The complete Digital Surveillance System relies on several key technologies:
- Image capture using IP-enabled video cameras
- Image encode, often provided by newer generations of IP enabled cameras to minimize bandwidth requirements
- Image decode for playback
- Data storage and database analysis
- Image and scene analysis
- Threat/anomaly reporting
The first four requirements provide an updated technical equivalent of an analog Video Cassette Recorder. Technological advances provide smaller, lower cost cameras combined with vastly greater record times. What makes digital surveillance essentially different from previous recorders is the ability to perform unattended monitoring that alerts operators to discovered occurrences of defined events.
The long term goal of realtime computer monitoring of video feeds holds the potential to shift the balance towards gathering critical information in realtime with minimal human monitoring. Some types of video surveillance can be automated in realtime today using products such as huperLab’s huperVision 4000 and operates under Linux. The realtime system is used in retail stores to monitor for theft and shoplifting, identify consumer purchasing behavior, and watch employees for fraud and theft. Other uses of the system include: building and security checkpoint monitoring, facilities monitoring against vandalism, and student monitoring.
The most visible portion of the huperLab solution is the video camera. Cameras are preferentially managed through an IP connection such as Ethernet. Until recently, cameras for security systems were closed circuit analog devices that were hardwired or wirelessly connected through a proprietary protocol. Now digital cameras based on an Internet connection have exploded in popularity. A generic Network IP Camera is a stand-alone device which permits viewing live full motion video from anywhere over the Internet, using a standard web-browser. In the case of surveillance systems, connection to the camera is made via IP to receive the video data. Data analysis for surveillance is implemented in a dedicated surveillance-specific networked embedded system that is accessed via a networked connection.
High data rates, or at least high demand for bandwidth requires well crafted IP support. There are a number of methods to achieve minimal overhead processing for IP communications. Real Time Operating Systems (RTOS) vendors often include TCP/IP support optimized for real time operation. Green Hills Software (1) not only provides a TCP/IP stack, they also offer a conversion service to move code developed for another vendor’s RTOS including pSOS, Nucleus, POSIX or Wind River Software’s (2) VxWorks to one supported by Green Hills.
TCP/IP stack optimizations within software are fundamental improvement for network cameras, but purpose-built hardware for surveillance usually relies on hardware for managing TCP/IP connections. For example, Lanner uses up to five Ethernet cards based on the Intel 82574L Gigabit Ethernet controller. Altogether the card stacks can support 40 simultaneous high speed connections.
huperLab’s huperVision operates on a variety of embedded cards including the Gaia 404/408/416 board with Intel Atom processor. Available cards support PCI and PCIexpress card formats. hupeLab also employs high quality video/audio capture chips to improve throughput performance. The Gaia board is the smallest and lowest power server board designed for security surveillance use. The Gaia Server Board based surveillance system reduces electricity expense, consuming less than 60 watts system power.
huperLab is one solution to video surveillance. Intel’s Technology Journal offers a PDF document that includes surveillance as one computer vision application. Titled “Computer Vision Workload Analysis,” the article provides basic computer vision technical information which is just as leading edge as when it was first published, just a few years ago. The document introduces a variety of topics that bear on video surveillance, but doesn’t incude all of the features present in current digital surveillance products . Open Source software offers an alternative for portions of video surveillance software. Sun (3) provides Open Source video surveillance software focusing on the storage aspects of the system. Twenty four other freely available software packages for portions of video surveillance functionality include:
SecureCam for Windows
Electric Eye for Linux
Carnegie Mellon University maintains a web site dedicated to computer vision topics. The compendium of sources includes the CMU source library.
Intel Open Image Processing library
Combining parts of several of these open source packages with your own code may provide a more complete digital surveillance system, but combining disparate software sources is a difficult task. These software sources are probably best as an aid in gaining experience with the technology. You may find that one of these packages meets your objectives. But, a caution when dealing with these open and free sources: video surveillance technology is changing rapidly and these sources may not be current or complete.
Technically, video surveillance is commercially run on an ATOM, Core 2 Duo, and Core 2 Quad embedded processors. More and faster processors don’t necessarily mean faster application execution. Video processing is one application that benefits from the Intel QuickPath Technology. QuickPath provides additional paths between memories. Each core can access shared memory and local memory. As can be seen, the memory controller offers better and faster interconnect between processing and I/O blocks.
Computer ethicists have long expressed concerns and misgivings over the intrusion into personal and private actions. But the uses discussed for automated video surveillance have been in settings previously decided by Courts around the world to be permissible venues for video surveillance. Some professionals believe that computerized video surveillance improves privacy by replacing the continuous monitoring by people with an algorithm operating on an embedded system. In effect, automated video surveillance simply alerts human operators to events that meet criteria requiring their attention.
Where can you use video processing techniques in your next projects?
(1) Wind River Systems is an Associate Member of the Intel® Embedded Alliance
(2) Green Hills Software is an Affiliate Member of the Intel® Embedded Alliance
(3) Sun is an Associate Member of the Intel® Embedded Alliance
Roving Reporter (Intel Contractor)
Intel(r) Embedded Alliance