Photo: Tim Dees

Photo: Tim Dees

A "police science" textbook in use during the 1970s listed among the necessary tools for a surveillance assignment a pair of binoculars, a thermos of coffee, and a mayonnaise jar for the "used coffee." In that era, the author assumed his only readers would be men. The game has become a little more sophisticated since then. Cops still sit in cars watching houses and businesses, but they are just as likely to be using clandestine tactics to see inside buildings and over fences, scan faces in a crowd, and review security camera footage.

Sometimes it's not possible to conduct surveillance without being seen, and sometimes it doesn't matter. Organized crime "wiseguys" have been known to be on a first-name basis with their watchers, especially when a surveillance goes on for months or years. Like the cop on the street corner or a parked patrol car in a school zone, the perception of being watched can deter crime. Even so, most surveillance is covert, so that the watched will let down their guard and leave an opening for discovery of their misdeeds or a breach of their sanctuary.

The newest and most innovative surveillance tools work without the bad guys ever knowing they are being watched until it's too late.

Seeing Through Walls

Recent research at the Massachusetts Institute of Technology (MIT) developed a radio-based system that detects movement through walls as much as eight inches thick and 20 feet away. S-band radio waves, similar to those used in wireless Internet networks, move through an array of 21 antennae (13 transmitting and eight receiving) aimed at the barrier the watcher wants to see through. The reflected radio energy still displays the barrier as the brightest point, but the waveform of that return is a function of the distance between the antenna array and the wall. A wall 20 feet away might show up as a 20KHz sine wave, while objects on the other side of the wall show as a 30KHz sine wave. By suppressing the lower frequency return with an analog crystal filter, objects on the far side of the wall show up on the detector's display.

This technology is new and the discrimination between objects is fairly crude. Output consists of video at 10.8 frames per second, and the system is able to detect only movement by comparing one frame to another. Experimenters can show two humans moving behind concrete and cinder-block walls, and a person swinging a metal pole, both from about 20 feet away. The biggest downside thus far is the size of the device. The antenna array is 8.5 feet long, so it requires a good-sized truck to move it. Refinement will almost certainly reduce the device to a more manageable size.

A different approach by researchers at the University of Utah uses radio transmitters deployed around, on top of, and even within a building to detect movement within the building. Tomography is the technique for producing focused images of structures at specific depths within an enclosure by the use of reflected energy. It's the "T" in a CAT scan (for Computerized Axial Tomography) used to detect injuries or growths inside the body. This passive detection system is called variance-based radio tomographic imaging (VRTI). The scientists who developed this technology visualize a process of throwing, launching, or dropping small radio sensors throughout a structure, which would form a mesh network and report their location to a central monitor against a database from Google Maps or a building diagram.

As people move through the network, they disrupt the radio signals between the sensors, and a measureable value called the residual sum-of-squares (RSS) variance (don't ask-the math is deadly) produces an image of the motion path. The effect is to show movement as if the people moving were carrying tracking devices, only without the tracking devices.

This system can be pre-deployed in high-security buildings or where there is a special need for life safety in a fire situation, or deployed on the go when operators needed to detect movement in a building that hadn't been prewired. The sensors would establish a coarse network by placement around the perimeter and on the roof, and searchers would drop more sensors as they moved through the building, refining the network and giving greater precision to location monitoring outside. Proof-of-concept was demonstrated at the home of one of the researchers, using 34 nodes distributed around the perimeter of a 30-by-30-foot enclosure.

If you want to work a little closer to the action, the Prism 200 through-wall radar unit from Cambridge Consultants is about the size of a small backpack, with rigid handles where the shoulder straps would be. The operator holds the flat side against a wall and sees a display with blips indicating any moving objects on the other side of the wall. The manufacturer says the device will detect anyone up to 20 meters away from the base wall, with a field of view 120 degrees horizontally and 90 degrees vertically. A similar line of through-wall imaging systems is the Xaver series from Camero.[PAGEBREAK]

Another such device is the Range-R from L-3 Cyterra. This handheld sensor is powered by four AA batteries and weighs less than 1.5 pounds, making it easy to transport and maintain. You hold it up to a wall and it sends radar signals through to locate people on the other side. The Range-R acts as a Doppler motion detector using L-3's patented stepped frequency continuous wave (SFCW) radar technology and proprietary target detection algorithms. It detects any moving object within a conical field of view of 160 degrees in one scan, which takes only a few seconds. And "moving" includes breathing, so it's difficult to hide from.

