|
DIGITAL ORTHO
Automating Image Mosaic Production
New software revolutionizes image mosaic process
By Dave Knopp and John Gerhard Introduction
There are fundamental requirements and considerations associated with the economic production of large seamless ortho image mosaics. Seamless implies continuous radiometric and geometric properties across user defined sheets. To facilitate this process, a photogrammetric mapping firm is successfully using new tools and procedures to automate its mosaic generation process. The results from this automated mosaic process have met all processing requirements for a production project with 78 color infrared images. Industry Needs
Today, clients demand seamless, aesthetically pleasing ortho image databases for a variety of applications. Years ago, it was okay to deliver independent image sheets - sheets which would form a checkerboard pattern when the client viewed many images over a large area integrated within an application. Now, however, with today's sophisticated GIS users, it is no longer acceptable for a production photogrammetric firm to deliver checkerboarded imagery. Clients want and specifications now demand seamless image products which cover an entire project area. Mosaic Requirements
Increasingly, the remote sensing and photogrammetric industries are working with larger data sets and are shifting to the use of color imagery. Therefore it is now important for the production service firm to produce large image mosaics. It is also necessary to work with both true and false color imagery in addition to black and white photography.
Color image processing places additional demands on production processing, especially when assembling color imagery into a large seamless mosaic. However, the fundamental requirements of color mosaic production are the same as those for panchromatic image mosaics. Namely: The mosaic must be cartographically precise. Typically it must meet government, industry and/or project standards of accuracy and quality.
It must be geometrically seamless. This is especially true for large scale imagery where building lean and other relief features are discordant from one exposure to the next.
The mosaic must be radiometrically seamless. This requires performing a variety of image intensity corrections within individual images and between adjacent images.
For color imagery, complex, non-linear color relationships must be accommodated. The processing must also address perceptual idiosyncrasies in the human vision system. This is the subjective part of the production process that is also the most difficult to maintain radiometric consistency over an entire project area. Technical Challenges
The large-scale, efficient and cost-effective production of aesthetically pleasing and cartographically correct image mosaics is a demanding process fraught with challenges. Mosaic production must satisfactorily accommodate a wide range of image intensity and color variation. This is complicated because source photography is subject to time dependent changes in illumination caused by sun angle and intensity. Therefore, photography often contains image "hot spots", effects of lens vignetting, multiple exposure dates, and exposure variation on each flight line.
Image quality is affected by film type and photographic processing differences. Based on certain ambient light conditions, film type (grain size) and exposure (aperture) are altered to obtain the best possible tonal quality. However, these two situations have a major impact on the down-stream mosaic processes.
Frequently film is disposed to subjective modifications introduced during photographic reproduction (ortho diapositives) and/or scan digitizing rasterizing. Manual Process
The conventional approach for image mosaic production relies heavily on manual interaction. The standard manual procedures typically involve a three step process: 1) image processing of each image until it "looks good"; 2) manual definition of image "Cut-lines" which zig and zag between non-obtrusive physical and cultural features such as streams and roads, relief features, and along region boundaries; and 3) the pair-wise matching of one image to the next and of one submosaic to the next.
In general, this manual process is extremely inefficient, unproductive and just plain painful. Because the results are based on individual human perception and taste, they often exhibit inconsistency within a project as well as from one project to the next.
In addition, the manual process consumes a massive amount of human analyst time. As a result, the analyst is unavailable to work on critical jobs which require advanced skills and professional judgment.
Interactive mosaic production is often performed on expensive photogrammetric workstations. As a result, manual mosaic production often consumes valuable equipment resources which are then not available for other work. Automated Process
To overcome the limitations of manual mosaic production, Merrick & Company of Denver, Colorado, has implemented a fully automated mosaic generation capability. This automated capability is integrated with Merrick's conventional photogrammetric production environment and has now been successfully demonstrated on several color, color IR and black & white production digital ortho projects. Existing Workflow Integration
Merrick's automated image mosaic process is completely integrated with a standard digital orthophoto production workflow. Photographic images are digitized by scanning directly from negative film rolls using a Zeiss SCAI photo scanning system with automatic roll feed. This allows the unattended scanning of approximately 90 photos during an overnight batch process.
Utilizing the results of either conventional or automated air-trig aerial triangulation processing, the scanned photographs are processed with a terrain elevation model to produce digital orthophotos. Merrick's orthophoto rectification is performed using several PHODIS-OP workstations. Automated Mosaic Generation
Merrick has automated the image mosaic generation process using OrthoVista mosaic processing software. OrthoVista addresses the requirements of automated image mosaic generation by providing the following capabilities.
Fast, easy setup: Merrick generates mosaic sections in "production units" containing approximately one days' worth of orthophoto production. This tracks scheduled orthophoto production rates and is also convenient for overnight processing. After gathering image data from multiple orthophoto production systems the analyst spends about 5-10 minutes configuring OrthoVista for overnight processing.
Internal image balancing: This processing capability compensates for image hot-spots, lens vignetting and other per-image effects. Of particular significance, this capability compensates for effects associated with scanning negative film. This is an important consideration to fully optimize the economic advantages of the SCAI's film autowinder.
