10 Mar 2009
At got geoint?, we often featured submitted blog posts from industry experts, USGIF members and organizations doing cutting-edge things in the GEOINT arena. Today, we are excited to share a submitted post from Dr. Eiji Yafuso, PHD, President, PSI Origin. PSI Origin is a USGIF member and is an OEM specializing in capturing and transforming Broad Area Persistent Surveillance data into actionable intelligence, bridging the gap from theory to field-level operations. Below is the guest post:
CONVERTING VISUAL DATA INTO A MEANINGFUL PACKAGE
Acquiring visual data is only half the battle……packaging data for meaningful intelligence is the real challenge.
Meaningful intelligence means more than acquiring visual data of a huge coverage area. A minimum Ground Sample Distance (GSD) requires billions of pixels to convert the data into actionable information. Storing and converting the data into actionable information provides real-time processing opportunities such as mapping to a google map with real time movers. This kind of action in real time is ideal for forensic analysis. While this is desirable, packaging the data is extremely difficult. Gigapixel FPA’s are not readily available and packaging their data requires management of an enormous incoming data rate (terabytes/hr). This creates massive processing and storage issues.
The good news is there is a solution. A system approach providing imaging, memory/storage, data management, metadata correlation and a coordination of inertial and GPS data provides the opportunity to convert visual data into a meaningful package.
Due to recognized security needs, Broad Area Persistent Surveillance has come into the common vernacular for military and US Agency applications. Aerial data acquired over large areas for prolonged periods provides analysts the ability to forensically identify activities of interest along with positions, movements and behavior patterns of those involved. In this way, the success of the Army’s Constant Hawk program has led to an explosion in development of systems for acquisition of high-resolution video data covering areas extending to and beyond 10 Km square. Multiple terabytes/hour of incoming data bandwidth is a direct result, making storage of the acquired data over extended mission periods difficult, and access to tactical imagery in real-time even more so. Moreover, modern focal plane arrays large enough to acquire such large imagery in a single frame are expensive and too slow to achieve video frame rates. The solution utilized by leaders in the industry is use of several large area video-rate-capable focal planes which are then combined in a mosaic to form a much larger “virtual” focal plane.
The early lesson learned is that while technology to acquire huge amounts of image data is readily available, (e.g. gang together hundreds of cell phone CMOS cameras), the ability to meaningfully manage the resulting data flow lags behind. The imagery in a virtual focal plane must undergo stringent processing before it can be accurately stitched, derotated, orthorectified, mapped to digital elevation information, etc. The ability to accomplish this in real-time requires highly precise calibration, GPS, and inertial measurement information (e.g. lat/long, altitude, roll, pitch, date, time, etc.), along with the capacity to combine and process the imagery and metadata. Processing of incoming data along with all correlated metadata must occur within the acquisition system and at extremely high speeds if real-time access of any kind is to be realized. Ideally, such real-time processing will allow a ground user access to any part of the acquired video they had the communication bandwidth to explore, and in real-time. This approach, referred to by PSI Origin as the Digital Data Cube, provides bandwidth-agnostic tactical access to the time and space axes of video during acquisition.
Important to note is that the capability sought by the industry cannot find its solution within a neatly delimited body of expertise. As a result various aspects of the solution are being developed by specialists from disparate areas. Large contractors and government labs such as MIT and BAE are leading the development of extremely large virtual focal plane systems, while smaller companies such as PSI Origin lead the development of integrated technology to provide acquisition, real-time processing, tactical data access during acquisition, and smaller, inexpensive tactical systems for persistent surveillance. Furthermore, highly specialized companies such as Pixia Corporation have long been providing unique pc-based solutions for manipulating and viewing these massive video data sets as though one were paging through landscapes on a supercomputer-based video game.
These and others are being brought to bear on the complexities of broad area persistent surveillance. The growth of interest in the industry is spurring innovation into a common and growing area of difficulty in the internet age: how to deal with massive volumes of disparate, yet interrelated data. As such broad area persistent surveillance stands to make significant contributions to the future of high-speed data management.