Of course, the purpose of this paper is to answer this question with a resounding, "No!" and offer details as to how orthophoto products vary.

There are obvious differences: film type (for example; color, color infrared, black and white), and various pixel resolutions. These have significant impact on the finished product and how it may be used. However, there are more subtle factors which may escape casual inspection, including the aesthetics of the imagery, and the accuracy of the imagery.

This paper shall concentrate on rectified orthophoto imagery derived from relatively small scale aerial photography. In particular, it will go into how accuracy standards apply to orthophotos.

Accuracy is a quantifiable property in orthophoto imagery, but it is often misunderstood. Firstly, orthophotos have no vertical accuracy. There is no way (within an orthophoto) to display the elevation of a feature. Only the horizontal position of ground features may be quantified.

Orthophoto imagery requires multiple input data sets. These are Aerotriangulation (AT), a digital terrain model (DTM), camera information, and the scanned image (with fiducial measurements). An error in any of these data will introduce error to the final orthophoto product.

The output of AT is an air station, or the exact position of the camera, (x, y, z, yaw, tilt, roll) when it took an exposure. This information is critical to orthophoto rectification, as it controls how the photo will be aligned with the terrain model (and to the mapping plane) to remove distortion. An error in this data will distort the entire orthophoto, either radially or in one direction.

The use of Internal Navigation Systems (INS) has become popular to replace AT, because this eliminates the need for ground control points. While this is a paper topic in itself, using INS is, in general, less accurate and more unforgiving than traditional AT. Lower flying heights may be required to meet accuracy when using INS than with traditional AT.

Also note that without ground control points, there is no way to compare the position of features in the resulting orthophoto to the ground. These factors do not preclude the use of INS, but it is important to review the process before hand, as well as inspecting the output data from this system, to ensure that errors are not introduced which will throw the resulting orthophotos out of the accuracy specification.

If the terrain model is inaccurate to the ground, the orthophoto imagery will be removed from true position, only in the (hopefully localized) area of inaccuracy. This problem is apparent only when away from the nadir of the photograph.

A rule of thumb is that a vertical error in the terrain model at the outer edge of the usable area of a photograph (flown with 30% side lap), will result in half the horizontal error in the orthophoto. Therefore, a 6-foot vertical DTM error at the edge of an ortho will result in a 3-foot horizontal error in the orthophoto.

The focal length of a camera must be included in the rectification calculations, as well as the position of the fiducial marks of the camera. An error in these values will result in a radial distortion throughout the orthophoto.

The fiducial marks of each photograph must be measured on the scanned image to relate that scan to the camera, the air station, and the terrain model. An error in these measurements will cause a large scale distortion in the orthophoto, found in relation to the erroneous fiducial mark.

Absolute orthophoto accuracy is quantified by measuring the distance between where ground features appear in the orthophoto and where they appear in reality. An accurate measurement of where a feature is in reality may (arguably) only be captured through field survey techniques.

The number of required ground control points will vary, depending on the accuracy standard. Obviously, the desire is for a good statistical indication of overall accuracy. The ground control used for this measurement should not be included in the AT solution, as that would skew the data to being more accurate in the vicinity of each point.

Can the absolute accuracy of an orthophoto be measured without surveyed points? No. The relative accuracy can be measured. This could be done by comparing the orthophoto to planimetric data (where ground features are visible). Or, the orthophoto could be edge matched against adjacent orthophotos, and the relative displacement would be quantified.

The orthophoto industry largely follows three types of accuracy standards.

1. American Society for Photogrammetry and Remote Sensing (ASPRS) Accuracy Standards for Large-Scale Maps
2. U.S. National Map Accuracy Standards (NMAS)
3. Customized Accuracy Standards

ASPRS measures accuracy as a Root Mean Squared Error (RMSE). First, the distance between the position of features in the orthophoto and the ground survey is measured. The RMSE is calculated by squaring the discrepancies, averaging the squared values, and taking the square root of that average.

ASPRS accuracy varies depending on the scale of the map, and the class of accuracy.

Minimum number of check points per orthophoto: 20.

There are values for Class II and III as well, which double the acceptable RMS and blunder values for each category.

What does this mean to the appearance of the orthophoto? It is most obvious when overlaid with planimetric data, or when edge matched with an adjacent orthophoto. The shift may be measured with a GIS tool, or pixels may be counted.

NMAS accuracy is defined as 90% of the measured points being within a certain distance of true ground position.

