The field of video surveillance has seen explosive growth in the last 3 years. The convergence of heightened security demand and innovative technology, in the acquisition, transport, analysis and storage of quality video has resulted in a massive deployment of cameras and systems in a number of venues. Major cities, transportation centers, highways, military installations, retail and business centers are all covered by the un-blinking gaze of millions of cameras. According to some reports, the UK alone has over 15 million security cameras.
Digital technologies were the key to this massive deployment since the transport, monitoring and recording of the output of more than a very few analog cameras was clearly going to outstrip the resources and budgets of even the most determined and well funded security organizations. The traditional CCTV systems were expensive and bulky with very limited capabilities for automation of the security function, thus requiring banks of monitors (and security officers) cycling through images from multiple cameras. The surveillance problem can be described as a bandwidth-limited network of sensors tasked with detecting brief events of interest (intrusions, accidents, criminal behavior) and responding to them, against a background of entirely un-interesting footage. Similar to the old “needle in a haystack” analogy, but with the complication that you must watch the entire haystack at all times, because the needle could appear anywhere, including the location you just finished examining.
Fig 1. A networked digital video surveillance system
The design of a modern surveillance camera node requires a number of components in addition to the actual optical and sensor hardware that make up the camera. Typically the sensor contains an ASSP that handles basic camera functions such as auto-exposure. A separate embedded processor provides the power to implement the system level features that define the capabilities and differentiate the system for the end user. A powerful processor opens up the possibility of many advanced features and provides performance headroom for eventual upgrades to new ‘smart camera’ technologies. The system solution will most likely require external flash and DRAM to provide supplemental storage for program code and image processing. A standardized, network interface is also required to transmit images and alarms to the monitoring nodes of the system solution.
Figure 2. A basic ‘smart camera’ node in a
The illustration in Fig 2 shows a very basic block diagram of such a system solution. The processor will likely connect to the video source through a bi-directional peripheral bus for passing image data to the processor and control data to the camera. The memory devices typically communicate through a dedicated memory bus and the flash through a low pin-count SPI-type connection. The network interface may be either an internal or external component depending on the processor choice. Ethernet is the most commonly deployed interconnect for compressed surveillance video, offering the lowest cost implementation for most industrial and commercial networks, while delivering sufficient bandwidth and range for common applications.
The processor selection in the ‘smart camera’ to a large extent dictates the magnitude of the local image analysis. Depending on the specific application and location, this processor may implement image enhancement and compression, detection and tracking, object recognition, etc. Performance is obviously a key selection factor. Since the workload is almost entirely image processing, a processor optimized for handling a number of different video tasks is needed. A flexible processor with the ability to download and execute new application code is a key attribute in this quickly changing market where codecs, proprietary analytical functions, and other additional features can change on a constant basis. Other important selection factors include power dissipation, built in peripheral devices, development tool infrastructure, and available application code. While standard video codecs and filters can often be used in the surveillance market, analytical algorithms are often developed as proprietary value-add features, requiring an integrated set of tools for development and security features in the hardware for protection of Intellectual Property.
Smart surveillance networks will continue to evolve. Market drivers will include expanded distributed processing of video streams, multi-camera tracking through 3D mapping, wireless networks including meshes, and advanced video analytics such as object and facial recognition. Markets will also continue to expand as technology becomes more affordable and total deployed system costs reduce, due to automation of monitoring.
Sidebar - Avnet Introduces Digital Video Security Development Kit