How to get best results photographing aerial photographs for use in GIS?

[Robertdbuckley]

Good morning. This is my first thread!

I am starting a photographic project and need a bit of advice regarding how to set everything up. I work in an cartographic office for regional government in Germany and have received many large format transparencies of infrared aerial images. I would like to scan them, but unfortunately do not have a large format scanner with lighting. I have tried scanning them with a normal scanner but this just produced dark images. I am now trying to photograph the images with my olympus e-420, a light table, and my two lenses. I have a pancake and a 14-42 mm lens. Both lenses produce great images, but unfortunately they are distorted around the edges. I would like to produce images with minimal distortion because later I have to georeference them. This mean I have to make the images fit a "real" area in the world using GIS software. I have heard that a macro lens produces images with less distortion . Is this correct? has anyone done this work before?

thanks,

Robert

[Manfred M]

Robert - The first question would be what is the size range of there transparencies?

Two techniques come to mind; the first is to use a copy stand attached to a light table (i.e. a traditional way of doing this work) and the second is to use a large format drum scanner designed to scan large format negatives and transparencies.

Instead of trying to do this yourself with inadequate tools / experience, what about trying to find someone who has the proper equipment to do this and send out the images to have this done. I would start by asking who does this type of work at other cartographic offices in the country, especially the federal government or state government ones. I'm sure you are not unique in having this issue.

One of these are bound to have a photoreprography unit to handle this type of work with the experience to do what you are attempting to do and if not they may have some company where they contract this type of work out to.

[Robertdbuckley]

Robert - The first question would be what is the size range of there transparencies?

Two techniques come to mind; the first is to use a copy stand attached to a light table (i.e. a traditional way of doing this work) and the second is to use a large format drum scanner designed to scan large format negatives and transparencies.

If were to use the first traditional method using the copy stand, the question still stands what lens is most suitable. Is a macro lens more suitable? Does a macro lens produce less distortion on the edges than normal lenses with the same f-number?

[Manfred M]

First of all, based on the lenses you list, I suspect you are shooting a micro four-thirds (mFT) camera, as the lenses you list are standard offerings. The 14-42mm lenses that Olympus and Panasonic put out are kit lenses that are fairly decent for general photography, but are designed to be quite inexpensive to manufacture, so I would suggest that that some optical quality trades offs were made to hit the required price points. The pancake lenses are designed to be small and compact, so optical quality has been a trade off for size.

Micro four-thirds has another significant twist throw in as these cameras automatically apply software corrections for distortion, vignetting, etc so if you are still seeing distortion there must be some other issue. This is difficult to discuss without seeing high quality images. I wonder if your setup might not be the cause. When the sensor plane is not absolutely parallel to the image, you will introduce some distortion. Also smaller apertures (f/8 to f/11) will tend to help in reducing some of these issues, but going to smaller apertures is going to start giving you some diffraction issues (softening).

Macro lenses are designed to shoot at 1:1 ratios, so while they are fine for general shooting as well, I don't know a lot about them as I've only used them a few times.

As a general rule, higher end lenses will give you less distortion and fixed focal length lenses tend to be sharper than zoom lenses, but these are generalizations and there are examples of where this is not necessarily true. I shoot a mFT camera (Panasonic GX7) and would probably turn to my full frame Nikon D800 for this type of work because of the larger sensor and because I have higher quality lenses for it. I do understand that some of the Leica branded mFT lenses that are built by Panasonic are extremely good and have low distortion.

Unfortunately, you haven't given us a lot of information to go on. Seeing the setup and equipment you are using would help as well as knowing the size range of the transparencies that you are trying to reproduce, software you are using to edit your images and of course your photographic knowledge and experience would be helpful.

[Dave Humphries]

Hi Robert,

It is difficult for us to answer since you have not supplied any additional information as requested; crucially, this was what size are the transparencies?

You say 'large format', so do we assume that means larger than 6 x 6 cm 'medium format', in film photography; 'large format' suggests negatives of sizes of 4 x 5 inches to 8 x 10 inches.

It occurs to me that you really need a 'flat field', or copy lens - one where the plane of focus is designed to be flat, I believe many macro lenses are configured for this, compared to lens designed for normal photography, which if pushed to shoot at close ranges, may exhibit a curved field of focus.

Beyond that, Manfred is correct saying that the lenses for 4/3 format cameras, such as your Oly E-420, do correct lens issues in camera, which begs the question, are you reviewing jpg images, in which the corrections have been applied, or RAW, where they will not, especially if viewing via a third party (non-Olympus) software?

The E-420 is 10 MP, not a lot by today's standards - and those are crammed in to an area of 17.3 x 13 mm, so expanding those up again to view is likely to cause issues in demanding 'copy'/scientific situations. Otherwise; I might have suggested that if you have edge softness problems, simply shoot from a bit further away and crop in post to use the better area of the lens, but that may be counter-productive.

As Manfred has also asked; we need to know far more detail about your set up and see an example image to assist more usefully.

Best regards, Dave

[Robertdbuckley]

Hi,

first of all thanks a lot for all your replies, I really appreciate the time and effort you are putting in to help.

The transparencies are 240mm x 260mm infrared aerial images taken in 1989. I work for a GIS (Geoinformation system) Office in a government agency which has been content to hold these transparencies up to the light for 20 years. I am now interested in digitizing them and georeferencing them. This means that after the images are digitalized, they have to be "fitted" onto the earth using GIS software.

There is no official contract to digitize the images. I am basically taking on this project in order to see if it can be done without the use of scanning equipment. I know that it is a lot of work, but that is my choice. For a small region the aerial photos have in fact already been scanned using military scanning equipment and to be quite honest, the quality of my images are just as good.


My set up is as follows. Olympus E-420 with f/2.8 pancake lens as shown here. (http://www.dpreview.com/lensreviews/olympus_25_2p8_o20). I have found that the quality of this lens is better than the 14-42mm kit lens. I have a light table to put the images onto and have set up a tripod with horizontal arm (Manfrotto XPROB 190 - http://ecx.images-amazon.com/images/I/41jmCSbasOL.jpg). Once the camera has been attatched, I level it using a very small spirit level. The light table has also been levelled of course.

I open the aperture as far as it can go to increase the field of view. The whole image should be fully focused using manual focusing with 10x magnification. This gives very good images. I have attatched a sample image:

This is a whole image. I am finding that I don´t get the resolution I want for the whole image. I need to come in closer and capture a smaller area to get the same resolution as the scanned images. This means taking more photos of a smaller area and then stitching them.

I basically want to get the camera setting right so that I can be happy that I have the best set up possible with the equipment available.

I am not a photographer. But have learnt about the technology. I am an IT Specialist in the field of Geoinformation systems and am well instructed when it comes to file formats and raster image manipulation. My software includes professional Raster manipulation , Image analysis tools, and Spatial data manipulation software. The Adobe suite is available too.

The only thing I don´t have is a Nikon D810 with a high quality macro lens. I do have the opportunity of buying a ZUIKO DIGITAL 35mm Makro 1:3.5 macro lens though if anyone thinks this is a good idea and could ensure a flatter image.

[Manfred M]

The equipment and setup you are using is probably going to work fairly well enough if you are trying to do a few dozen shots as a "proof of concept", but if you are looking at doing hundreds or thousands of shots this way, no way. You'll drive yourself crazy as things will tend to move around a bit. The extension arm on the tripod has its uses, but decreases the stability of the whole setup. Weighting the tripod would help, but the solution is still a bit awkward.

The first bit of gear I would suggest, if you don't already have it and your camera can accept it would be a cable release to trigger the shutter. Pressing the shutter with the camera mounted on the tripod is going to cause some grief (by moving something), so this is an easy way to avoid that problem.

I still recommend using a copy stand, like this oen:

http://www.bhphotovideo.com/c/produc...py_Stand.html#!

