What's New At The Edge

How network cameras are reshaping the surveillance landscape

• By Fredrik Nilsson
• Jun 04, 2012

Even though the first product launch of a network camera happened in 1996, it was a little more than a decade ago before the device began appearing in the physical security arena. Crude by today’s standards, the pioneering technology boasted the ability to stream video at a sluggish one frame per second. Even more, the cameras required a minimum of 20 lux to deliver any sort of image clarity. Nowadays, network cameras stream HDTV-quality video at speeds up to 60 fps, can operate in lighting conditions as low as 0.008 lux and offer a long list of other features.

So, how did IP-based cameras achieve such a quantum leap in performance in such a short time frame? Like most other computer technology, network camera performance follows Moore’s Law, which describes the trend where digital electronic devices double in power and speed every 18 months. In the case of network cameras, this trend specifically encompasses processing performance, image sensors and pixel count. With so much computing power now residing in cameras, manufacturers have been able to push more processing out to the edge of the surveillance network to provide better image quality, better scalability and greater functionality at an overall lower system cost. But how have these performance advances affected surveillance in the real world? Let’s look at six key features.

Higher Resolution and Frame Rate

Higher resolution means capturing more details in the image area and therefore increasing the forensic value and usability of video. True analog systems are limited to the NTSC/PAL standard, which means their maximum resolution is 720x480—corresponding to 0.4 megapixels—and often only a quarter of that resolution is recorded. However, in the IP world, camera resolution has undergone an exponential evolution to megapixel and HDTVquality image clarity even faster than Moore’s Law predicted.

Megapixel-resolution cameras first appeared around 2005, providing more detail for identifying people and objects and covering a larger field of view. The first megapixel cameras offered 1280x1024-pixel resolution, basically a scaled-up version of VGA with the same 4:3 aspect ratio. But as the technology improved from 1.3 to 5.2 megapixels, resolution jumped to a 2560x2048- pixels format. The drawbacks came in two forms: the aspect ratio didn’t always match new 16:9 monitors, and the frame rate for higher megapixel cameras was limited to around 10 fps for 5 megapixels and even lower for higher megapixel cameras, such as the 8- and 10-megapixel products now on the market.

Like the megapixel cameras, HDTV cameras deliver much higher resolution than VGA cameras. Unlike megapixel, for a camera to call itself HDTV it must strictly adhere to SMPTE standards for resolution, frame rate (30 fps), color fidelity and 16:9 aspect ratio. Like the TVs we buy for our homes, this ensures that every camera classified as HDTV will deliver consistent performance no matter who manufactures it.

In addition, HDTV-compatible cameras support advanced H.264 compression technology, which drastically reduces bandwidth consumption and storage requirements. HDTV network cameras come in the same three formats as flatscreen TVs today: 720p (1280x720 pixels), 1080p (progressive, 1920x1080 pixels) and 1080i (interlaced, 1920x1080 pixels).

Better Video Compression

Advances in compression standards, along with improved processing power at the edge for real-time compression, have also evolved to significantly reduce image file size—and therefore bandwidth consumption and video storage—without adversely affecting visual quality. In other words, without H.264 compression, HDTV-quality video wouldn’t be possible in the surveillance world. The compression standards evolved to focus efforts on frame consistency for better efficiency, such as reducing color nuances and color resolution, removing small invisible parts of the picture, and comparing adjacent images and removing details that are unchanged between video frames:

Motion JPEG treated each frame as a still JPEG picture. It prevented dropped frames during transmission, but the compression ratio was low for video sequences because it made no use of video compression techniques.

MPEG-1 used a more efficient coding of video sequences, but the focus was on compression ratio rather than picture quality.

MPEG-2 employed more advanced techniques to enhance video quality through resolution and frame rate, but it was done at the expense of higher bandwidth usage. MPEG-2 is used for standard-definition DVD movies.

MPEG-4 accommodated both ends of the spectrum, streaming lower-quality video to mobile devices requiring lower bandwidth consumption and streaming extremely high quality for applications with almost unlimited bandwidth. The MPEG-4 standard has multiple parts.

H.264, aka MPEG-4 Part 10, is the newest video compression technology in IP video, representing a huge step forward for video surveillance applications. Without compromising image quality, H.264 can reduce the size of a digital video file by more than 80 percent compared with Motion JPEG compression and as much as 50 percent compared with the MPEG-4 Part 2 standard. With far less network bandwidth and storage space required for a video file, users save money and achieve a much higher video quality for a given bit rate. This advanced compression standard is being used in the entertainment industry for Blu-ray movies and online video.

