ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume I-4, 2012 XXII ISPRS Congress, 25 August – 01 September 2012, Melbourne, Australia THE MICROSOFT GLOBAL ORTHO PROGRAM Wolfgang Walcher a, Franz Leberl b, Michael Gruber c a Microsoft Bing Imagery Technologies, 1900 15th Street, Boulder, Colorado, [email protected] b Graz University of Technology, Inst. of Computer Graphics and Vision, Austria c Microsoft Bing Imagery Technologies, Graz, Austria Commission IV, WG IV/2 - Automatic Geospatial Data Acquisition and Image-Based Databases KEY WORDS: Transnational, Photogrammetry, Automation, Ortho-Rectification, Aerial Camera, Digital Elevation Model ABSTRACT: Wide area and thus continental mapping extending beyond national borders is a novel concept in civilian photogrammetry. The Microsoft Global Ortho Program was launched in the Spring of 2009 as a result of Microsoft’s need for global geo-data at a high geometric resolution and radiometric excellence. By fall of 2012 more than 10 million km2 of the USA and 14 European countries will have been covered by seamless 30 cm GSD color-, 60 cm GSD false-color infrared ortho-mosaics and a 1 meter GSD digital surface model. The ortho-maps are being published to Microsoft’s Bing Maps Internet mapping portal. The Global Ortho Program was designed for highly and unprecedented automated mapping of essentially entire continents. In 2011, exclusive of flight operations, the product output per person has been measured in excess of 275,000 square km per year. We describe research efforts that made this achievement possible. Those include a specially designed aerial sensor (Ultracam G), logistics simulation for fight planning and optimization, in-flight blur detection and subsequent automatic blur removal, modeling and removal of atmospheric and environmental conditions, automated shear detection and DTM refinement, an IT architecture to process > 200,000 aerial images/day, and for creating over 1,000,000 km2 ortho-imagery and DSM data in 24 hours. While addressing these issues, we provide ideas how this might affect the future of spatial infrastructure initiatives. 1. Motivation for the Global Ortho Program the USA National Agriculture Imagery Program (NAIP) [NAIP, 2009]. That program creates 1m GSD color imagery of Online mapping portals like Microsoft Bing Maps have about one third of the continental USA every year. At one pixel distinguished themselves from earlier services by an imagery- per square meter, this equates to a multi-year average of about rich global experience. Users are expecting high-resolution, 3 trillion pixels per year. For comparison, in calendar year recent imagery of any place in the world. With commercial 2011, Global Ortho was creating 112 trillion pixels. This very satellite imagery being restricted to resolutions of 50 cm or likely makes Global Ortho the largest commercial aerial less, aerial imagery may be preferable. However, such coverage mapping program to date. has not been very conducive to global applications. This, in 2009, triggered a Microsoft-decision for the Global Ortho Program. The goal: Cover the USA and Western Europe with 30 cm resolution orthophoto imagery in 2 years or less. The Bing Maps Imagery Technologies team (BITs) in Boulder, Colorado was tasked with executing the program. Global Ortho is not a one-off project; it plans to extend into new territories and continuously update aging imagery. 2. Program Scope Phase 1 of the Global Ortho Program covers the 48 states of the conterminous USA and 14 countries in Western Europe. Figure 1 (USA) and Figure 2 (EU) show the respective maps overlaid with the 1° x 1° cell grid [see section 3]. Phase-1 comprises Figure 1: 1° grid in USA. Colors describe priorities by population close to 1,400 complete and partial 1°-cells with ~10.4 mio km2 density (H-high, S-standard, R-rural). of land. By April 2012 the program covered > 99% of the US and 66% of the European terrain, well ahead of schedule. Transnational aerial mapping of this scale has no precedence, although one can assume that there are past or present non- publicized governmental satellite mapping projects of even larger size and much higher costs like the $7.3 Billion EnhancedView contract awarded by US National Geospatial- Intelligence Agency (NGA) [EnhancedView, 2012] Aerial mapping programs could be related to one another via the number of pixels generated. In 2011, Global Ortho created 8,400,000 km2 of in-spec 30-cm orthomosaics, Color InfraRed (CIR) mosaics and other by-products. The 2011 Global Ortho pixels were > 30 times bigger than the average annual output of Figure 2: 1° grid in EU. Colors as in Figure 1 53 ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume I-4, 2012 XXII ISPRS Congress, 25 August – 01 September 2012, Melbourne, Australia 3. Program Logistics and Aerial Acquisition cells or bodies of water. Each color-ortho has a twin CIR mosaic, even though at a lower resolution [1 CIR-pixel = 4 Irregular national borders complicate any orthophoto project. RGB-pixels]. The CIR product is not published to Bing Maps Global/transnational coverage supports efficiency. Therefore but can be purchased from DigitalGlobe. the well- established 1° x 1° grid as defined by the US-NGA serves well to spatially organize the Global program. The 1° cell or “block” is a production unit with an average area of ~8,000 km2. All specifications, planning, processing, quality control and product publishing are based on those blocks. Partial cells exist near water and along denied areas. The North-South dimension of each cell is approximately 111 km while East-West dimensions decrease with latitude. Flight direction is standardized to N-S unless terrain or airspace restrictions force a different pattern. Flight plans minimize the number of flight lines for a cell. High forward overlaps at 85% are needed for fully automated aerotriangulation and dense image matching. Therefore, a typical cell will be covered by 12 to 20 N-S flight lines producing ~ 1,700 exposures. Population density causes slight variations in specifications. About 10% of the areas are urban with denser flight lines to minimize building lean in ortho-products [thus High, Standard and Rural population densities in Figs 1 and 2]. To collect imagery per cell within a single 6-hour-aerial sortie, the novel UltraCam-G [UC-G] camera was built [section 5]. High productivity also was made possible by means of a novel project simulation tool. Individual airplane characteristics, flight plan (altitude, lines, turns, etc.), historic weather patterns, daily sun angle, leaf on-off seasons, seasonal snow cover, technical, logistical and localized environmental factors enter Figure 4: Color RGB orthophoto mosaic, Presidio in San Francisco, into scenarios for the forecast of a globally optimized solution. CA. Area shown is ~ 2.5 km by 3.0 km Figure 3: Simulated flight pattern over the N-E US for a 6-month period. Color indicates different flight dates. The forecast over a period of 12 months has proven to be within ± 4% of actual acquisition totals. As a result of this optimization, only 12 UC-Gs, operated by experienced regional aerial photography firms, were required for all aerial work. These UC-Gs and aircraft collected > 1 mio km2 in Oct 2011. 4. Global Ortho Data Products Global Ortho produces a 30 cm RGB true color and 60 cm false Figure 5: Color Infrared (CIR) ortho mosaic of area in Fig.4 color infrared (CIR) orthophoto (Figs 4, 5). A digital surface model DSM [and DTM] gets created for use in image A main achievement has been the level of automation of rectification. Only the RGB-product is published to Bing Maps. production processes while meeting specifications in radiometry, geometry and environment. This needed automated The 1° grid, as the smallest unit of data production, is a Quality Control (QC) and equally automated problem fixes. seamless image mosaic with a 200 meter buffer into adjacent This in turn required all specifications to be quantifiable and 54 ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume I-4, 2012 XXII ISPRS Congress, 25 August – 01 September 2012, Melbourne, Australia measureable in a way that can be implemented in computer Centeno, 2009]. A wide swath with almost 30K cross-track programs. At the same time the old adage applies, namely that pixels and high photogrammetric elevation accuracy was 100% perfection is impossible to achieve. Conversion of achieved. The later was especially important to avoid the standard aerial trade practices into computer code reveals their dependency on pre-existing elevation models for the creation of often very vague language, and that expectation of flawless high-quality ortho products. perfection is unrealistic. The UC-G came about in 2009 at Microsoft Photogrammetry in A discussion of all quality measures and specifications is Graz, Austria. This team has a history of developing innovative beyond this publication. However, we can describe the most aerial cameras [Leberl et al., 2003]. Figure 7 and 8 illustrate a 7- important specification, namely on absolute positional cone camera with a larger format, but lower resolution accuracy. Table 1 below lists the minimum requirements for panchromatic as well as a smaller format higher resolution absolute horizontal errors. It is important to mention that to color component. The panchromatic image is formed using the date, with 90% program completion, all tests so far have shown proven technology of the UltraCam Lp with 2 camera heads, an RMS error of less than ± 2 meter in x and y directions. using precision calibration [Ladstädter et al., 2010]. The focal length is at 40 mm for a relatively wide field of view of 80° Block CE95 Accuracy ~ Equivalent across the swath. This produces a strong photogrammetric Designation Requirements RMSE in X and Y solution and serves as the geometry backbone of the UC-G.