I have been trying to Orthorectify a number of images (in a flight line) taken from a personal aircraft that has had a Nikon D200 inserted into the body of the plane that is looking directly NIDAR. I have basic information such as the lens (42MM), altitude, micron size and center GPS point. I have a Quickbird Image for the same area as a refference image.
In the Frame Editor I have set the Micron size to 6.1 and as i do not know the interior information i have set these variables to 'Unknown'. I have then collected various GCP's using the QB image as the reference. I have also ran the Automatic Tie Points which has gathered many more points between the 2 images.
When i try and run the triangulation i get RMSE in the order of hundreds, not the usual 1 or 2 i get when doing this with imagery with Fuducials. I appreciate that using this type of imagery is not ideal but ihave imagery going back 20+ years thatt was taken like this and it will help in a numberof research projects if i can manage to orthorectify this to a relatively 'good' standard.
Does anyone have any expereince of using 'simple'imagery such as the type i am trying to use? Why is it that the first image int he block file usually appears quite good and they get worse as they go along?
Any advise would be much appreciated!
Regards
----------------------------------
You were sent this email because you opted to receive email notifications when someone created a new thread, or replied to a topic within this forum, ERDAS Software Forum.
To unsubscribe:
To unsubscribe from this item only, visit the item link and turn off notifications.
To stop all notifications of any updated items from ERDAS Communities, go to Edit Profile, click on the Site Options tab, then under Email Configuration, set Send Notifications to "No".
I am extremely interested to hear answer too. Doing photogrammetry with normal tools is very interesting topic to me also
I have no such photogrammetry background that I could immediately say what is the issue. However I have same feeling as you - you should have all the balls needed in hands and question is now just how to use those.
The first thing I notice that you talk about lens - is that same thing as focal length or something else. To me if you know the focal length, pixel size in microns, center GPS points and reference data you should have everything needed. Of course with low end cameras all calibration etc. is not as accurate as in hight end sensors but to me that shoulc be compensable with bigger amount of ground control and tie points.
----------------------------------
You were sent this email because you opted to receive email notifications when someone responded to this thread.
To unsubscribe:
To unsubscribe from this item only, visit the item link and turn off notifications.
To stop all notifications of any updated items from ERDAS Communities, go to Edit Profile, click on the Site Options tab, then under Email Configuration, set Send Notifications to "No".
I don't agree that GCPs and tie points would compensate the lack of camera calibration; the lens modeling errors remain... And in interior information, can the amout of pixels in the images or something fixed variables, e.g. physical dimensions be defined?
----------------------------------
You were sent this email because you opted to receive email notifications when someone responded to this thread.
To unsubscribe:
To unsubscribe from this item only, visit the item link and turn off notifications.
To stop all notifications of any updated items from ERDAS Communities, go to Edit Profile, click on the Site Options tab, then under Email Configuration, set Send Notifications to "No".
I've used a Nikon D200 with a 28 mm lens fixed to a helicopter and achieved good results using an lidar data for my control points (overall rmse of 0.38 pixels). I used an image pixel size of 6.1 microns (as you did) but check your lens and, as Timo suggests, hopefully it has a fixed focal length. Triangulating non-metric imagery can be difficult as the software has to estimate both interior and exterior camera parameters at the same time. In this regard, I'm wondering about the quality of your control data... deriving an x- and y- coordinate from a QuickBird image should be fine but what about your z? Are you using a separate DEM for this and what kind of precision does it have?
----------------------------------
You were sent this email because you opted to receive email notifications when someone responded to this thread.
To unsubscribe:
To unsubscribe from this item only, visit the item link and turn off notifications.
To stop all notifications of any updated items from ERDAS Communities, go to Edit Profile, click on the Site Options tab, then under Email Configuration, set Send Notifications to "No".
I'm gald someone else is trying the same thing somewhere! The lens has a fixed focal length, a prime lens, so that is covered i think. I am not using a DEM as the one i have for the area is particularly bad and the majority of the scenes i am looking at are reef scence so the average elevation is between 0 and 1 M so i was hoping that this would not have that much of a bearing.
The problem appears to be that when i use a block file, with say 10 images in, the fist image is fine when i overlay it onto the QB image and then the rest get progressively worse as it goes through the block file witht he last image in the block being out in the order of 10's of meters.
Thanks for the input thus far . . . . .
----------------------------------
You were sent this email because you opted to receive email notifications when someone responded to this thread.
To unsubscribe:
To unsubscribe from this item only, visit the item link and turn off notifications.
To stop all notifications of any updated items from ERDAS Communities, go to Edit Profile, click on the Site Options tab, then under Email Configuration, set Send Notifications to "No".
However in theory you should be able to compensate these by adding ground control and tie points. Or is sensor inaccuracy and missing calibration information so big factor that even bigger amount of ground control will not work it out?
timo
----------------------------------
You were sent this email because you opted to receive email notifications when someone responded to this thread.