You don't need the baseboard, but they seem to always come with one. I would remove the column and bolt it to the light table. Your camera will be fixed in two axes and constrained by the column in the third axis. Once you bolt your camera on and level it, your are in business. One other thought, these copy stands are often sold with a lighting setup; you definitely won't need them for what you are trying to do.

When it comes to lenses, the only type of lens that I know of that are designed for flat surfaces are projection lenses. I'm not thinking of projector lenses as these will be outside of your working range, but rather enlarger lenses. They were designed to project a piece of film that was held flat in a negative carrier and project down onto photographic paper that is held flat on an enlarger easel. The Schneider Componon line is considered to be an outstanding lens this type of work, and any focal length under 50mm would work out well. The only issue with this approach is that one needs to have a set of bellows between the camera and the lens to focus the lens.

http://www.bhphotovideo.com/c/produc...Enlarging.html

Other alternatives would be any high quality lens that focuses close enough. You are definitely not into macro territory here as your reproductions will be well over 1:1 size. As I mentioned before the Leica mFT lenses have had excellent reviews for image quality. Whatever you do, don't stop down too far, as you will get diffraction softening in the image. Any of the software stitching methods are simply going to be too much work for this initiative.

[Robertdbuckley]

ah...forgot to mention that I have a remote control for the dslr!!

[mastamak]

Hello Robert,
In my professional life I worked as a Surveyor and (part time) Photogrammetrist, so I know what you are trying to achieve. Yes, the scanner option is the best solution to obtaining a precise copy of the original aerial photograph but you have to keep in mind that the original photograph, itself, contains a number of lens distortions such as spherical aberration, radial distortion and numerous others which I have thankfully forgotten. In addition, unless you have ground control targets that appear in the image, you have no way of accurately adjusting an individual photo for the scale and angular distortions that will occur as a consequence of the aerial camera not being in a precise vertical position at the time of exposure, due to aircraft movement. As the photography was captured in 1989 it will not have GPS control which was not available at that time, so you don't know the ground coordinates of the image centre - another problem. The camera may have had a gyroscope to maintain verticality but the distortion problems I have mentioned above will still exist to a greater or lesser degree in the original image. These were normally resolved (in the olden days) by the photogrammetric process known as a block adjustment whereby individual photographs were "stitched" together and the resultant mosiac of photos was then distorted to fit to ground control marks which were accurately surveyed and coordinated. Targets were placed over these marks so they appeared on the images.

What this all means is, regardless of how accurate you are able to photograph the images the inherent distortions in the original photo will make it impossible to fit the photo accurately to your GIS, unless it is properly rectified, and the greatest error will be evident as you move away from the centre of the image with significant distortion at the edges.

However, all is not lost. With a bit of effort you may be able to get an acceptable result with relatively cheap gear. Firstly you need to know what is the expected order of accuracy of your GIS. I am guessing a bit because I don't know the details of your photography (flying height, focal length, etc.) but I assume you would be lucky to get better than +/- 30m at ground scale. Your photos should show (around the edge of the frame) the camera details, focal length and flying height. The scale of the image is the flying height divided by focal length. This will give you an idea of the accuracy possible - for example a typical focal length for Lieca aerial cameras was 6" so for a flying height of 12,000 feet the photo scale is 1:24,000. This means a 1mm error on the image will translate to a 24m error at ground scale. So you can see it is a daunting task.

What I would do is send the images to a professional photogrammetrist and ask him to rectify them (remove distortions) and fit them to ground control (you may need a surveyor to coordinate features in the image to achieve this). You will have to get your GIS Dept to allocate some money for this but the resultant file will probably give you sub-metre accuracy and will slot into you GIS with a minimum of fuss. If this option is too expensive you can try the following:

Photograph the images as described above, making sure you keep the focal plane of the camera parallel to the image. This will reduce the distortion. There should be four small crosses in the corners of the images. These are called fiducal marks and the distance between them is very precisely known. To remove the distortion inherent in your photograph (as opposed to the original camera distortion) you need to transform the photos so that the distance between the fiducal marks in your image is exactly the same os for the original image. You can do this in Photoshop using the Transform tool.

