David Brostuen, Scott Cox
Minimizing Subjectivity in Digital Orthopho Imagery
Digital Orthophoto Imagery (DOI) has become an affordable and highly effective component to the modern GIS. However, a variety of considerations must be addressed in order to ensure the DOI youre investing in will meet your expectations and intended applications. This paper will identify issues critical in producing quality DOI, and reveal ratings-based criteria useful in evaluating the quality and suitability of the product. Identifying and implementing these criteria into your project workflow will result in highly accurate, non-biased DOI that will meet your expectations the first time around.
The past several years have seen an explosion in the use of digital orthophoto imagery (DOI) as a source for capturing various thematic layers and as a visual backdrop in a GIS. DOI is aerial or satellite imagery that has been processed to remove distortion caused by variation in terrain, camera/sensor orientation and perspective and earth curvature. Use of DOI by utilities, municipalities and county governments is commonplace. Emerging markets utilizing DOI include real estate, insurance, and automobile location systems. Many areas throughout the United States are being mapped several times over, creating DOI at a variety of resolution and accuracy. The Internet has facilitated marketing and distribution of "pre-packaged DOI" with many flown "on-spec" in an effort to capture this rapidly expanding segment of the photogrammetry market. Accuracy, quality and pricing of these products varies widely, even with imagery of the same geographic area and pixel resolution. Differences in cost are based on factors including:
|Lower Cost DOI||Higher Cost DOI|
|Use of existing/old DEM||New/accurate DEM produced using a variety of methods|
|Control points from existing DOI||New control using field survey methods/AT|
|San every other exposure||Scan every exposure (less relief displacement)|
|May not meet NMAS/ASPRS standards for accuracy||Meets ASPRS and/or NMAS standards for accuracy|
|Distortion of buildings/bridges between exposures||Minimal building/bridge distortion between exposures|
|No color balancing||Color balancing applied to project area|
|No true-ortho rectification||True-ortho rectification for some or all of project area.|
Costs can vary for producing 1 foot color DOI from $80/sq. mile to over $500/sq. mile, depending on the types of processes and data used. These inputs ultimately impact quality, accuracy and other factors. The next time you need to purchase new DOI, identify specifically what your needs are. Before you buy, be convinced that you are confident that the product youve identified will meet or exceed your accuracy requirements. Be suspicious of products that are extremely low-cost and do not specify accuracy. If the company you intend to purchase the DOI from is unwilling or unable to provide details regarding accuracy or the process used in creating the product, watch-out! Purchasing DOI production services through a comprehensive Request for Proposal (RFP) process is beneficial, but only if you ask the right questions. Common requirements include:
Failure to request many of these parameters may result in surprises when receiving your pilot area for review. The majority of RFPs do not address this level of detail. The best way to determine accuracy of DOI is toperform an independent field survey of various locations throughout the project area. Only in this way can assess the accuracy of the product since you are removing dependant variables derived by the scanner, DEM, camera, AT, and orthorectification. Radial Distortion
An important factor that is often overlooked and a cause for client dismay is radial distortion. In both aerial and high-resolution satellite imagery, features are displaced outward from the center of the photograph due, in part, due to elevation variations in the terrain and the perspective geometry of the sensor. The resulting distortion is worst at the outermost edges of the photograph. When mosaicking multiple images together, the end result is often imagery with above ground features such as buildings and bridges being distorted along the seam line between two or more exposures. The problem, commonly referred to as edge displacement, is due to features being displaced in one direction in one image and in the opposite direction in a neighboring image. Figure 1 shows an example of edge displacement of a building divided between two exposures. The problem can be amplified if adequate sidelap/overlap is lacking or, in an effort to save cost, the DOI producer uses every other exposure (rather than all exposures).
Figure 1: Building divided between two exposures.
An effective process to eliminate edge displacement is to produce "intelligent" seam-lines that are placed to avoid cutting through buildings and other raised features. However, depending on the software being used, producing these seam-lines can be expensive. Even still, the process is often necessary if maintaining the integrity of raised features is important and it should be requested in the RFP. When considering off-the-shelf imagery, determine whether or not edge displacement problems exist before you purchase a product.
Radial distortion causes buildings to lean outward, obscuring important features such as sidewalks, road edges, storm drains and other infrastructure. For core business districts in downtown areas of large cities, radial distortion can eliminate vast areas of the ground from being visible. Figure 2a shows high resolution DOI of downtown Manhattan, New York.
