3-D City Scenes
DIGITAL MODELS OF URBAN AREAS PROVIDE SITUATIONAL AWARENESS FOR SECURITY PLANNING.
As they jockeyed to gaze upon their favorite celebrities at last year’s Academy Awards ceremony in Los Angeles, Calif., few onlookers probably realized they were being watched almost as closely. The entire scene, from the crush of paparazzi to the procession of limousines and hybrid cars, was being monitored by cameras strategically placed on surrounding buildings. The surveillance devices had been installed especially for the ceremony to allow security officials to spot suspicious or unexpected activities.
Beefed-up security at high-profile affairs like the Academy Awards has occurred for decades, of course, but what was new about the 2005 event was the application of geospatial data and technology from the National Geospatial- Intelligence Agency (NGA). Months before the ceremony, local law enforcement officials had requested the Department of Homeland Security (DHS) provide them with a three-dimensional digital city model of the buildings and terrain around the Kodak Theatre, where the ceremony was held.
Los Angeles law enforcement technicians loaded the 3-D data into computers to model lines of sight from multiple perspectives to determine where the best positions were to place the security cameras. The goal was to provide officials with multiple unobstructed views of the theatre itself and all access routes leading into the neighborhood.
“It was a situational awareness tool,” said Todd Cummings, deputy director of the NGA’s DHS Support Team.
The Academy Awards is just one in a growing list of nationally recognized events using 3-D geospatial information, comprising high-resolution aerial orthophotography and digital surface models, for security purposes. These two data sets are among the most critical that NGA is obtaining for inclusion in the Homeland Security Infrastructure Program (HSIP) for use in a variety of emergency preparedness and terrorism response planning initiatives.
Future awards ceremonies, however, may take advantage of 3-D data acquired through more efficient means than now used. With so much emphasis placed on visualization in homeland security applications, NGA continually improves the methods by which these, and other, key HSIP data sets are created.
Late last year, NGA contracted Earth- Data International, a mapping and GIS firm, to provide 3-D geospatial data in support of HSIP using an alldigital approach. A pilot project completed early this year demonstrated that these data sets can be created faster and less expensively with the alldigital approach than through conventional methods.
3-D APPLICATIONS
In the wake of the September 11 terrorist attacks and the formation of DHS, the U.S. government began ranking major national events that warranted security support from federal agencies. Presidential inaugurations, political conventions, the Olympics and a few others received the highest designation as National Special Security Events. Others, such as the Academy Awards, fell into a prioritized list including everything from Super Bowls to county fairs.
“DHS requests for support usually come from local authorities through the Homeland Security Operations Center in Washington,” said Cummings, explaining that when security involves maps, imagery or geographic information, DHS can task NGA to obtain them, if the data sets do not already exist.
For most major metropolitan areas in the United States, however, much of the geospatial data has already been collected and is being updated regularly. A variety of one-foot-resolution imagery, three-footresolution terrain models, building information and demographic data is stored in the HSIP geodatabase for the 133 largest cities in the country. Image and terrain visualizations covering the Kodak Theatre area, for instance, were provided by DHS to Los Angeles officials from the HSIP database for the Oscars.
The application of 3-D data to prepare for and secure these types of event is definitely increasing, confirmed Cummings. “Three-dimensional city models are used at all of the higher-level events.”
The proliferation of commercially available visualization software and computer processors capable of extremely fast rendering—combined with the availability of highly accurate geospatial data—have made 3-D modeling more routine, according to Travis Hardy, business development executive at EarthData, who works closely with NGA on HSIP acquisitions.
Line-of-sight analysis is one of the most common 3-D applications for security planning. Originally developed for radio wave propagation studies in telecommunications, line-of-sight models allow law enforcement to identify locations, such building rooftops or windows, that could potentially give a sniper a clear shot at a podium or stage. Conversely, security personnel may use the same multi-dimensional views of a venue to decide where to post their own sharp-shooters for a maximum, unencumbered field of view.
Simulated walk-throughs and drivethroughs have also become extremely popular with law enforcement and security agencies. By simulating the driving route of a motorcade through a 3-D city on the computer in advance, experienced security personnel can pinpoint locations where attacks could be launched at the passing vehicles. This enables the forces to pre-position resources around those locations to ensure the safety of the motorcade occupants.
