Continue Saving with Orthorectification Alternatives

"EI Technologies offers GIS Training by manufacturers’ certified trainers. Our training workshops can help users of all levels improve their skill in GIS."

Scanning the Right Photos

The community Development and Parks Recreation departments had combined budgets totaling about $5,000 to map new zoning lines. Instead of farming the work out, the GIS group determined it could purchase ER Mapper, an image-processing package from San Diego-based Earth Resource Mapping, and create the necessary maps within the budget.

Public Works bought the software from EI Technologies, LLC a GIS consulting firm and ER Mapper distributor in Aurora, CO. The department also purchased a JAZ storage drive with several 1GB cartridges, which were critical in handling file storage challenges.

The first step was to decide which photographs to use – a 1:420-scale set from 1995 or 1:1,000-scale photos from 1992. The GIS group chose the older photos over the newer, higher resolution set because of time and disk storage considerations.

In general, scanned air-photos produce huge digital files, especially after they’ve been rectified. In addition, the scanned 1:420-scale photos would have created a much larger file size than the other set. Although this posed no problem for the image processing software, it would have complicated the data storage and production time issues.

Time, however, was the prime consideration. Public Works wanted to produce the final photomaps in one-section sheets. The 1:1,000-scale photos were ideal because each covered slightly more than a full section. As a result, they could be scanned, cropped, processed and printed as single sections without technicians mosaicing and color balancing them. Conversely, the group would have had to mosaic about four of the 1:420-scale photos to get a full section, and the project would have taken five times longer to complete.

The GIS group scanned 50 air-photos in-house, using an 8.5 by 14-inch flatbed scanner from Hewlett-Packard CO. Palo Alto, CA. The group used the scanner to crop the photos, so only the area within the section was captured. The group also straightened the photo on the scanner to remove the rotation induced by the aircraft’s crab.

“The quality of the photos was good, so they were scanned at 400 dots per inch to achieve a high-quality product,” notes Brian Sullivan, a GIS coordinator in the GIS group. “The goal was to print final maps with the zoning lines accurate to within one to three meters.”

Each scanned photo generated a 15-25 Mb TIFF file and was named according to an established section numbering system.

Registering to Rectifying

The TIFF files were imported into ER Mapper to begin the registration process. The image processing software was run on the same hardware as Arc/Info, which allowed the group to simultaneously access raster and vector files, and to transfer them from one software to another. As each file was loaded, the image processing system prompted a group member to key in information about the preferred coordinate system. Because the photos would be registered and rectified with existing vectors from the Greeley GIS, its datum/coordinate system, NAD27/Northern Colorado State Plane Coordinates in feet was specified. Registration is a command common to image processing systems and usually accessed via a pull-down menu. A group member first selected the TIFF file to be registered and then browsed through corresponding vector files in the GIS. The scanned photo was registered to these GIS layers, which had an accuracy of about one meter. “We typically choose our roads and hydrology layers to register with the photos, because these features are the easiest to find on both the photos and the vector maps,” explains Sullivan. The aerial photo and vector layers were displayed on screen in separate windows. The group first looked for obvious features, such as road intersections, preferably near the photos corners. Using a mouse, a group member clicked first on the point in the photo and then the same point on the vector map. The computer then began locking the two files together based on these points.

After four points were chosen and matched, the image processing system calculated root mean square (RMS) error for each point, giving the group an idea of how accurate the registration was. High RMS errors indicated low accuracy. Another way to view the accuracy of the registration was to overlay the vectors on the photo.

“You won’t notice distortion in the registration until you overlay and see where vectors don’t line up with the roads,” relates Sullivan.

Next, a GIS group member zoomed in to fine tune the areas where features weren’t matching by editing or deleting existing points and adding new ones. For example, in the zoom mode a group member might see that control points weren’t placed precisely at the center of certain intersections. Points could be moved on the photo and on the vector to try and get them matched as close to the exact same location as possible. This step greatly enhanced overall accuracy.

The group member then hunted for additional control points, such as the centers of cul-de-sacs. Once four or more points were established on a photo and the corresponding vector layer, the software automatically found additional points on the vector layer as each new point was selected on a photo. On average, the group picked about 16 control points evenly distributed across each Greeley section. If the registration proceeded correctly, RMS errors dropped closer to zero with every new point added. When the RMS for one point remained high, a group member usually just deleted it entirely; concluding there was an error in the vector layer that threw off the registration.

The group usually was satisfied with RMS errors of about one to two feet. Once all RMS errors were in this range, registration was considered complete for that section. Rectification of registered photos is an automated process. With the click of the cursor on the “Rectify” command, the registered photos were rectified to the NAD27 datum.

Drawing Zoning Lines

After each photo was registered and rectified, it was converted from the ER Mapper file format to the Arc/Info file and stored on a JAZ cartridge. About seven of the 110-120MB rectified photos could be stored on a single cartridge. Eventually they were written to CD-ROM to free up space on the cartridges.

In the project’s next phase, the files were opened in Arc/Info, and technicians began drawing the zoning lines around parks, streams, wetlands, and other protected areas. Based on the accuracy of the vector-overlaid photomaps, the city estimates the new zoning lines are accurate to within one to two meters – adequate for this type of zoning project.

The GIS group printed the photomaps in color with the zoning overlays on 18 by 24 inch sheets and provided them to the Community Development Department and the Wildlife Committee for final approval. They made few changes and submitted the new zoning lines for inclusion in the 1998 development code.

“We were very impressed and a little overwhelmed with the amount of information the GIS photomaps could provide,” exclaims Lohnes. “Already there have been instances when we had developers come in to look at the new zoning lines and they were satisfied that [the lines] weren’t ambiguous or arbitrary.”

Developers, planners and citizens are buying copies of the maps for $14 each and using them for various purposes.

The city departments whose master plans were consulted to initiate the zoning project now use the maps to fine tune their plans and make sure they didn’t overlook any sensitive lands.

Public Works plans to begin building a GIS photo archive by scanning, registering and rectifying aerial photographs every couple years. Ironically, the project has piqued the city’s interest in orthorectification. Now the GIS group is confident it can perform full orthorectification in-house, using the image processing system to prepare engineering grade maps for other Greeley projects.