Over the fields.Terminated by fully automated approach loop, as on all my flights (no reason to get nervous anymore).Used 7 waypoints for landing loop and 3 waypoints for the mission itself (overhead 200m agl, remote fixed position at 200m agl, overhead 200m agl then landing).Shoting interval 4.4s, Sony Webbie (that long shoting interval is uncomfortable at lower altitudes for planes flying around 50km/h - photo overlap suffers).Used around 50 photos out of 120 taken (only one direction, all resized down),all in order to save processing time.
I am fully aware of stitching software limitations. This is a different class than GIS software. This time I have obtained results good enough for amateurs and I have posted them.
I used transparent color, downgraded to 256 colors, imported in GE then resized to my taste.
Reason: even CNES imaginery in my area has apparently 25m precision (misplaced road etc).
You cannot get too much precision when importing overlay bitmap to GE, anyway.
This 3km map is barely fitting and has 40m errors at both ends, (what from another point of view is quite good considering it is 60 images long).
When you overlay your mosaics in google earth, do you simple use the overlay tool and make the image slightly transparent and then attempt to match underlying features by re-scaling the mosaic? Or do you have a more sophisticated method?
If you're interested in geolocalizing (or georeferencing) it does not require a full-out GIS suite- I've used a georeferencing program that's part of an open source suite called MapWindow. I've actually referenced some of the pictures I've taken from my plane with this program. I use GE to get the lat/long positions of easily found features in my pictures and input those values as well as precisely pick the point on the photo. I've not done this yet with multiple 'near' photos, but it would probably have a similar 'stitching' effect
There is a google earth community where you can request adding your map to the global pool, but at the end if satellite imaginery covers half of the country and is 4 years old, they always use sat in order to have better consistency.
So indeed it must be stored on personal servers as I cannot imagine Google storing all possible small maps from the world for free - they don't have a business model for that task I think.
PTGUI can handle as much as you like (theoretically some 64Kx64KPIX), but 120 5MPIX photos are cooking since 12h without much success (600M B ram when raw...).
The largest obtained in practice is some 20KPIXx20KPIX.
4.4s is one shot every 4.4s, typical for most cheap cameras
with wides field of view, at 121m it creates something like 85x60m on the ground (don't remember now exactly).
This is by eye 19% overlap at 40km/h at most (a photo every 4.4s*11m/s=every 48.4m, (60-48.4)/60*100%=19%) but if you add some wind here or there plus roll plus yaw...
Very cool. I don't have a camera that will do sequential photos so my idea was to use a Flip video camera aimed down and trim the resulting video to maybe 100 pixels tall. Then I was going to output the lot of them as a numbered sequence. Can PTGUI handle 3-4000 narrow images?
Comments
Reason: even CNES imaginery in my area has apparently 25m precision (misplaced road etc).
You cannot get too much precision when importing overlay bitmap to GE, anyway.
This 3km map is barely fitting and has 40m errors at both ends, (what from another point of view is quite good considering it is 60 images long).
3km is here:
http://www.aerialrobotics.eu/examples/LineMap-s1.png
The image is 256 colors because I wanted to have transparent border.
There is a google earth community where you can request adding your map to the global pool, but at the end if satellite imaginery covers half of the country and is 4 years old, they always use sat in order to have better consistency.
So indeed it must be stored on personal servers as I cannot imagine Google storing all possible small maps from the world for free - they don't have a business model for that task I think.
The largest obtained in practice is some 20KPIXx20KPIX.
4.4s is one shot every 4.4s, typical for most cheap cameras
with wides field of view, at 121m it creates something like 85x60m on the ground (don't remember now exactly).
This is by eye 19% overlap at 40km/h at most (a photo every 4.4s*11m/s=every 48.4m, (60-48.4)/60*100%=19%) but if you add some wind here or there plus roll plus yaw...
BTW stitching software is PTGUI.
This is south-west Poland, flatland in all directions, the snow as you see has totally melted,
compare to the mapping made a week ago:
http://www.rcgroups.com/forums/showpost.php?p=13926816&postcount=5
http://www.aerialrobotics.eu/examples/WinterPhotomapping-Large.jpg