Note the excessive lean of the tallest buildings and how they obstruct entire streets and other features. Flying an area at a higher altitude using a larger focal length camera, increasing overlap/sidelap, and acquiring overhead "spot-shots" are common ways to reduce building lean. Using only the central "sweetspot" from each exposure will also reduce building lean since the problem magnifies outward from the center of an image.True-Ortho Imagery
An innovative process developed by Analytical Surveys, Inc. (ASI) eliminates radial distortion by orthorectifying buildings, bridges and other raised features. This largely automated process, called Method for Elimination of Terrain and Relief displacement in Orthophotgraphy (METRO) relies on the use of an enhanced digital terrain model (DTM) that includes buildings and bridges which have been captured photogrammetrically. Using the enhanced DEM and other standard DOI inputs, METRO automatically positions buildings and bridges to their true location. Figure 2b shows an example of the same area of Manhattan using the METRO "true-ortho" process.
Buildings and bridges have no lean and are in their correct geographic location. The primary area of expense for METRO true-ortho imagery is the need for increased photo coverage (overlap/sidelap) and development of the enhanced DTM.Regular Grid vs TIN
Another consideration when purchasing DOI is the type and accuracy of the DTM used in creating it. Currently, there are two common forms of DTM used in producing DOI; a regular grid and an irregular grid, usually modeled with a Triangular Irregular Network (TIN). A regular grid is a matrix of equally spaced points with each point having x, y and z coordinate values. The accuracy of the regular grid in depicting terrain is based on the process used in originally creating it and the distance from one point to another. A variety of processes are used in producing a regular grid including stereocompilation, LIDAR, auto-correllation and conversion from existing contours. In some cases, the grid spacing may be too coarse to accurately model the terrain, and distortion may result. Figure 3a shows an example of a regular grid DEM in 3-D.
Figure 3a: Regular grid DEM
Using irregularly spaced mass-points and break-lines to represent a terrain surface model is the preferred method. Mass-points and break-lines are captured through photogrammetric compilation techniques. The terrain surface defined by a TIN is produced by interpolating elevation features between point and breakline features. Figure 3b shows an example of a TIN for the same geographic location as the regular grid DEM. The fact that a TIN accommodates irregularly spaced points results in a more accurate terrain model than that achieved using a regular grid DEM.
Figure 3b: TIN model replacing DEM
The end results of the orthorectification process using both a regular grid and DTM can be see in Figures 4a and 4b. Although this example may represent an extreme case, it shows there is no apparent distortion with the DOI produced using the TIN in Figure 4a. Conversely, there is distortion of the railroad track using the input regular grid DEM in Figure 4b.
Figure 4a Figure 4b
Another important consideration includes the date when the DEM was produced as compared to the photo acquisition date. If recent photography has been acquired and significant changes have occurred in the project area due to growth, environmental factors, mining, etc., the DEM should be updated to reflect this prior to being used for orthorectification. If not, distortion may occur in areas of change.
The accuracy of the output elevation model can vary depending on how it was produced. In many cases, an elevation model may be perfectly suitable for producing DOI, but may not meet the more stringent requirements for generating contours. If submitting an RFP, emphasis should not be placed on the method used in producing the DTM, but the desired positional accuracy needed. When purchasing existing DOI, determine the type of DTM that was used, the stated positional accuracy and currency in relation to the original imagery.Aesthetic Issues and Quality Acceptance Criteria
An important aspect of DOI is the aesthetic quality of the imagery. This can be very difficult to quantify, as every viewer of the data may want it to appear somewhat differently. Performing pilot projects and establishing up-front acceptance criteria, can minimize subjectivity.
For traditional metric cameras, film scanning is an important step once aerial photography is captured. Photogrammetric scanner settings must be carefully fine-tuned. Capturing good detail in shadows and bright areas reduces the overall contrast of the image. Higher saturation removes the "haziness" of color imagery. Panchromatic (black and white) imagery should have a setting agreed upon for brightness and contrast, understanding the impact this has on shadow or bright area detail. Color imagery is impacted by brightness and contrast, but also by the overall hue of the imagery, and the saturation level (does it appear dull or vivid?).
Once the overall appearance of the DOI has been agreed upon, acceptance criteria may be defined for subsequent deliveries of DOI. ASI has developed a framework for such criteria, based on five aesthetic categories. The acceptance criteria are based on the production of new DOI and assume a pilot project will be performed. However, when purchasing existing DOI, many of these criteria can still be applied to determine if the product meets your requirements.Radiometric Consistency
Inconsistent radiometry is one of the most noticeable aesthetic problems with DOI. Overviews can take on a "checker board" appearance, and high-resolution views may show seams between photos such as an abrupt line between distinctly different color, brightness and contrast levels.
