• Font Size:
  • A
  • A
  • A

Feature Articles

Security and Machine Vision Merge in the Infrared

by Winn Hardin, Contributing Editor - AIA

This is Part 2 in a series of articles on Infrared (IR) imaging. This article looks at the growing relationship between security and machine vision as illustrated by new data fusion, global coordinate registration, and visualization techniques that build on similar techniques used in robot vision. Part I looked at how integrators are using advanced filtering systems and contour geometry to enhance industrial IR inspection systems.

Until recently, when you thought security, you didn’t think machine vision. Then came 9/11, and with it, a whole new worldview.

Security isn’t just to counter corporate espionage or a checklist item for a lower insurance policy premium. Security has to be effective. The days of simply putting up a number of cameras with the lowest possible labor investment is no longer a viable answer for chemical, financial or technology companies, or for national civilian and defense infrastructures.

Today, security systems have to be able to detect intruders at a distance that will actually allow preventive measures to be taken. Arresting the culprit after the terrorist bomb has exploded is no longer a hedgeable position because the likelihood of such an incident has increased considerably.

Now that security faces the same efficiency metrics as say, industrial production in which defects lead to unacceptable loses, security is looking to machine vision and its methods to increase the capability of a human security guard in consideration of his or her average focused attention span.

Why Security Likes IR
‘‘Studies show that the average security guard has a 3- to 5-minute attention span before their attention is drawn to something else,’‘ explains Glen Francisco, product manager of L-3 Communications Infrared Products. ‘‘The security industry is automating the detection process to fix that through a suite of video analytics that support automatic alarm overlay, camera identification, and more. But to date, we haven’t combined the best of the world’s intelligent analytics with the capability to detect intruders at long ranges [in non-military applications].

‘‘If you’re in a security environment, and you’re trying to watch and monitor things,’‘ Francisco continues, ‘‘you want to proactively react rather than forensically react to determine what you would do different next time. To proactively react, detection range is a necessity, and visible surveillance can’t see far at night without giving itself away through active illumination.’‘ L-3 Communications Infrared Products, previously Texas Instruments, then Raytheon before being acquired by L-3, produces uncooled long wave infrared (LWIR) sensor technology. L-3’s Cincinnati Electronic facility produces cooled mid-wave infrared (MWIR) systems for the U.S. military used mainly for Littoral environments where water meets land and fog and rain are common.

Visible CCD cameras have been the mainstay of security applications because they have historically been cheaper than IR cameras, however, visible light cannot penetrate fog, smoke and other obscurants with the efficiency of longer-wavelength IR radiation. The same principle also gives IR cameras the ability to see through thin films and coatings, such as paint used to obscure marks on shipping containers or vessels, according to SUI Goodrich Corporation’s Douglas Malchow, manufacturer of near infrared (NIR) and short-wave infrared (SWIR) InGaAs cameras.

This ability to penetrate over distances can actually reduce the total cost of ownership for a security system when comparing IR cameras to visible cameras. At night, passive uncooled LWIR cameras can see 3000 ft or more without the need for active illumination while visible cameras can only see about 200 ft with active illumination. Uncooled LWIR cameras may cost more than a simple CCTV camera, but 2000 meters of fence line would require 5 to 10 times the number of visible cameras compared to one LWIR camera, in addition to the poles, wires, monitors, illuminators and personnel needed to support the extra visible cameras.

‘‘Yes, the purchase price of an uncooled LWIR camera is higher [than CCTV], but the cost per foot of surveillance goes down as you move to passive thermal cameras’‘ notes L-3’s Francisco.

True Data Fusion for Security
While it’s true that thermal cameras can see just as well at night as they do during the day, IR cameras are not likely to completely supplant visible CCTV security systems. As traditional security is enhanced with longer-range IR security systems, the challenge becomes how to efficiently display and analyze the various data streams.