Adjacent image adjustment: This capability simultaneously corrects for radiometric and color differences between all images within the mosaic. This 'tilting correction' compensates for differences in color and intensity between adjacent images in order to produce radiometrically seamless image mosaics.
Automated image fusion: After images have been radiometrically adjusted, it is necessary to fuse them into a single geometrically seamless mosaic. OrthoVista incorporates innovative image processing techniques to eliminate the concepts of cut-lines and feathering. Instead, OrthoVista utilizes autonomous image fusion technology to merge overlapping images while recognizing relief features, discrete cultural and physical features, such as buildings and terrain surface breaks.
Output tile generation: After creating a large seamless mosaic, output files are produced in accordance with a project tiling definition. Tiles are typically used to represent cartographic sheet boundaries or to define image units with a size that is convenient for the client. Efficient QC Interaction
As a professional photogrammetric production firm, Merrick must assure the quality of its work. Therefore, the automated mosaic generation process must provide efficient and effective quality control tools.
For QA/QC, Merrick utilizes OrthoVista's QC Tool. This software tool allows the image analyst to systematically review mosaic output tiles. The display provides interactive viewing of the mosaic with a variety of options to superpose data stereo-model limits, exposure boundaries and other pertinent image information.
Occasionally the mosaic may contain a defect (e.g., pixel smear in steep terrain area, urban development between photo epochs, etc.). When this occurs the analyst can draw a polygon around the effected area and extract corresponding pixels from an appropriate image. The "patch" source image need not be part of the original input image collection. Therefore, it is practical to use an "in-between-image" to fill mosaic polygons. Project Review
After initial testing and verification of the automated mosaic generation process, Merrick began utilizing automated mosaic processing for full scale production projects. Now that the OrthoVista software and Merrick's automatic mosaic processes are well established, significant savings in time and costs are possible. The following describes the results of a comparison of utilizing automated mosaic production on the Fort McCoy, Wisconsin, project which contains 78 color infra red images flown for 1":1,000' 1"=1,000' (1:12,000) scale and scanned to 14um pixels resolution.
Table 1 presents the amount of analyst and computer time required to complete the first project with the automated mosaic using OrthoVista. For comparison, Table 2 presents the time required by the "old" method based on interactive, manual image processing and "cut-line" definition. The times presented for the manual process are based on Merrick's extensive experience on past projects with comparable photography.
According to Merrick's Digital Image Manager, John Gerhard, "What we like about the mosaic automation software is that it's so easy to set-up a project. We just tell it what to do.
Our analyst simply inputs project parameters and then lets it do all of the work. The machine software makes the mosaic while our people perform other, more productive work while the processing runs off-line or at night."
Using the OrthoVista software to process the first project deliverable of 21 input images for 35 output image tiles required an elapsed time of 14 hours. Of this, approximately 13.5 hours were machine time - mostly overnight and weekends. Only 0.5 hours were analyst time.
Concerning the quality of results, Gerhard noted, "We closely review the mosaic image. But so far on this project, we haven't had to change anything." Perhaps the most significant quality indicator is that the client has reviewed the imagery and accepted the complete delivery. Speaking from a clients' perspective, Mr. Jon Kragt of the U.S. Army Corps of Engineers, Omaha District, "Merrick's imagery produced using the OrthoVista tools is seamless and will satisfy our requirements better than their conventional method." Evaluation of Mosaic Automation
Production experiences with automated image mosaic generation have produced several observations which can be used to evaluate the performance and the potential of automated mosaic production. According to Brian Raber, Merrick's vice president of photogrammetry, survey, GIS, "As one will see, in Merrick's attempt to provide the highest quality imagery at the most competitive price, the OrthoVista software was a viable investment." Advantages
Automation provides immense savings in human labor. Not only does this reduce production costs substantially, but it eliminates a tedious, inefficient and unproductive task from the production workflow.
Automation frees the analyst to perform productive work where their skills can be put to good use. Because processing can be completed as a background computation, automated mosaic generation frees valuable workstation seats for other uses.
The quality of intensity and color correction is as good as or better than the manual process. Automation removes inconsistency attributable to analyst fatigue and individual perception and preferences.
Automation provides significant economic benefits. The savings in analyst time alone per 100 photos can amount to as much as 5 work days! This is a saving of about 0.4 analyst-hours or potentially 25 minutes per photo given the correct set of photography circumstances. Disadvantages
A skilled image analyst operator is capable of placing mosaic cut-lines intelligently based on scene content and anticipated use of the final product. A fully automated process cannot achieve this same level of intelligence. In these cases, Merrick utilizes OrthoVista to perform only the radiometric and color corrections. The radiometrically corrected orthophotos are then merged using manually defined cut lines.
Investment in automation software is an initial disadvantage. However, Merrick expects to completely recover its OrthoVista investment in significantly less than a year. About the Author:
Dave Knopp is president of Stellacore Corporation. He has 15 years experience in remote sensing, with a focus on developing software applications for remote sensing analysis. John Gerhard has experience in aerial photography as a camera operator and progressed into the digital world. He currently is digital image manager at Merrick & Company where he's been employed for the past two years.
Back
|