Custom accuracy standards are becoming very popular, where the accuracy is defined to meet the needs of the user. Both absolute and relative accuracy can be given a specification. For example, it may be specified that a 1-foot pixel orthophoto not have an absolute positional error exceeding 2.5 feet, and that the edge match with adjacent orthos be within 1 foot. Where the burden of proof falls is up to the user.

Absolutely not, if one is comfortable working with a dataset which can only be proven accurate in relation to other digital data sets. Orthophotos may line up very well with a set of planimetric data, and a GIS user can profitably use the data for years even if both data sets are significantly shifted from true ground position. If the data is never checked against an accurate ground measurement, or used in conjunction with an accurate dataset, no one will be the wiser. Ignorance is bliss.

It is possible to have a 1-foot pixel orthophoto that meets 100 scale map accuracy, or a .5 foot pixel that only meets 200 scale map accuracy. This may be done to have orthos match the accuracy of a separate dataset, but it may throw off those who are used to certain conventions.

This is where ugly mis-match errors come into play. First, when comparing orthophotos to adjacent orthophotos, the relative accuracy specification is based on the absolute accuracy spec (unless a custom specification is put in place dealing specifically with edge matching).

If one orthophoto may be shifted 2 feet from true ground position, and the adjacent orthophoto may also be shifted the same amount, it is possible that the adjacent orthophotos be shifted in opposite directions, giving a 4 foot relative shift (which is acceptable with the accuracy spec). In fact, if a inaccurate terrain model, erroneous camera information, or bad AT is causing the error, such a scenario is likely.

This can be a bit of an aesthetic shock, so the user should be aware that the standards allow for this.

There has been a recent interest in creating Second Generation Orthophotos. This involves calculating an air station by measuring ground features from both existing orthophotos and new scanned imagery. The advantage of this process is that AT need not be run on the new photography, which also eliminates the need for new ground control and the use of airborne GPS or INS.

Because there is no AT for the photography, models cannot be set to produce new DTM. Therefore, this process requires the use of existing DTM. While this saves the cost of producing new DTM, it also means that the DTM may be erroneous due to changes to the terrain.

Limited update of the DTM is possible. This involves performing change detection on the imagery to define areas of change significant enough to warrant updating the DTM, then running AT on that area, producing the new terrain model, and merging it with the original terrain data.

Obviously, the largest cost savings is realized with no DTM update, and the schedule can be abbreviated greatly. Some may go so far as to suggest flying photography with minimal overlap (for example, 30/30). As long as there is photo coverage of the project area, a second generation ortho may be made. However, the film cannot be used as stereo pairs with this small an amount of overlap, which precludes any AT, DTM update, or using the photography for photogrammetry in the future.

The process can use varying film types (go from black and white to color, or vice versa). Or, planimetric data may be used as the base data to orient the photos (there must be a significant number of photo identifiable plan features for each exposure).

Where no significant terrain change, and the process is run correctly, the new orthophotos should align very closely with the original data set (usually no more than a 1 pixel shift over less than 10% of the area). This means that any planimetric data compiled from the original ortho data will align very well with the new dataset.

There are conventionally acceptable flying heights for different scales of orthophotos. As of late, these conventions have been stretched further and further to reduce cost (flying higher means fewer exposures). The higher the film scale, the more difficult it is to maintain accuracy, and the lower the aesthetic quality of the imagery.

* based on professional opinion, of course!

Flying height is above mean elevation. 1"=3333’ is recommended because flying any higher scale photography would require a flying height over 20,000 feet, which is higher than most vendors will fly. Scan resolution is in microns.

There are standard tile sizes for varying scales of orthophotos. They result in a 25 megabyte file (black and white), or 75 megabytes (color), uncompressed.

These standards are often not adhered to, and there is no reason to be stringent about the size. However, in the past users have been faced with a legion of 1 megabyte files, or have received monstrous 1.2 gigabyte files which wouldn’t fit on a CD.

Here are the standards most commonly seen.

It is hoped that this paper has pointed out aspects of orthophotography that are not commonly thought of when planning a base mapping project. A user of orthophotos should ensure that the basic building blocks (the specifications) of an orthophoto project are in place before soliciting vendors, or beginning production.

Without such specifications, much effort and expense may be put forth to produce something that does not meet the user’s needs.

Scott Cox has participated in and managed orthophoto production with ASI and Sanborn for the last 8 years. He is now branching out into other types of GIS products.