Computer chip advances and better analysis of images captured by security video systems are two important trends identified by Technical Insights, a division of Frost & Sullivan of San Antonio. The research firm recently completed analysis on the report, Security Technology: Advances in Video Surveillance.

According to Technical Insights, video surveillance applications have been vitalized by the power of digital processing and storage. The move to digital format facilitates the adoption of the Internet for remote monitoring, smart sensors, wireless networks and the integration of intelligent software such as facial recognition and others.

The addition of infrared thermal sensors, movement detection, frame grabbers, spectral analyzers and behavior interpretation has energized the basic security system. Streaming video goes over the Internet and cameras are remotely controlled. Very small wireless camera systems can be anywhere.

Advances in security video technology affect access control and benefit surveillance activities. Such research has resulted in cameras that are smaller than ever before, with superior color capability and crisper images. Higher picture resolution has been achieved through fully digital technology – a new way to record a signal. Beyond these kinds of benefits, there have also been improvements in camera quality and reliability. The main advance here has been the development of CCDs, an imaging process developed at Bell Labs in the late 1960s. The CCD camera produces superior picture quality, is anti-static, and eliminates the possibility of image burning. Such cameras last longer and are also easier to maintain. However, a few manufacturers, primarily those producing cameras for highly specialized purposes, still use the old vacuum type of tube camera.

Cameras and monitors remained monochrome for a long time because they could deliver sharper images and were also cheaper. Now, industry-sponsored research has provided color cameras that are more sensitive, deliver sharper images and are priced more reasonably at between $200 and $400.

Traditionally, video cameras relying on CCD required more chips. Now camera-on-a-chip or complementary metal oxide semiconductor (CMOS) chip cameras have emerged as major competitors to CCD cameras. Also developed at Bell Labs, the camera-on-a-chip approach, based on the same CMOS technology found in today’s computer chips, has the capability to produce real-time video images that rival the quality of images produced by camcorders. Aside from requiring less space, CMOS cameras use less power than CCD cameras. A 9V battery, for example, can run a CMOS camera for five hours, but can run one of today’s computer-based desktop cameras for a maximum of 30 minutes.

As a result, CMOS cameras are best suited to hand-held devices or security cameras requiring low power. And because the camera-on-a chip is an offshoot of silicon-chip technology, semiconductor manufacturers can produce the chip using existing facilities. This was not the situation with CCD technology because it requires specially designed manufacturing facilities. Future uses of CMOS might include 3-D imaging and collision avoidance, such as cameras detecting other vehicles in a car’s blind spot. One day the camera may get integrated into the computer screen and revolutionize the tech world.

The world image sensor market is currently estimated at $4 billion with a 17 percent growth rate. 2006 revenues have been forecasted at $6 billion. The market is blooming as the demand for high-end applications increases. Emerging CMOS technology is also anticipated to contribute substantially to the revenue, its entrance into the market accompanied by automotive and handheld communication applications. CCD is anticipated to provide the foundation for CMOS growth with its maturing market.<

Better than biometrics

Several experiments on facial recognition by video imaging and database search have been disappointing. Yet the use of a full arsenal of biometrics and imaging technologies will no doubt improve these systems. For example, 3-D imaging, color and IR and UV spectral imaging provide far more information about a face than just a video. Hair dye, cosmetics, movements, gestures, padding in the cheeks and temperature variations on the skin can all become part of the equation. The development of software and algorithms that can analyze these data, and make quick and accurate decisions, will be a major activity in the field.

One such system is A4Vision technology, which enables the real-time capture of 3-D images of a subject’s face. Using near-infrared light, an invisible structured light pattern bathes the face while the video camera precisely records the reflected and scattered energy. The distorted pattern is input into a reconstruction algorithm, which extracts a biometric template from the facial geometry. This template is based on the unique rigid tissues of the skull, which do not alter over time. Verification is performed by matching the biometric template, stored in an ordinary database, against a template stored on a smart card.

This A4Vision technology has some advantages over biometric technologies. It is not affected by lighting conditions, background colors, facial hair or make-up, provides higher performance at different view angles, and is more accurate in real-life environments.

The human visual system is highly non-uniform in sampling, coding, processing and understanding. The spatial resolution is highest around the point of fixation (foveation point) and decreases rapidly with increasing eccentricity. Real-time video surveillance may not require maximum resolution of the complete image. A video security surveillance system may have hundreds of cameras, microphones and biometrics stations. The design and development of such a system is quite complex and includes the encryption and decryption process, the image data compression and special data processing like real-time foveation. Currently, most image quality measurement methods are designed for uniform resolution images and do not correlate well with the perceived foveated image quality.

A surveillance system must enable the efficient transmission of events and data of interest from a network of cameras to a central processing unit. For example, a face detection and tracking module that runs at each camera server can find a human presence within an observation area, and provide 2-D face coordinates and an estimated scale to the video transmission module. The captured video is then efficiently represented in logpolar coordinates, with the foveation point centered on the face, and sent to the connected client modules for further processing.

The actual movement of the eye as it focuses on details of an image follows a saccadic path and some imaging systems mimic this movement by small, rapid scanning motions of the sensor. A surveillance system may employ active cameras that track the detected person by switching between smooth pursuit and saccadic movements, as a function of the target presence in the fovea region. The high resolution of the data in this region enables successive application of recognition/identification modules on the transmitted video without sacrificing their performance.