Don’t look now, but megapixel cameras are taking our industry by storm. According to many integrators I have spoken with, a great deal of their customers have moved to IP — primarily to be able to use megapixel cameras. Why? Greater fidelity, fewer cameras, digital PTZ — the reasons are numerous. IMS Research estimates that, by 2014, half of the network cameras sold will be megapixel. This is just the tip of the iceberg.
First, some perspective. The evolution of video surveillance has been a mix of consumer-market driven developments and home-grown innovations. On the consumer side, MPEG-2 (the basis for DVDs, direct broadcast satellite and studio-quality video), MPEG-4 (borne out of the desire for a low-bandwidth solution for cell phones and limited-bandwidth networks) and H.264-driven enhancements (making high definition deliverable at reasonable bandwidths), have had successive impacts. On its own, the security industry made extensive use of wavelet and MJPEG technology in implementing DVRs and is now using high-pixel-count imaging technology to deliver greater functionality in its next generation of surveillance cameras.
Megapixel cameras can be obtained that provide HD format, alternative formats or both. Pixel count is obtained by multiplying the number of pixels per line by the number of lines. If that result is more than one million, you have “mega-pixels.” While HD represents a 16:9 (1920 x 1080 or 1280 x 720) aspect ratio and is defined by SMPTE standards, there are plenty of megapixel cameras on the market delivering non-HD aspect ratios, usually 4:3 (1600 x 1200, 2048 x 1536, 2592 x 1944, 3628 x 2752). If the pixels are there, you can theoretically derive an HD-compliant image — say 1920 x 1080 from a 2048 x 1536 camera — by selecting a properly sized portion of the image, effectively trimming lines and line widths from the full image.
So what do you do with all these pixels? Clearly, the cameras give you better granularity, making digital PTZ more effective and enabling better image discrimination. Combined with the proper lens system, a camera can provide a variety of coverage options up to 360 degrees. A greater number of pixels provides more information for analysis, creating more opportunities for analytics to extract information from the scene. However, all this comes with a technical price — since it often means reduced frame rates at higher resolutions to limit bandwidth usage. As an example, Arecont Vision’s 10 megapixel camera at 3648 x 2752 resolution is specified to run at an unthrottled data rate of 70 Mbps @ 7 frames per second. The greater the information being analyzed, the greater the need for more processing power to analyze the scene. But as available network bandwidths and processor power climb, surveillance systems will take fuller advantage of these resolutions, likely at higher frame rates, analyzing and storing richer, fuller pictures.
So what’s next? Look again to the consumer market, and you will see that the most likely answer is 3D video. The first TV implementations have used stereo glasses, employing a shutter effect to allow each eye to see a different frame. The brain then processes the information to create the 3D effect.
It is hard to imagine a bank of control operators sitting in their chairs with 3D glasses, and future implementations will not be so constrained. Glasses-free 3D video — called autostereoscopy — is achievable through several competing approaches. For example, parallax barrier consists of a layer of material with precision slits, forcing each eye to see a different set of pixels, but it requires the viewer to sit in a well-defined spot to get the effect. Lenticular lens technology places lenses in front of the pixels, expanding the viewing angle to dictate what each eye can see, and providing more leeway on the viewer’s position. 3D range imaging technology uses an IR illuminator and corresponding sensor, measuring the round-trip travel time of the light beam to determine the distance of an object measured at multiple points, and providing information on distance, height, width, shape. Further, it can calculate direction and speed, while being relatively immune to current lighting conditions. Whatever the underlying technology, 3D will enable more accurate scene reconstruction and expand the scope of video analytic capability.
A 3D approach that does not try to mimic the human eye uses multiple cameras feeding a processing engine for scene reconstruction. This technique is used for production of animated films, capturing actors going through the desired motions; and advanced medical diagnostics, fusing data from different types of imaging instruments to produce a more complete picture of the patient’s area of concern. The science behind this will continue to evolve, leading to increasingly accurate images and better overall scene construction. This is known as multi-sensor fusion — the blending of information from similar or disparate sensors. Another example of this is combining video data with a prior knowledge of the scene using techniques such as geo-registration and Google Earth images. This is not unlike today’s PSIM, where different systems provide input to obtain a more holistic view of a facility or situation.
An area that has always fascinated me is holography. A hologram is constructed by recording interference patterns of reference beams and reflected beams of light. When reconstructed by illuminating the recording media with a reference beam with the same properties, the resulting image is unbelievably realistic and can be viewed from any angle. There’s the promise of holographic TV — which would be 3D on steroids — providing an even more exciting and lifelike viewing experience. To read more, I recommend an article published by MIT and found at http://web.mit.edu/newsoffice/2011/video-holography-0124.html.
I can’t help but think that future video surveillance systems will create a feel not unlike video gaming — an environment where future generations of security professionals will begin to experience video surveillance as teenage kids.
