Video analytics and image fusion are key tools in intelligence and surveillance
Technology focus -- Today's video security technology can provide impressive detailed surveillance in many ways, but system integrators are still looking for better ways of interpreting the detailed night and day images these modern tools provide. Video analytics, image fusion, and high-definition capability are just some of the methods they are developing.
Today's video security technology can provide impressive detailed digital video surveillance in many ways, but system integrators are still looking for better ways of interpreting the detailed night and day images these modern tools provide.
Video surveillance for homeland security and military installations consists of much more than a slovenly guard with several monitors and an endless supply of coffee.
The threats the nation faces today are more varied and complicated than ever before and detailed video analysis is needed in addition to the naked eye watching the video monitor.
There are many different tools for surveillance users, such as infrared -- short wave and long wave, image fusion, satellite links, and video streamed from unmanned aerial systems (UASs).
The way to bring all that together is through video analytics, says Charlie Morrison, director of full motion video solutions at Lockheed Martin Information Systems and Global Services in Gaithersburg, Md. It is the idea of tracking objects and performing object recognition, he explains.
"There is so much information out there," that it needs to be organized and property analyzed for the surveillance operator so he is not overloaded, Morrison continues. This technology is already out there in the commercial world, he says.
Tomorrow's video surveillance screens will resemble the television feeds of ESPN and CNBC where sports scores and stock tickers are streaming across the bottom of the screen, Morrison says. There will be "multi-intelligence collected within the video so that it will be more than a picture."
Morrison's team at Lockheed Martin along with engineers at Harris Corp. in Melbourne, Fla., are taking that technology and applying if to military surveillance and intelligence applications, Morrison says. "We're looking to integrate commercial technology for DOD video."
"Full-motion video has exceptional potential for intelligence collection and analysis," says Jim Kohlhaas, Lockheed Martin's vice president of spatial solutions. "Thousands of platforms are collecting important video intelligence every day. The challenge is to collect and catalogue that huge volume of footage, and give analysts the tools they need to find, interpret, and share the critical intelligence that can be gleaned from that mountain of data."
The main Lockheed Martin video analytics tool is called Audacity. It tags sorts, and catalogues digital footage. According to a Lockheed Martin public release it also has intelligence tools such as video mosaic creation, facial recognition, object tracking, and smart auto-alerts based around geospatial areas of interest.
Harris brings to the effort its Full-Motion Video Asset Management Engine, or FAME, which integrates video, chat, and audio directly into the video stream, according to the Lockheed release. The too, which is used by the broadcast industry, forms a digital architecture from an integrated processing and storage engine to provide the infrastructure for enhanced video streaming, according to the release.
"What we've done is take the best of FAME and Audacity and combine them together to that point that you cannot say which is doing what," Morrison says.
Lockheed specializes in the "filtering of information from multi-intelligence sources" to analyze video and provide object recognition. The joint effort will focus on research and development of video capability and focus on analysis in real-time and through archival footage, according to the Lockheed Martin release.
The team will also develop solutions for "cataloguing, storing, and securely sharing video intelligence across organizational and geographic boundaries to include bandwidth constrained users," according to the release.
More and more assets are being put out there, "infrared, day/night, vision, object recognition, etc., and we are providing the filter, Morrison says. The software will organize it for the operator to make his decision and reaction processes quicker and more efficient, he adds.
"We already have done a deployment" Morrison says. However, he declined to name where it was deployed or who is using their solution.
Improving decision time
The joint effort will bring surveillance operators ease of use through graphical user interfaces, Morrison says.
In a port security application for example, operators will be able to access on their video screen not only the ship coming in, "but what it's supposed to be carrying and what it actually is carrying," says Tony Morelli, program manager at Lockheed Martin Information Systems and Global Services.
It will also provide a chat have, which could include an audio chat between two commanders or operators chatting within the video stream, Morrison continues. The chat will be seen as have within the mission video stream, he adds.
Archiving today for video involves chopping up pieces of the video into 30-second to two-minute bits, Morrison says. This can take a long time in retrieval, especially when you are looking for a two second frame, he adds.
It is akin to searching for a sentence in a search engine that only gives you the 10,000 word document the sentence is located in, he notes.
Meta tagging the data is much faster, Morrison says. For example imagine a red Ford Mustang that is sitting outside a building, he says. An operator can search to see where that red Ford Mustang was three days ago, and then play out the video before and after the car appeared -- much the same way your digital video recorder or Tivo allows you to go back and see what you missed" when stepping away from the TV.
This an example, of geospatial awareness, Morrison continues. The system conducts temporal (time), keyword, and geospatial searches, he adds.
Morrison notes that their solution does not search for spoken words, but rather for the transcribed texts of the audio data. Morrison elaborates with another red Ford Mustang car example. An operator can "search for any and all references to the red Ford Mustang he sees on his screen for the last 10 days and up it will come, he explains.
Down the road the goal is to get more interactive with the video with touch screen capability, Morrison says. Operators will be able to touch a piece of data in the stream to pull more detailed information -- whether it is a person of interest or a car or even a chat between two commanders in the field.
One of the advantages "our solution will have is its ability to work with a variety of surveillance tools," Morrison says. The Army, Navy Air Force, and Marines all use different tools to exploit information and the joint Harris/Lockheed Martin solution can work with all of them, he adds. "The key to this is standardization," Morrison says. If the users all use the same standards of metadata it easy to integrate no matter the tools involved, he explains. The video standards are set by the Motion Image Standards Body or MISB, Morrison notes.