Similarly, certain agencies now immerse their personnel in 3-D walkthrough environments simulating the venues or locales where upcoming events will take place. The ultimate objective is to attain better situational awareness by giving the security personnel a familiarity with the terrain and cityscape even if they have never been there. This walk-through simulation is, in some cases, taking the place of costly advance visits by out-oftown security teams.
“Three-D modeling was the last tactical mile spanned by geospatial technology for homeland security applications,” said Hardy.
DIGITAL ELEVATION MODEL
The 3-D city models included in the HSIP database have traditionally been created with a two-step acquisition method. In the first step, an airborne camera collects digital imagery at one-foot special resolution. During a separate flight, laserbased light detection and ranging (LiDAR) is used to acquire elevation values for the ground terrain and structures.
The LiDAR data is processed to create a digital elevation model (DEM) for the area of interest. The DEM is usually applied to the orthorectification of the aerial photography to create digital orthophotos. Additional digital processing then overlays, or drapes, the orthophotos on top of the elevation data to generate a 3-D rendering that accurately shows ground features, such as streets, fields and lakes, in their correct locations as well as the accurate height representations of vertical structures, such as buildings and houses, and the natural terrain.
“The results of this traditional process are quite accurate, but the problem is that two aircraft flights are required to collect the data, which increases the cost and time needed to produce the 3-D model,” said Adam Cohen, EarthData business line manager.
EarthData proposed eliminating the LiDAR acquisition flight and extracting the DEM from the aerial imagery using advanced processing techniques. Intrigued at the prospect of saving time and money by generating 3-D models from a single airborne dataset, NGA requested a pilot project over St. Louis, Mo., with imagery collected over a 67-square-mile area and the DSM and 3-D model processed for the 2.2-square-mile city center.
“We flew the pilot area in a single day with a Leica ADS-40 aerial scanner acquiring six-inch-resolution imagery,” said Cohen. “The only major difference in the airborne portion of the acquisition procedures was the scene overlap, which we increased to 80 percent side to side and front to back.”
The 80 percent overlap was required for application of EarthData’s proprietary ISTAR processing technique, which extracts highly accurate elevation data through stereo correlation of ground points imaged from at least six different look angles. This softcopy photogrammetric process yielded a one-meter DEM used to orthorectify the aerial image, creating what is known as a true orthoimage. Produced from highly accurate DSM, a true orthoimage does not have the building lean distortion common to standard orthoimagery.
In the city core, technicians turned to manual stereo compilation of the multiple look angles in the overlapping image scenes to extract dense elevation points from the rooftops of buildings, including sloping surfaces and mechanical objects such as vents or air conditioning units. The roof surface, mechanical feature and parapet elevation points were each compiled as separate data layers during this step.
“The building elevation extraction is performed manually because human operators are required to compile all of the different features that define roof surfaces of buildings,” said Cohen.
With the photography orthorectified and the elevation models extracted, Earth- Data then created the 3-D city model for St. Louis using a procedure it developed for its standard product line called SIMmetry. In this process, the firm obtained a GIS layer of the building footprints in the city’s downtown area and matched the photogrammetric compilation data for each building with its correct footprint.
“Here is another aspect of our 3-D modeling technique that departs from a traditional methodology,” said Cohen. Rather than simply extruding the 3-D model of each building from the footprint up to its spot height elevation, resulting in a poorly defined “shoebox” shape, the GIS technicians used a CAD package to build the geometry from the top down. Using this technique with the dense rooftop surface data created extremely detailed renderings, or shells, of the physical traits of each structure.
“The process of building the 3-D models is very manually intensive, but the commercial CAD software is easy to use,” said Cohen. “Creation of the 2.2-mile St. Louis city center model, which included 1,000 buildings, took about three weeks for a small team to complete.” He added that technology now exists to link 3-D models with geodatabases containing information on the occupants, floor plan and design of building interiors, although NGA did not request this for the pilot.
EarthData delivered the first sample image tile to NGA in November 2005, about a month after the acquisition flight. The orthophotos and DEMs were delivered in December followed by the completed 3-D models in February. All of the three deliverables met or exceeded specifications for their mission.
Future taskings by NGA for city models will be based on the results of its official evaluation of this project as well as available funds and needs commensurate with NGA’s mission. ♦
