In a previous section I showed how to make one type of DEM overlay using satellite ortho images. In this section, I will demonstrate a similar technique for overlaying terrestrial ortho images produced by aerial photography.

Until recently, gisdatadepot.com offered free digital ortho quarter quarter quads (DOQQs) for free download. DOQQs suitable for this exercise can be downloaded there for a charge. However, several states offer DOQQ libraries for free. My target state, New York is one of those that maintains an extensive GIS library at the New York State GIS Clearinghouse. See the instructions in the News section on where to look for other states.

A DOQQ is one quarter of a DOQ, which in turn covers one quarter of a USGS quad. So there are sixteen DOQQs per quad. A quad is the same area covered by a USGS topographic map or DEM. So the idea is to merge sixteen DOQQs into one ortho image and overlay it on a USGS DEM.

The technique that I will outline is a quick and dirty approach. As I have stated in previous articles, the quality of a DEM overlay is directly related to the quality of the raster image, so you can improve your results to an extent limited only by your computer resources and patience (this technique requires lots of both). The technique is mainly aimed at using free DOQQ data for those states that offer them. However, these techniques work just as well for DOQs. In fact, there is less work to do with a DOQ and your results will be better because there is no merge to perform.

DOQQs are archived in something called MRSID format. This is a clever multi-layered scheme for efficient compression and presentation of image data. So, in order to use DOQQs, you will need a MRSID reader. Your best choice for this exercise is to download Arc Explorer. Arc Explorer can read, merge and export MRSID format so this makes it the ideal tool for making DOQQ overlays. You can also download Geo Viewer, a program offered for free by Lizard Tech, the owners of the MRSID format patent. Note that Arc Explorer can only export BMP format, so your image will be limited to a color depth of only 8 bits. GeoViewer allows you to export Tiff format which supports a 24 bit or 16 million color depth, but cannot merge images.

So if you want the highest quality image, you will have to use GeoViewer. But be prepared to do a lot more work because you will be performing sixteen manual merges instead of just four as in this exercise. GeoView is also handy for determining the UTM corner coordinates of the ortho image in case you are not overlaying an entire quad, as we will here. (The corner coordinates are necessary for registering the image onto the DEM.) In the case of overlaying less than 16 tiles, you will also need to convert UTM coordinates to decimal latitude and longitude. The best way to do this is to use an online UTM Conversion Applet. Again, this will not be necessary for our simple exercise. And finally, a good image-editing package such as Paintshop Pro or Photoshop is essential.

I used my familiar target of Salem, NY for this example. You will need to download all sixteen of the DOQQ tiles for the Salem quad. They will be grouped NW, NE, SE and SW and each of these will in turn be numbered 0 through 3. The tiles need to be assembled in the counterclockwise order 0,1,2,3 starting from the upper left corner of the quarter and ending at the lower left. The four NW tiles are shown to the right. (These images are shown much smaller than the actual DOQQs).

The technique for tiling an entire USGS quad is as follows: load the DOQQs four at a time into Arc Explorer. You do this by clicking on 'Theme' and adding a Theme. When you do this a file directory will come up. Browse to the directory where your .sid DOQQs are and select the four from the NW quadrant. Arc Explorer will edge match the tiles automatically. Copy the merged tile to the clipboard by clicking the selection under the 'Edit' toolbar. Create a large blank new window (about 4000 X 4000 pixels) in Paintshop Pro. Then open the merged quad as a new window in Paint Shop. Each time you import a quarter quad, resize the image to 25% of its original size by making the proper selection under the 'Image', 'Resize' choices. This will prevent the merged overlay from being too large. When this is done, copy the resized image and add to the blank window as a new layer by making the appropriate choice under the 'Edit' selection. The four merged NW tiles are shown to the right.

Repeat the steps for the other three quarters, merging, resizing, and adding as new layers as described above, until you have all four quarter quads assembled. (You may wonder why I do not just merge all sixteen tiles in Arc Explorer first. The problem is, Arc Explorer will only copy the visible section of the image to the clipboard. If I merged all sixteen tiles, I would have to reduce the size of the image until it was all visible in the window before exporting. This would result in unacceptably poor resolution in the overlay.) You will need to edge match the four quarter quads manually. This is done by a combination of cropping and nudging the images, bringing the appropriate image to the front as required to get the best match.

When you are done, merge the layers, crop the border, and save as a .bmp file. Now you are ready to overlay the image using 3DEM. This is done in the usual fashion. Download and open the SDTS 1:24,000 Salem, NY DEM into 3DEM. When the DEM displays, move the cursor to the upper left and lower right corners and record the latitude and longitude of the corners of the quad. Then choose 'Operation', 'Apply Map Overlay' and 'Load'. Browse for your .BMP image and load it. Choose 'Georeference' and move the red and blue markers to the corners of your image. Key in the latitude and longitudes that you noted earlier and click 'OK'. The .BMP will be automatically registered to the DEM.

Now it is only a matter of creating the 3D view and positioning the viewing perspective. Choose 'Operation', '3D View' and set 'Terrain Position' to the 'Far' setting. Render the view. Now choose 'Operation' and 'Change Position' to position the viewing perspective to best advantage. You can 'fly' the 'camera' across the terrain to get the best view.

The result of our efforts are shown in the two middle images. The first image is looking North and the second is looking Southwest from the Vermont direction. The result is an interesting image that offers a more realistic perspective than an overlaid topo map. You will notice that there are a lot of red patches on the image. I believe that this is a result of the characteristics of the sensor used to record the image. The red areas seem to represent row crops, as plowed fields and trees do not seem to reflect the same spectral band that the corn fields do. (I think this is how illegal marijuana patches in the woods are detected as well.)

I was not happy with the red mottled effect and unnatural colors resulting from the sub-visible spectral sensitivity of the sensor used to collect the images. So I modified BANDPASS to correct the colors back to something more natural. The results of the image processing are shown in the lower two maps. (Of course the image processing should have been done before the overlay, but I did not have the modified program and it would take more time than I have to redo them.)

BANDPASS does this by looking at the magnitude of the red component of the RGB value of each pixel. If it is greater than a threshold, and the green and blue components are less than another threshold, BANDPASS reduces the red component (why am I telling you my secrets?). This produces much better results than a straight color substitution. Notice how BANDPASS cleverly leaves the white text alone, and only fixes the offending red pixels. (Big deal, you might say: we are talking about red, not white. But don't forget that since tif uses an additive color model, each white pixel has 100% (0xff)of each of the three color components including 100% of the red. BANDPASS knows to leave the red component alone in this case.)

I really need to modify this program to allow user-input for the threshold values. I suspect that there is considerable variation in DOQQ coloration, which means that the settings I have hard-wired into the program may not work well in all cases. If I get any indication that anyone cares, I will modify the program.

It is interesting to note that even the processed images look somewhat unpleasing to the eye. This highlights one of the principal underpinnings of cartography: don't make your maps too realistic. Maps are supposed to represent an abstraction, i.e. just the important information is abstracted and presented to the user. This is much more useful than presenting too much information. Anyone who has done any flying knows how confusing things can look from the air, and how easy it is to get lost, despite the bird's eye view. It takes aeronautical charts and other navigational aids to find your way back to the airport.

See the excellent tutorial offered by the NY State GIS Clearing House for more information on processing DOQQs and Digital Ortho Imagery.

Note: for an alternative technique for transforming IR band images to pseudo true color composites, see the excellent article at Topoimagery