Figure 5: Image on left has radiometric blending considered unacceptable. Image on right shows acceptable radiometric blending.
New tools to better control and blend radiometry are continually being developed. However, drastic changes in photo capture (such as the time and date), or differing processing techniques may result in some minor radiometric differences, even after blending.
Typically for DOI, no radiometric correction is done in water bodies. Sun reflections from different angles can create abrupt changes when mosaicking DOI. Figure 6 is an example of acceptable DOI with sharp radiometric differences in water.
· Sharp contrast is visible at photo seams.
· There are significant radiometric differences between photographs.
· Overall radiometry is significantly inconsistent between photographs.
· Seams between exposures are blended.
· There are only minor localized radiometry differences.
· Overall radiometry is consistent between photographs.Brightness
In areas of DOI that are very bright (rooftops, for example), some detail may be lost if scanner settings are defined to obtain higher overall contrast. Based on the settings agreed upon at the pilot, if bright area detail is lost, the DOI may still be acceptable. The ability to interpret ground features differs from the ability to see the features plainly. It's still possible to interpret features that are somewhat obscured.
· Features aren't interpretable in bright areas.
· The imagery's contrast setting does not meet specifications set in the pilot.
· Features in bright areas are interpretable.
· The imagery's contrast setting does meet specifications set in the pilot.
It's nearly impossible to obtain orthophotography without shadows from buildings or trees. However, capturing imagery when the sun is at an angle greater than 30 degrees can minimize the effects. In addition, if the aerial film is exposed and processed correctly, then it should have acceptable detail in shadow and bright areas.
After the film is processed, scanning the original negatives (rather than diapositives) can help capture detail in bright and dark areas. A diapositive lacks detail in these areas, because it's a reproduction from the negative. When subsequent generations of film are created, contrast decreases and fine detail is lost.
Similar to the brightness criteria, shadow areas may lack detail and still be acceptable, depending on the scanner settings used.
Figure 7: Image on left has unacceptable shadow detail while image on right shows acceptable detail in shadow.
· Ground features aren't interpretable in shadow areas.
· The imagery's contrast setting meets specification set in pilot.
· Most features in shadow areas are interpretable.
· The imagery's contrast setting meets specification set in pilot.
Desired contrast levels vary among users, depending on the orthophotography's intended use. Users who overlay vector data on imagery often prefer flat-looking imagery so the vector data are more visible. Also, users who plan to make hard-copy reproductions often prefer low-contrast imagery. However, many users say high-contrast imagery is more aesthetically pleasing to view and interpret.
There's a balance between conserving detail in bright and shadowed areas, and producing high contrast. By understanding and maintaining this balance, acceptable imagery usually can be achieved.
· Overall imagery contrast doesn't meet specifications set at pilot.
· Overall imagery contrast meets specifications set at pilot.
*When purchasing existing DOI, verify that the contrast is acceptable for your needs.
Foreign Artifacts and Scratches
Foreign artifacts that don't obscure ground features may be considered aesthetic problems. An acceptable level of artifacts may be defined per unit of area in DOI. ASI recommends a default of 5 scratches or artifacts as acceptable.
Hair, lint, dust and scratches can be introduced to DOI during photo capture, film processing and scanning. Scanning the original negatives can minimize scratches and artifacts. Planning flights at lower elevations allows scanning at lower resolutions, which can help minimize scratches
After scanning, artifacts and scratches can be carefully removed. This operation should be kept to a minimum, however, because ground detail can be removed along with the artifacts.
· More than five artifacts and scratches are present per unit.
· Important ground features are obscured.
· Less than five artifacts and scratches are present per unit.
· Important ground features are not obscured.
The acceptance criteria for one unit of DOI may be displayed on a chart or stored in a database for future reference:Digital Imagery
Evaluation of Aesthetic and Functional Quality
No significant differences
Pilot spec, features interpretable
Pilot spec, features interpretable
Meeting pilot specification
5 or fewer
Accepted X Rejected__ Evaluated by
There can be tremendous variation in the quality, accuracy and cost of DOI, even with products of the same apparent spatial resolution. It is important to recognize these differences when submitting an RFP or purchasing existing DOI products. When requesting a new DOI product, establishing specific technical guidelines and acceptance criteria will reduce variability and provide both the customer and product developer with a mutual understanding that should result in first-time quality acceptance. When purchasing existing DOI, you do not have the luxury of establishing parameters up-front. Dont be swayed by cost alone. Be convinced that the product will meet your requirements in terms of resolution, accuracy, and other issues.
Senior Director of Technical Services
Analytical Surveys, Inc.
Digital Imaging Operations Director
Analytical Surveys, Inc.