Greater Light Sensitivity

While higher resolution and more effective compression have a major impact on image quality streaming from the camera, image processing technology also plays an important role, especially in difficult lighting conditions. Network camera manufacturers today have greatly improved a camera’s ability to capture quality images in fairly complex lighting conditions—from very low light to wide variations in light throughout the day or within a single scene—and have surpassed analog in light performance.

Low light. In the past, manufacturers have addressed low-light problems by integrating more light-sensitive sensors, day/night filters, IR illuminators and thermal imaging into their cameras. As new cameras have come on the market with higher processing power, manufacturers can employ even more advanced filtering techniques to further improve light sensitivity.

Lightfinder technology is the latest innovation in extremely low-light surveillance. It works in concert with a network camera’s sensor and lens to find light in a scene that it can use to stream color video even at night. Sophisticated image processing software sets the degree of filtering and sharpening to capture the best image possible. Highly sensitive to low light, a network camera enhanced with Lightfinder can maintain tight focus with minimal noise and lifelike color fidelity from dusk to dawn as well as in full sunlight.

Wide dynamic range. WDR incorporates techniques for handling a wide range of lighting conditions within a single scene, such as extremely bright and darkly shadowed corners or backlit situations where a person is standing in front of a sunlit window. A standard surveillance camera would inevitably produce barely visible images of objects in dark areas. A network camera equipped with WDR, on the other hand, combines different exposures of different objects in a scene, depending on the prevailing light to ensure nearly uniform visibility across the field of view.

In-camera Intelligence

With the convergence of improved image quality and sufficient processing power, manufacturers have started to incorporate intelligent algorithms in-camera to push video analytics to the edge. The power of the latest chipsets have made it possible for network cameras to detect motion, sound and tampering attempts, like blocking or spray painting the lens; recognize license plates; and identify objects crossing an imaging line.

Intelligent network cameras also can count people, perform dwell time analysis for retailers and even track customer flow through the aisles of a store. This is all being done in the camera today. Some of the more advanced motion detection analytics can also filter out the natural rustling of leaves and the swaying of branches for better success rates.

Because video analytics often require very specific knowledge about the surveillance application, camera manufacturers typically partner with expert software companies. The additional processing power built into the camera makes the edge an attractive and robust platform for third parties to develop any number of custom analytics applications—think of it like an App Store for surveillance.

There are several advantages to performing analytics at the edge. First, raw and uncompressed video contains more information that can be used in an analysis. Second, analyzing the video before compressing it and sending it over the network reduces bandwidth consumption. Third, in-camera analytics provide better system scalability because they avoid overloading a central server with too many video streams requiring analysis.

Local Storage Option

Advances in SD memory card technology, formerly found only in consumer electronics, have created new possibilities for storage at the edge.

A few short years ago, a 1 GB card could cost upward of $100. Today, a 32 GB card can be purchased for less than $50. SD cards with the potential to hold upward of 2 TB of storage are already on the horizon, which could equate to years’ worth of video storage at the edge.

Applying H.264 video compression, a customer can now record 15 images per second of high-quality, 1080p HDTV resolution for days and even weeks on a single card. A network camera will be able to offer a level of fault tolerance for network outages by recording locally—even for security applications that require high resolution and real-time recording rates.

The industry is taking this one step further. While SD cards traditionally were used for redundant storage in critical surveillance applications, this year at ISC West we saw manufacturers and software developers leveraging the IP camera as the recorder. Individual cameras with a single switch or router can now become complete security systems unto themselves without the need for central storage or even a computer running the system. This camera-as-the-recorder model will be a major trend for moving IP video into small-camera-count installations by eliminating the cost for recording hardware.

Smaller Form Factor

The miniaturization of integrated circuit technology has allowed manufacturers to deliver more processing power in a smaller chipset. The smaller chips generate less heat, a primary culprit in picture noise.

But better image quality is only part of the story. Smaller, more powerful chips allow network camera manufacturers to downsize their camera form factors while maintaining the same capabilities as their larger cousins. Today, a palm-sized PTZ IP camera can discretely monitor a retail store, bank or hotel lobby without being an obtrusive menace to aesthetics—yet it will deliver the same HDTV-quality and intelligent benefits as the largest cameras today.

Where Moore’s Law Will Lead Us

Based on past progress, industry experts foresee network cameras maintaining the same forward trajectory as other computer technology. The degree of light sensitivity will become even more acute while resolution and compression will continue to improve across an ever-wider dynamic range.

In addition, with advances in chip technology and processing power, the potential for third-party development of video analytics applications will become even more prevalent. If we accept Moore’s Law as an accurate predictor, improvements at the edge will continue to grow exponentially for the foreseeable future.

This article originally appeared in the June 2012 issue of Security Today.