This article first appeared in the January 2009 issue of it magazine. This is part of the Folk Technology series that I write every month.The terror attacks in Mumbai have highlighted the need for better surveillance in our cities. Technology could play a major role here.
The terrorist strikes in Mumbai have exposed the vulnerability of our cities. It is essential that areas like hotels, train stations, bus stations, airports, markets and malls be provided with better surveillance to enable authorities to monitor and catch suspicious behaviour.
The Chattrapati Shivaji Terminus (CST) at Mumbai has over three million passengers passing through it every day. Most of these people, almost all of the time, perform normal activities like buying tickets, boarding and getting off trains, buying magazines, talking on mobile phones, and so on. Statistically, an abnormal event like the terrorist attack happens once in many millions of activities. But that does not lessen its impact or the need to spot it and nip it in the bud because, although rare, such an event can cause immeasurable damage to life and property.
Is it possible to add more ‘eyes’ to the city? Is it possible to use closed circuit television (CCTV) and other imaging technologies to catch terrorists before they strike? Is it possible to use machine vision technology to automatically detect abnormal activities in the images that have been captured and alert the police? Believe it or not, the answer to all these questions is ‘yes’, provided the right technology is in place.
An imaging and storage challenge
Each public place has its peculiarities. In some, it is easy to place CCTV equipment, especially indoors, where the cameras can be mounted on the walls. In others, like open markets and streets, it may be necessary to erect structures for placing the cameras.
Of course, this might need a large number of images. For instance, a place like CST has 18 platforms and many common areas. To cover each platform will require at least 30 cameras, and the common areas will require about 250 cameras. So, in all, CST will require about 800 cameras to cover it fully.
In open spaces like markets, tethered helium balloons, floating in the air, fitted with cameras can be used. One such balloon at a height of a few hundred metres fitted with multiple high-resolution cameras can cover the entire market very easily.
It is important to make sure that cameras can capture good quality images in different conditions like daylight, night light, cloudy or rainy situations, etc. The need is for imaging technology that has good resolutions under different imaging conditions, is inexpensive and is robust enough to be used outdoors and indoors.
Then there is the matter of storage. One hour of video from one camera can easily take up 30 GB (gigabyte) of storage space, so one would need to use image compression techniques to keep storage requirements at a minimum. However, the compression should not result in too much loss of clarity in the images, otherwise the very point of storing the images would be lost. What’s more, these image compression techniques need to be evaluated with CCTV images under different real life lighting and imaging conditions. Similarly, while a number of imaging technologies exist, it is still a challenge to determine the right solutions for camera placement and lighting.
The brain behind the eyes
Monitoring of CCTV images has traditionally been done by human observers, who often monitor multiple screens simultaneously. As the number of cameras increases, it will be necessary to increase the number of human observers as well. However, studies have shown that after the first 40 minutes, the alertness of the human observer falls drastically. Hence, automated solutions to recognise events that seem out of the ordinary, need to be developed.
Machine vision technology is maturing and today, a computer program can automatically detect stationary and non-stationary objects in an image. For example, people, luggage, trolleys, dustbins, kiosks and other objects in images can be automatically identified by a computer program. A computer program can even detect unattended luggage with reasonable accuracy—an important factor considering that bombs are often left in the form of luggage, transistors, etc. A computer can also detect when someone carries a suspicious-looking object. A human observer can be alerted to look at such images, verify that the system is correct, and notify security personnel on the ground.
It is possible to match the faces and gaits of people with those of known suspects, as well as identify suspicious behaviour like nervous or jerky movements by individuals, which can be indicative of a person who is about to carry out an attack.
It is essential that security experts work with the engineers writing such programs to define the scenarios of interest. The computer program needs to be given examples of normal events so that it can report anything that does not fit into this pattern.
Start-ups for security
But while security in public places needs to be stepped up, such security should not be disruptive. At CST, the authorities rejected a plan to frisk every person entering the premises—it was impractical to frisk over three million passengers every day. Security has to be effective but at the same time it should be unobtrusive as far as possible. Just placing guards at entrances or even frisking is not a solution in itself. We need to put technology to effective use and collect as much actionable data as possible to preempt terrorist strikes. ASSOCHAM expects that the security business in India will become a Rs 50,000 crore industry in the next four years. Much of this money will go into deploying technologies that can effectively fill the security gaps.
This represents an opportunity for technopreneurs to help make our lives safer. The first challenge is to reduce the cost of hardware. Today, a night vision camera costs Rs 10,000, while a camera for normal lighting costs Rs 5,000. These costs need to be brought down through innovative use of photo equipment and semiconductors. The next challenge is to develop better machine vision software. As it is not possible to monitor all images manually, machine vision algorithms for automatic surveillance need to be developed and deployed to aid the human observers. As in all matters these days, terrorism needs to be tackled by a blend of human effort and technology.
(Part of the monthly Folk Technology Series that I write for i.t. magazine)
Read: Fighting Terror Series