Remote Control Robots

There are times when what you need to see is out in the open, but you can't expose yourself to look without undue risk. This is a common problem for troops in Iraq and Afghanistan, where snipers and improvised explosive devices (IEDs) make hazardous duty of strolling around in the open. Many of the proposed solutions to this problem aren't far removed from radio-controlled toy trucks, with some improvements in endurance, range, durability, and the addition of a TV transmitter and camera.

One of the least-intimidating of these devices is the Recon Robotics' Scout Throwbot, which is the shape and size of a two-pound barbell. The "weight" portion of the barbell is actually the device's wheels and motor, and the camera, batteries, and antennae are on the connecting "bar." A stabilizing tail keeps the camera pointed forward and the antennas pointed up.

The whole device is 7.4 inches wide and three inches high, and weighs only 1.2 pounds, and it really is sturdy enough to throw into any environment you can walk into and be able to walk out. Its transmitter/receiver has a range of 100 to 300 feet and it moves at about one foot per second. The turnkey package, including control unit, chargers, and everything else you might need is $4,875.

If you would prefer a bird's eye view of your objective, there are quite a few radio-controlled aircraft capable of carrying TV cameras and transmitters. Quad-rotor aircraft use four helicopter-type propellers at the tips of a crossbar framework, with the props spinning in opposite directions to counteract torque. These take some practice to control precisely, but they're easier to master than traditional helicopter models, and they can hover as long as their battery will allow.

The Japanese Defense Ministry recently demonstrated a spherical aircraft that was built for about $1,400, using off-the-shelf components from hobby shops. The craft is a little larger than a basketball and is gyro-stabilized. In the demonstration video, the operator gives the hovering ball a shove and it returns to where it was, as if on an elastic cord. It was still able to dive and swoop over the indoor audience on command, and being round, it can land on any surface, right itself, and take off again unassisted. The effect looks like an anti-gravity device from a science-fiction film, but it's real and it carries a camera and transmitter.

Facial Recognition

It's great to be able to monitor large areas from on high, but seeing too much can be as much of a problem as seeing too little in some cases. Officers watching a crowd to detect the presence of a bad guy or a developing problem have difficulty discriminating between the innocuous and the important. Technology to help with this problem is being evaluated by the Brazilian Policia Militar in preparation for the World Cup soccer games scheduled for 2014.

With this technology, officers wear protective eyeglasses similar to safety glasses, with one eye covered by a camera and head-mounted display. The camera and attached wearable computer continuously scan the crowd for faces, capturing up to 400 per second at distances as great as 50 meters. The scans are compared to a database of 13 million faces. When the computer detects a match to a wanted person, the display turns red and alerts the officer. While some people object to this kind of "Big Brother" surveillance, it's still less intrusive than stopping everyone entering a venue to ask for identity documents.

The close-but-no-cigar accuracy of facial recognition systems has limited their practical applications. For instance, In a test during Super Bowl XXXV, Tampa Bay police were able to match only 19 faces captured by security cameras to records in their own system. Since then, computers have gotten faster and the data has gotten better. A system developed at Carnegie Mellon University is able to match faces from random photos on the Internet to those in Facebook and other social media profiles in under a minute, with no other data points. Take a moment to consider that. If there is a photo out there, anywhere, that has your name associated with it, it's now possible to go from random-person-in-the-crowd to you in under 60 seconds. That's powerful or scary, and maybe a little of both.

Blur Correction

Whether you're using facial recognition or not, you need an image to be clear before you can do much of anything with it. Images from surveillance can usually be fixed if they are too dark or overexposed, but blurred pictures are useless-until maybe the next edition of Photoshop. A new feature demonstrated at a recent Adobe users conference can correct camera shake-induced blur and restore most of the detail to the image.

The correction works by first analyzing the image to determine the "blur kernel" that caused the distortion. If successful, it uses a regression formula to reconstruct the original image. The effect is dazzling, even more so if you've ever tried to salvage a picture that had this problem. The developer of the technique has some test images posted on his Website, including a famous photo taken on Normandy Beach on D-Day.

You may never see some of these technologies in use, and others might become so commonplace that we'll forget about the time when we didn't have them. Ten years ago, most of us were still using film cameras and Eastman Kodak was a strong company. Ten years from now-who knows?

Tim Dees is a retired police officer and the former editor of two major law enforcement Websites who writes and consults on technology applications in criminal justice. He can be reached via [email protected]