The touch screen capability is still years away, but in the background the system will flash information it thinks the operator may need based on his tendencies such as past searches as well as what he is seeing on the screen, Morrison says. This is similar to how web sites like amazon.com analyze the buying purchases of their customers to push other items to them to buy, he continues.
Panoramic image fusion
Extracting and analyzing video data will be more efficient if the video is actually capturing everything in view. Engineers and scientists at GE Fanuc Intelligent Platforms (Bracknell) Ltd. (formerly Octec Ltd.) in Bracknell, England, are looking at ways to create a panoramic view that processes video the way the human brain processes images seen by the human eye.
It is a matter of presenting it to the human eye in a physical manner because the human eye reacts to physical movement, says Larry Schafer, vice president of business development for GE Fanuc Intelligent Platforms Bracknell. Currently most systems have so much symbology in displays; the operator is overwhelmed and cannot respond effectively.
GE Fanuc engineers figure if they can present the image in a more physical manner with less symbology it will be easier for the operator to determine what he is saying and make the proper decisions.
Schafer and his team are looking to re-create the way this information is presented by creating an array of cameras with several sensors active all the time, he says. "It is called distributed aperture sensing," Schaffer adds.
In a way it is not unlike a panoramic zoom camera, Schaffer says. The operator gets views from behind his head if he is out in the field in a tank or other vehicle, he adds.
The imagery is a form of image fusion, Schaffer says. Day and night imagery is overlaid in a 160 degree or 360 degree visual, he adds. Resulting image is much more contrast rich, thereby providing the eye with the necessary physical stimuli, Schaffer says.
This then "gets us back into data extraction since all sensors active all the time" essentially give the operator eyes in the back of his head, he continues. If there is an object of interest at his rear he should be able to hand touch a display screen whether in a building or driving and get that data in real-time, Schaffer explains.
"This creates a possible 3D environment," Schaffer says. Peripheral vision becomes just as clear as what is seen straight ahead, he adds.
The video processing is enabled by GE Fanuc's IMP20 video processing mezzanine card, which is an add-on module for the company's ADEPT 104 and AIM12 automatic video trackers to provide the cards with a supplementary image fusion capability, according to a GE Fanuc data sheet.
The device's image fusion algorithm produces faster execution times and reduced memory overheads, according to the data sheet. The IMP20 also has a "built-in warp engine that provides rotation, scaling, and translation for each video source to compensate for image distortion and misalignment between the imagers, reducing the need for accurate matching of imagers" while reducing total system cost.
It is true situational awareness enabled by overlaying thermal imagery with day pictures in image fusion then adding object recognition capability, Schaffer says. The technology has a lot of the same has of short wave infrared technology, which has been around for a long time Schaffer says.
While shortwave infrared (SWIR) sensors have been around for a long time, this technology is still in demand for mission-critical surveillance applications.
"The SWIR band has been shown to be valuable for imaging through atmospheric obscurants like fog, haze, dust, and smoke," says Robert Struthers, director of sales and marketing at Goodrich ISR Systems (formerly Sensors Unlimited) in Princeton, N.J. "Fusing SWIR with long-wave or mid-wave thermal cameras is useful for driver vision enhancement, airborne wide area persistent surveillance, and enhanced soldier portable night vision imaging systems. Fundamentally, the thermal bands detect and the SWIR band identifies.
"High performance SWIR cameras will image urban night scenes without blooming or washout from intense light sources," he continues. "Because of the longer wavelength sensitivity than visible cameras or night vision goggles (NVGs), InGaAs (Indium Gallium Arsenide) cameras combined with wavelength matching and covert illuminators can enable soldiers to avoid threats from NVG equipped subjects.
SWIR is also excellent for imaging during thermal crossover (near dawn/dusk), he adds.
"The SWIR spectral band produces images from reflected light, not temperature as thermal imagers," Struthers explains. "These images are more recognizable than thermal spectral bands because they appear much like monochrome visible images providing target/biometric identification capability.
Goodrich's SWIR indium gallium arsenide (InGaAs) cameras "provide the highest resolution available in the SWIR band; 640-by-512 pixel arrays," Struthers says. "Most camera models are available in enclosed housings or as miniature open-frame modules for embedding into surveillance gimbals and small optical systems."
Goodrich's latest InGaAs camera is the SU640KTSX-1.7RT, which has high-sensitivity and a wide dynamic range. According to a Goodrich data sheet it provides real-time night-glow to daylight imaging in the SWIR wavelength spectrum for passive surveillance and use with lasers. The camera's onboard Automatic Gain Control (AGC) enables image enhancement and built-in non-uniformity corrections (NUCs), according to the release. The Goodrich device also has an extended lower wavelength cutoff, enabling the camera to capture photons previously only possible with silicon-based imagers.
Funding from the Defense Advanced Research Projects Agency (DARPA) in Arlington, Va., is fueling Goodrich's SWIR camera developments "for drastically reducing the size, weight, and power)" for small UAS platforms and soldier portable night vision components, Struthers says.
"The proliferation of SWIR cameras has spurred development of covert illuminators and specialty AR coated lenses to optimize night vision capability for the warfighter," Struthers adds.