I hesitate to ask since it may turn out that the Romans had a solution, but I'm looking at some 2012 images from the National Agriculture Imagery Program and am wondering how others deal with the uncertainties in the pixel coordinates. The ultimate object is to extract positions of the banks of a river that show up quite well in the imagery and which would be good enough for certain purposes.
The images have 1m pixels and the horizontal accuracy of the rectification is claimed (in the accompanying metadata) to be "to a horizontal accuracy of within +/- 5 meters of reference digital ortho quarter quads DOQQ's) from the National Digital Ortho Program (NDOP)" As I interpret that specification, it is just a statistical measure of the residuals of image coordinates when compared to DOQQ coordinates, not a statement about the absolute uncertainty of the positions of pixel coordinates in the image.
I'm thinking that getting actual survey coordinates of image features is warranted to test/verify the accuracy of the images and possibly to do some secondary rectification of the images. Is there a better way to improve/verify the accuracy of the NAIP images that the Romans knew about?
> I'm thinking that getting actual survey coordinates of image features is warranted to test/verify the accuracy of the images and possibly to do some secondary rectification of the images. Is there a better way to improve/verify the accuracy of the NAIP images that the Romans knew about?
As it turned out, just dropping the boundary of a large tract recently surveyed in NAD83 coordinates onto the DOQQ showed really no discernible discrepancies where photo-identifiable features could be compared. The agreement was probably closer to +/-1m s.e. or less, which is plenty good. The formal accuracy statments that accompany the DOQQ may be very conservative.
That's interesting.. Are you inserting these under your surveys regularly? Can you share the source for the imagery?
ANdy
Andy,
I don't know about Kent...
I use USDA-NRCS Geospatial Data Gatewayto download the NAIP images.
DDSM:beer:
Kent,
I flew and acquired quite a bit of NAIP photography about a decade ago, including coverage from Austin south to the border and to Houston on the east and Del Rio on the west. We were based on NAD83-CORS and used an Applanix POS/DG system which when the GPS/IMU was post-processed positioned the camera at about the decimeter level. Other standard photogrammetric error budgets effect the ultimate ground accuracy. The ground checking we did showed we far exceeded the minimum specifications for this orthophoto scale, and was surprisingly good. Flights were 20,000 feet above ground. The specification is the standard, as you are seeing you may well have imagery that exceeds spec.
> That's interesting.. Are you inserting these under your surveys regularly?
> Can you share the source for the imagery?
The NAIP imagery I've got was for Texas, downloaded from the Texas Natural Resource Information System website. Most of it is 1m resolution and they also have the 0.5m resolution DOQQ imagery available for free, too.
I'm looking at using the images to map the meanders of the centerline of a waterway that would otherwise be very difficult to survey. After dropping the centerline of a segment of the waterway that I surveyed on the ground a couple of years ago onto the imagery, I discovered that the answer was so close to what I would have digitized from the imagery that for practical purposes of acreage calculations the answers would have been essentially the same. It's a riparian boundary, so the purpose of the meanders is for acreage rather than exact location of the boundary for all time.
I'll still survey a few photo identifiable features just to document the quality of the imagery, but there ought not to be any surprises based upon what I'm seeing.
I have noticed in the Central Zone, sometimes as much as 5 feet (it's noticeable), however, I think that is due to how Global Mapper reprojects it as I do not survey in the photo's native format.
> I flew and acquired quite a bit of NAP photography about a decade ago, including coverage from Austin south to the border and to Houston on the east and Del Rio on the west. We were based on NAD83-CORS and used an Applanix POS/DG system which when the GPS/IMU was post-processed positioned the camera at about the decimeter level. Other standard photogrammetric error budgets effect the ultimate ground accuracy. The ground checking we did showed we far exceeded the minimum specifications for this orthophoto scale, and was surprisingly good. Flights were 20,000 feet above ground. The specification is the standard, as you are seeing you may well have imagery that exceeds spec.
The metadata for the 2008 0.5m imagery describes the equipment and methods you describe. I'm going to survey some photo-identifiable features just to make sure that it is as good as at first impression it appears.
We use photos all the time for certain projects, one recently was a stream location. We had new 2011 photos, but not the government ones or google or bing. Those I've had issues with when I've checked them. We have a very large data base of photos flown for all different reasons. You might ask around and get something from an engineard, dot, utility, oil and gas guys, it's amazing how many flights go on all the time . I wouldn't use the 1m ones for much of anything. And don't forget the projection effect that can bite you
We use those often and over the years have found from 1 to 18 feet when we had a good ground point to check to. The worst I found was the golf course at Royal Oaks where I could easily identify the greens on the photo. I know they are ortho-rectified but I'm thinking they get worse as you get farther from the flight path. We recently used the TNRIS 0.5m photos for a portion of Webster Oil Field and the location of the wells and pads on the aerial matched within a couple of feet from where had shot them in.