Having achieved a reasonable copy of the original you now have to import it into your GIS. I am assuming your GIS software will allow you to fit your image to identifiable features on other layers of the GIS (this is similar to the ground control, mentioned above). Select 4 points at the corners of the image, and, if the software allows distort the image to fit the four identified points. Check other identifiable points within the image to make sure you image fits reasonable with data on other layers.

I hope I have not scared you off, Robert. It is a really interesting project and hopefully you can get something useful but don't expect sub-metre accuracy. If you need any further help send my a PM.
Grant

[Dave Humphries]

Thanks Grant,

It never ceases to amaze me of the collection of skills distributed amongst our members

[outwithmycamera]

I am also a GIS analyst and have a sense for what you want to do here. Grant makes a lot of good points and one of the questions you really need to ask yourself is how accurate do you need these results and what resolution do you need in the result? Depending on your application paying a photogrammetist is probably going to cost a lot for a result that may not be appreciably 'better' than scanning and georeferencing the images yourself. You can use the georeferencing tools available in some GIS software packages (such as ESRI's ArcGIS) to produce a more or less orthorectified image but as you don't have control over the control network your results are going to be in the "sort of close" realm and probably would be appropriate for qualitative rather than quantitative assessments of that landscape, i.e. you probably wont' be able to make accurate area measurements. Its an interesting project and I'd rather pay someone to scan the transparencies than fool around with trying to 'scan' them with a digital camera, as I think you will just introduce additional distortion that complicates matters further and will prove to be a time sink. I'd be interested to see the results you come up with.

[jcuknz]

My first and only thought on this subject is if the edges are distorted [ whatever ] then do not use them. And while you have heard of panoramic photography have you heard of the alternative method of instead of panning the camera around you traverse the camera parrallel to the subject*. I have used this technique with 3D subjects and had problems, but since you are copying a flat object onto a flat surface they would not occur .... as you simply use the central area of each frame. My old fashioned opinion would be to advise that you use an editor with layers for the compilation but that is because I have never used a modern compilation program that the likes of Kathey Lui has if she reads this thread. A PM to her could be worth while IMO
*In this case you do not need to get closer to the subject as the distortion is probably the same at all four edges so you need to be 'back a bit' so you can dispense with top and bottom as well as the sides.

The other technical point is to lock your mirror up as its movement can upset critical work .... and alternative to cable release is if the exposure is long enough to open the shutter with the lens capped [ a little distance in front of the lens ] and wait the count of thirty before removing the cap for the exposure [ when I did it my ISO was 25 and aperture f/64 ... so the odd second inaccuracy in say a 20 second exposure was neither here nor there ]

Hope these random thoughts are of help

edit .. I have subsequently read Grant's posting where he writes about 'stitching' of the originals which is pretty much what I am suggesting to avoid edge distortions .. a plane flying is traversing the ground. Much depends on how many originals you have to copy as to if it is worth the mamoth effort

[mastamak]

You make some good points, John, but I think the underlying problem is not so much the distortions introduced by photographing the original diapositives, as proposed by Robert, but those photographic distortions that are inherent in the original aerial photography. Consider radial distortion as one example. This is where tall objects are distorted outwards from the centre of the photograph. This is most obvious in vertical aerial photographs of modern cities where the tops of the skyscrapers do not appear vertically above their foundations, but are distorted outwards in a circular fashion around the centre of the photo. If you determine the coordinates of the top of the skyscraper from the unrectified photo it will tens of metres different to the base of the skyscraper. Same for your Southern Alps. Unless Robert's photos have little vertical relief this problem cannot be resolved without applying the appropriate photogrammetric solution. In my experience I have seen some good GIS products and many that are less than satisfactory due to a complete misunderstanding of the limitations of raw aerial photography. Chris has hit the nail on the head. It is a matter of determining what accuracy you can reasonably achieve with your available photography and if you consider that to be satisfactory for your GIS application then go for it. But don't be too surprised if you find errors of +/- 100m relative to your survey-accurate base layer.
Grant