Digital Video Motion Detection

Sept. 26, 2004
If you've considered video motion detection in the past and passed it up, you might want to look again.
It seems like just yesterday that I was in the process of trying to convince a customer that there was nothing wrong with his video motion detection system. He had 15 outside cameras, each on a pan/tilt system, each in auto-pan mode, each connected to the analog video motion detection system. He was told that the system would detect the motion of someone walking while the cameras were panning back and forth. What a hoot! Can you imagine? There was already motion in the image, due to the auto-panning of the camera, and yet he wanted to be able to detect a man walking independently at the same time. Couldn't be done! At least not 26 years ago.

Enter advanced digital video motion detection (DVMD). Twenty-six years in the making and the explosion is getting louder. Now I realize that the entire industry has been pumping the heck out of digital video recorders. Want one? Go to the corner and buy one—God knows there are enough of them out there. On the other hand, I don't hear a lot of fuss being made over DVMD, and I feel like I should. This stuff is hot. Between IP cameras and DVMD systems, the CCTV industry is stepping boldly to the forefront of electronic security design. Little or nothing that we have produced in the last 50 years compares or competes with the new face of CCTV. So let's talk DVMD.

Where We've Been
First, I always like to discus the history of a process. This helps me to understand how we got where we are and, more important, where we are going. Video motion detection started with analog systems some 25 or 30 years ago. The technology was simple: Monitor the video signal coming from the camera and look for any disturbance within a specific section of it. This was the beginning of the four VMD false-alarm nightmares.

• Contrast change due to fluctuating light or sudden lighting changes. Flash a headlight across a dark room and you have contrast change. Streak it across the area that you are monitoring and you have motion. At least, this is how we had to perceive motion in the beginning. The problem is obvious: Not all disturbances are motion. Many are fluctuating light—sweeping headlights, room lights turning on or off, lightning, cigarette lighters, flashing neon signs, the sun passing behind a cloud. The opposite side of contrast change detection also caused a problem. If a black man dressed in black walked past a dark or black wall, or a white man dressed in white moved past a bright or white wall, did either man produce enough contrast change to be detected? It depended on the sensitivity setting, which was almost always set to minimum.

• Flying garbage. A piece of paper on the wind passes through your secured area and bingo, you have an alarm. A legitimate alarm. An object, in motion, passed through your protected zone. What about a rainstorm, snowstorm or dust storm? No, you don't have to be in the desert. I have seen dust storms in every state in the U.S. and in many other countries. High, playful winds across open plains open the day to dust storms. Same with snow and rain. The result of such natural interference would be an alarm. Such storms could create a problem serious enough to disable your entire system for a while.

• Animals. Birds and other critters don't let your motion detection problems get in the way of their daily business. Hello false alarms. I always get excited when I see a robin redbreast in my yard. It is a living symbol of spring. It is also a living nightmare for video motion detection systems. But, you say, it's so small. True, but the original systems didn't differentiate between small things and large things. On the other hand, what if I needed to detect an object moving that were only two or three pixels in size? Couldn't be done with analog. Just a bit too small and precise.

• Constant motion. Fountains, revolving doors, escalators and bodies of water all create constant motion. To the old systems, motion was motion—there was no discrimination. If it moved, it created a disruption to the video signal and therefore created an alarm.

The net result of the early days was that video motion detection, for the most part, was kept indoors, locked away in chambers or areas where there was little or no chance for alterations.

The Birth of a Giant
Sometime in the very beginning of the '80s, digital video motion detection was born. The first digital system I was aware of was called the DAVID by the Senstar Corporation of Kanata, ON, Canada. For the most part, DAVID was developed for government applications. Think about this: DVMD was available before computers were available to the general public. Expensive? Yes! Accurate to the extent that it could be used outside effectively? Not by today's standards, but given the options of 25 years ago, yes. Multi-camera input, multiple zones per input, with limited motion discrimination and individual contrast sensitivity per zone pushed DVMD up to the plate and hit a home run. The only major drawbacks were cost, limited market acceptance, technical and theoretical acceptance, and false alarms.

The next 24 years would prove to be quiet in the world of digital video motion detection. Things were happening, but no one was bragging about them. Then, suddenly, so many improvements and so many new applications happened at the same time that we had to look. I have been promoting DVMD for more than 15 years. However, it seems that just lately, people are starting to ask how this technology applies to them. I have a simple answer. This technology applies to every application that you can dream of, from visual alarm monitoring, to access control, to traffic control, to wide-area protection, to temporary and flexible on-site demand coverage.

Advancements Galore
Now, let's apply modern digital technology to our four spooks of video motion detection mentioned earlier.

• Contrast change due to fluctuating light or sudden lighting changes. The beauty of digital technology is that it gives us the ability to monitor each pixel of every image individually and/or as a group. Because of this, we now have the ability to monitor and filter harassing lighting or contrast changes. If the sun goes behind a cloud, the system stays calm and determines that the area of surveillance is still visible and real and that the contrast change is scene-wide and so must be a shadow. If lightning strikes or a headlight or flashlight sweeps across the scene, the same logic is applied in reverse. It comes down to the quality of the system you are working with. In essence, a snapshot image is taken at different intervals and is used as a comparative study for all future shots. By constantly updating and comparing, the system detects only actual, physical change—not light. On the other hand, we are able to detect even the subtlest of changes since we can check pixel by pixel. This solves our black or white man dilemma as discussed before.