To unsubscribe:
To unsubscribe from this item only, visit the item link and turn off notifications.
To stop all notifications of any updated items from ERDAS Communities, go to Edit Profile, click on the Site Options tab, then under Email Configuration, set Send Notifications to "No".
My strategy when working with non-metric imagery is to first make a rough estimate of the exterior camera parameters, refine the estimate, and then triangulate the block using the derived exterior parameters. This can help overcome a lack of good quality GCPs in your block. You might have some luck by trying the following:
- choose a single image pair in your block with good quality GCPs
- estimate the x,y,z position for each image from your GPS data (if the GPS data is of poor quality you may be able to estimate the x,y coordinates from the QuickBird image and the z coordinate from the scale equation)
- assume omega and phi rotations are both 0 (i.e. the plane was flying level) and estimate kappa from the direction of your flight line and the orientation of the camera as mounted to the aircraft
- set the standard deviations for the exterior parameters to something relatively large as these numbers are only rough estimates
- double check your GCPs and make sure they're perfect
- same goes for your tie points and deactivate any that are questionable (poorly matched tie points will create headaches -- I'm guessing that you may have some points on the water surface and these may be affected by swells or waves between successive frames)
- set the GCP standard deviation to something relatively small and triangulate the image pair (experiment with the additional parameters and the standard deviations of both the exterior parameters and the GCPs)
- check the triangulation report and keep experimenting until the derived exterior parameters look appropriate (hopefully they match close to your initial estimates -- if they're nonsensical repeat the previous steps) and your RMSE is low
- once satisfied with your results, update your newly-derived exterior parameters, reduce the standard deviation of the exterior parameters and relax the GCP standard deviation
- triangulate the image pair again, check the report and hopefully the residuals associated with the exterior parameters are relatively small
- if your results look good, then start adding additional frames repeating the steps above for each
----------------------------------
You were sent this email because you opted to receive email notifications when someone responded to this thread.
To unsubscribe:
To unsubscribe from this item only, visit the item link and turn off notifications.
To stop all notifications of any updated items from ERDAS Communities, go to Edit Profile, click on the Site Options tab, then under Email Configuration, set Send Notifications to "No".
A few years back I was struggling with a metric camera that had bad calibration information (which I didn't know at the time). I did have the benefit of knowing the pixel size, focal length, and exterior orientation quite well. What I found was that the principal point offset (x0,y0) was bad in the calibration report, and this was causing a progressively worse alignment between images as you would move down a strip. What I wound up doing was using two overlapping strips flown in opposing directions and incrementally nudging the x0 and y0 values. Then I would look at the in-strip and cross strip overlaps in stereo to check alignment. Eventually I found some numbers that worked very well. And when we got the corrected calibration report they were very close.
In this case that might not work as well unless the things I cited are well known, but I wanted to point out that not modeling this offset can cause significant issues. In my case one of the offsets was over 50 pixels from image center!
The GCP from the satellite imagery may have some absolute positional errors, but they should be pretty consistent relatively within the same image. This can help with at least getting a good relative adjustment among all your images. Also if everything you are measuring is pretty much Z=0 or 1 you can just use those values for the Z in the GCP. Not having any vertical control can cause you problems with scale, especially if your focal length and flying height are not well known and locked.
Tom
----------------------------------
You were sent this email because you opted to receive email notifications when someone responded to this thread.
To unsubscribe:
To unsubscribe from this item only, visit the item link and turn off notifications.
To stop all notifications of any updated items from ERDAS Communities, go to Edit Profile, click on the Site Options tab, then under Email Configuration, set Send Notifications to "No".
I have been using data collected from digital cameras on-board aircraft in LPS for more than a decade, so I have some background on what may go wrong in this situation. But I will assure you that once you get things working well, you can expect very good results. LPS is certainly sufficient to perform this type of processing, so don't despair. Are you using GPS data for the camera station (location) for each image? Do you not have INS/IMU data (to provide Omega/Phi/Kappa angles)? Those can be estimated, and if you insert a value that is only a guess, LPS will improve it from there.
What's the data rate on the GPS? Typically it is 1 Hz, so your GPS position is always lagging your actually position by as much as 1 second. What's the frame rate on the camera? I typically collect data every few seconds, so unless you make some interpolation to the GPS data, several successive frames will each have the same identical GPS coordinates.
I see that you have received plenty of good advice from others here, so I just wanted to throw in that bit about the GPS lag that hadn't been mentioned.
All of us who are using digital cameras in aircraft should get together and share our experiences. It's quite different from satellite imagery or aerial photography processing.
----------------------------------
You were sent this email because you opted to receive email notifications when someone responded to this thread.
To unsubscribe:
To unsubscribe from this item only, visit the item link and turn off notifications.
To stop all notifications of any updated items from ERDAS Communities, go to Edit Profile, click on the Site Options tab, then under Email Configuration, set Send Notifications to "No".
