waters...

I have a feeling that the geo-referencing info on google earth is embedded in the images. I may be wrong, but my guess would be that the google earth images are in .ecw format which is highly compressed compared to jpg's of tiff's and tiled together seamlessly. (Although you can see the joins where the higher res imagery is blended into to lower res stuff.)
I know from experience (i'm a GIS manager) that ecw's have the info embedded inside them, cause you can open them with MrSIDviewer & get coords. This is different to the way most other raster images are geo-referenced - using a separate file to list the coords vs pixels.
ie the .map file in oziExplorer, the .tab file in MapInfo, the .sfw file in ArcView, etc...

Also, aerial photography for mapping purposes goes through a process called ortho-rectification which eliminates the perspective & restores the scale to a constant across the image - So that they can be used as a map in a horizontal plane. However, small errors will still persist due to the variations in height across the land in the image, which is why you'll get a better result if you use more control points in the polynomial solution.

Just my 2c worth...

MarkT.

jindydiver
25th October 2005, 06:01 PM
If the pics are ecw then you should be able to find a way to save them, and then you can import them directly into OziExplorer.