• Flying garbage. The first advantage of digital is multi-area coverage per scene. This reduces the area(s) that can be affected by outside stimuli. Although a small factor, it helps. And because we monitor the scene on a pixel-by-pixel basis, we can filter the images by object size, object direction, object color, object density, object motion and pattern, and object speed.
1) Object size. If I am looking for a man-sized target, then I have two size discriminations: larger and smaller. I do not want to see something the size of a car or truck. I do not want to see something the size of a squirrel or floating trash (unless it's the sports section of a really big newspaper). So, I have just eliminated or filtered out a good percentage of my potential false alarms.
2) Object direction. If I can tell my system to monitor, during specific time periods, only those objects that are moving east to west or up or down in the scene, then I again am lowering the affect of floating garbage or outside interferences. The garbage must now be of specific size, in a specific area of my scene, and moving in a specific direction at a specific time.
3) Object color. Although not a feature that you will find in the average, off-the-shelf DVMD, this would be a fairly easy and possibly effective algorithm to add to a good system. For instance, you sell John Deere® tractors, and you want to monitor visual activity in and out of your storage lot. You're not as interested in people as you are in equipment driving off. So, you set the filters for objects the size of your tractors, moving within a 50-foot area by the front gate, going away from the plant, and painted John Deere Green®. False alarms are becoming obsolete.
4) Object density. This is where your dust storm or fog comes into play. As long as I can see through the fog or dust at the percentage that I decide is acceptable, there will be no alarm. This is a good thought, but what about a person, dog, truck, fly or whatever moving between the camera and the object of protection? Let's go into a museum and protect a statue. A person walks in front of my statue and effectively blocks the majority of it from view of the camera, so an alarm is sounded.
About 16 years ago, I had the pleasure of doing some work in Australia. While there, I was approached by a friend that had a friend that was a modern dance director. Seems this dance director wanted to give his modern dancers the ability to create music through their motions on a stage. So he took two cameras, aimed them at the stage from opposing directions, and melded the two video signals into a computer program of his own design.
Using the same tactics of basic video motion detection, he was able to create multi-sized, invisible, three-dimensional cubes of air that were “hot.” He tied his multiple outputs to a sound synthesizer. The net result was that his modern dancers could float around the stage hitting different points and create their own music as they danced. The net result for the world of security was the birth of 3-D DVMD. Give me two or more opposing views of the same area, and I will give you one of the most advanced video motion detection features available today. Three-dimensional alarms detected and confirmed by multiple sources, simultaneously. This puts the bag on false alarms from one more perspective.
5) Object motion and pattern. We are now seeing systems that not only determine if an object is moving left or right, but also calculate its progress or pattern of movement. A floating piece of paper, regardless of size, does not move like a man walking. The cool part of this technology is that it is giving us the ability to tell the difference between a walking dog and a crawling man. Doesn't sound like much, but this is huge. Additionally, systems are being experimented with that read or predict an individual's intent based upon his actions when he enters a store or bank. Is he there to rob the place or do a general transaction? Does he demonstrate violent tendencies or is he of a passive nature? These systems are on the brink of introduction and will ultimately advance and change the direction of security's preventive nature. So cool!
6) Object speed. Somewhat self-explanatory. I want to detect motion but only if it is moving in a specific direction slower or faster than a predetermined speed. This could be used for various applications from parking lots to monitoring forklift drivers in sensitive, highly populated areas. We could use it for ticketing speeders on our highway systems, but the debates—oh, the debates.

• Animals. The above six filters also address the problem of small, flying, crawling, creeping animals and insects.

• Constant motion. This is one of the biggest, hardest, and yet most significant improvements in video motion detection since its inception. Almost 30 years of area protection have been slighted by fountains, rainstorms, snow, sleet and hail. How can I detect motion of an individual if the camera is auto-panning or it is snowing? We've discussed density as a form of filter, but snow and rain are different. They are not stagnant like fog. They are in motion, sometimes in several directions at the same time. A boat on the lake needs to be protected, but the waves make it move—heck, the wind makes it move. And what about all those ripples? Wouldn't all their sparkly indifference create false alarms? Absolutely. Hence the introduction of electronic semi-intelligence: the ability of a DVMD to learn a pattern or type or style of constant motion.
Now when the rain comes, the system continues to protect. It still sees the object of concern and filters out the mess between it and the camera. The boat is moving, but the system calculates its expected range and looks for the parameters of the defined threat. A fountain or escalator is in the middle of the image, but the system looks through the water or past the moving stairs and looks for change in the expected pattern of movement. It analyzes the change, and if it is beyond the acceptable parameters as defined by your filters, it creates an alarm. Man, oh man how I love the kinky thrills of modern technology. How I wish I could be here in 50 or 500 more years to look it over again.

What's It Cost?
OK, so what's the cost of this modern technology? It's not cheap. However, when you consider that we have the ability to replace or remove a huge array of various types of electronic alarm mechanisms—leaky coaxial, fence alarms, microwave motion detection, outdoor photo beams, infrared motion detection systems—the cost of the systems drop quickly.

If you considered video motion detection in the past and passed it up, start looking again. I think you will be amazed and pleased with what is available to you now.

Charlie Pierce is the President of LeapFrog Training & Consulting, a company dedicated to training the professionals of the CCTV industry. Visit its Web site at