Data fusion, or the intelligent combination of different sensor data into a data set that represents more than the sum of the parts, has traditionally been confused with data overlay with the result of more hype than benefit. Data overlay is simply where two fields of view are closely matched and displayed simultaneously.

MWIR vs LWIR: A Tale of Water and Cooling

The choice of uncooled LWIR versus cooled MWIR cameras is one of price, performance, and application.

LWIR cameras containing microbolometer arrays made of vanadium oxide (VOx), barium-strontium-titanate (BST), or amorphous silicon (a-Si) are sensitive to radiation in the 8 to 12 micron range. MWIR cameras made of arranges of quantum well infrared photodetectors (QWIP), indium antimonide (InSb), mercury-cadmium-telluride (MCT), and lead salts (PbSi), can be sensitive to light in either the 3 to 5 micron range or in the 7 to 14 micron range, with lead salts climbing up to 30 microns and beyond, mainly for chemical analysis and spectroscopy.

The application (or how the imaging system is to be used), defined by what is being imaged and the distance of the object from the camera, along with the atmospheric water absorption of IR radiation, all guide the selection of MWIR versus LWIR. Infrared radiation in both the midwave and longwave bands are less susceptible to absorption by water particles in the air, however active heat sources like humans and vehicles, at ranges historically under 2 miles, ‘‘leap’‘ out using uncooled LWIR cameras because of how their heat signatures interact with the atmosphere, resulting in a strong signal at the detector.

To detect faint heat signals, at ranges beyond 2 miles, MWIR cameras use active cryogenic coolers (picture if you can a miniaturized version of your home heat-pump HVAC system) that cool the detector focal plane array to between 80 and 100 degrees Kelvin (K), while uncooled LWIR cameras operate without an active cooler at between 180 to 250 K, or approximately room temperature. Both active cooled and uncooled camera technologies have their advantages, depending on application. The ability to detect unique IR signatures that are important to defense applications, along with the significantly higher cost of cooled MWIR cameras makes these systems historically well suited to long range (greater than 2 miles) military and airborne applications. The ability to detect IR signatures that are important to commercial applications, along with significantly lower cost of uncooled LWIR cameras, make these systems historically well suited for machine vision, corporate surveillance, border, transportation or critical infrastructure applications.

Companies like L-3, however, are developing true data fusion algorithms that evaluate each scene on a pixel by pixel basis, and displays only the pixels from sensors that have the best data compared to all the rest of the sensors. Instead of simply overlaying data, the new methods make value judgments about the weighted importance of each pixel’s piece of data and display it in an intuitive representation. 

Sensor Level Detection Table

Data Fusion Picture

‘‘If you overlay two different video images you may combine a red and a green pixel and get a messy purple pixel as a result, but if you do a pixel by pixel image fusion, where you take the best that one wavelength has to offer, perhaps visible for spatial resolution of facial detail and IR to see through a smoky area, then you have the power of image fusion,’‘ Francisco explains.

Throughout much of machine vision’s early years, thermal imaging has suffered from the return on investment (ROI) argument, but in today’s changing security environment, and given the drop in prices uncooled LWIR cameras, fiscal momentum has switched in favor of the IR vendor and customer. ‘‘What drives cost is not just the technology, but also volume and as we get into big volume commercial markets, like automotive manufacturing (L-3IP was the first company to put a NightDriver passive uncooled LWIR system on a commercially available automotive platform), costs will be driven down. Ten years ago, you couldn’t get an IR camera under $18,000. Now it’s under $10,000, and we can expect that trend to continue as new technologies come along,’‘ said Francisco.



There are currently no comments for this article.

Leave a Comment:

All fields are required, but only your name and comment will be visible (email addresses are kept confidential). Comments are moderated and will not appear immediately. Please no link dropping, no keywords or domains as names; do not spam, and please do not advertise.

First Name: *
Last Name: *
Your Email: *
Your Comment:
Please check the box below and respond as instructed.

Search AIA:

Browse by Products:

Browse by Company Type: