Capabilities that were at the forefront of image capture for security a decade ago, such as movement detection and target tracking, are now the bread and butter of the industry. Although such simple systems are adequate for many applications, heightened concerns about security in public places is driving the demand for intelligent systems that are able not only to distinguish between different individuals in a crowd, but characterise the behaviour of each. Such systems can quickly spot individuals acting suspiciously, and alert the appropriate authority to the potential risk.
In the latter half of the 20th Century, airplane hijackings and bombings became a prime method for extremist groups to draw attention to their causes. This, as well as drugs and other illicit goods being smuggled, means that every air passenger passes through a series of security checks, such as metal detectors, random baggage searches and pat-downs. The sight of specialist drug and bomb sniffing dogs is not uncommon at most major hubs. However, all these can be bypassed if an unauthorised person can enter the grounds of the airport and gain access to planes on the ground. For this reason many airports have sophisticated systems that monitor perimeter fences to ensure that security services are alerted should anyone attempt to enter the airport grounds without proper clearance.
Thermal imaging cameras from Flir Systems are being used for just such applications at airports in Houston, Texas. Flir’s ThermoVision Sentry II and ThermoVision Integration Series cameras have been installed at the George Bush Intercontinental Airport and William P. Hobby Airport as part of a comprehensive video security system that protects more than 30 miles of perimeter at the two airports. The infrared cameras are integrated with the airports’ video processing and radar surveillance systems, linked to a geospatial map, to provide video coverage in areas where ground-based radar is obstructed by buildings, or other objects.
Although the drive to improve airport and aircraft security has led to many significant developments in the application of intelligent imaging systems, CCTV surveillance still relies predominantly on the pan and tilt camera. However, a new system devised under the OpTag project, a three-year research project funded by the 6th European Framework Programme, is capable of generating full colour, high-resolution panoramic images in real time, by bringing together the very latest image processing hardware and software.
The camera system itself comprises eight CMOS sensors in an octagonal arrangement. The field of view of each camera module just overlaps with each of its two neighbours. This gives full 360° coverage, which is optimised for a viewing distance of 10m. In addition, by offsetting the optical axis of the lens relative to the sensor, the camera can be made to look downwards in a ‘rising front’ configuration. This arrangement has the advantage that vertical parallel lines remain parallel in the image, reducing the amount of subsequent geometric correction processing required. The system has a total frame size of 16 megapixels over the whole panorama (2 megapixels per sensor), and a data refresh rate of 20fps.
Within the OpTag system, images from the cameras are geometrically remapped to take out lens distortion before being stitched together to form a seamless panorama. The processed data is then made available over a network to multiple users, either in full panorama, or smaller areas of interest if bandwidth size restrictions are a concern. The full high-resolution panorama can be made available over the network at a slower frame rate for archiving or other purposes. The OpTag system can be combined with object or feature recognition software to help security personnel locate left-luggage or suspect packages. The panoramic views offered by the system also make it ideal for locating and tracking individual passengers, helping to locate lost or missing passengers – or monitoring suspected wrong-doers.
With increasing concerns about identity theft in the media at large, the concept of biometric testing – identifying people using inherent characteristics or traits – is now talked about as a serious method of identity confirmation. There are many different traits that can be measured, such as a person’s fingerprint, DNA, facial structure or even voice. To be a useful method of identity confirmation, any biometric test must be reliable, quick and easy to perform. The characteristic must also be unchanging with time. None of the above methods meet all these criteria. However, security firm IrisGuard has a method that can quickly and accurately identify one person out of a database of many hundreds of thousands, by recording the pattern of the iris of the eye. The iris is the elastic, pigmented, connective tissue that controls the pupil. Its unique random texture results in a unique pattern — the IrisCode. It is the only internal human organ visible from the outside and is protected by the cornea; this protection prevents the pattern of the iris wearing off over time, which maintains its stability over a person’s lifetime.
The IrisGuard IG-H100 system uses a sensor from Sony Image Sensing to capture an image of the pattern of someone’s iris. The system then compares the image to all those stored on its database. The largest deployment of this system so far is for border control in the United Arab Emirates. Upon arrival, every foreign visitor with a new visa submits to an iris-check at all the land, sea and airports in the UAE. This project involves more than 35 border control centres and 22 deportation centres, with a total of 130 iris cameras, distributed throughout the UAE.
The system checks the passenger’s iris against the central iris database of deported persons. If a match is found, the person is further investigated; otherwise, the person proceeds to normal passport control. This live iris-check takes no more than two to three seconds. So far the system has stopped more than 73,000 people returning to the UAE after being deported, despite them trying to re-enter the country using different passports or identity documents.
Seldes, a Belgian frame-grabber company, and Micromedia, a French software specialist, have jointly developed a new alarm alert system. The system analyses images to identify the nature of an event and automatically inform the correct staff to deal with the contingency. Lynx, from Seldes, analyses the images – to check to see whether the event should trigger an alarm and, if so, to assess the nature of the event. Contingencies can range from intruders, through fires, to regulating the speeds at which vehicles move inside large plants. Seldes uses an FPGA system allowing demanding image-processing to be done on the board and hence reducing the computational demand on the industrial PC.
Once the nature of the event has been ascertained, Micromedia International’s software comes into play. The company has more than a decade’s experience in producing ‘call management’ software for scheduling interventions after alarm and event alerts. According to Mike Walters, manager of the company’s Belgian office, such systems have traditionally been oriented to reporting readings from sensors. But now, he says, there is another market arising because of the decision-making capabilities that are possible with video and imaging processing.
Micromedia’s software is designed to get the right information to the right person at the right time – there should no longer be a need for security guards, when they notice some event on a screen, to look up directories to find out who is the responsible member of staff to contact. The Alert system automatically sends emails, text messages, or can even contact mobile phone numbers in order to ensure that the right person is informed. Jpegs of the event can be attached to the email, so that the member of staff does not have to search through many images before taking action.
The frame-grabber takes the decision on whether or not to trigger an alarm and the Alert software decides whom to call. An advantage of using images rather than traditional sensor data is that frames taken before the alarm was raised can be preserved for later inspection and analysis – allowing security managers to understand just what happened before, rather than after the event.
Imaging technology has had to evolve at great pace in order to keep pace with heightened security concerns, and hardware and software developers alike are responding to that demand with ever more innovative products.
The popular SeaWorld resort in San Diego is positioned right on the sea front. The park even has its own beach. But this can cause particular headaches for the security personnel in charge of controlling access to the park. The beach is only around the headland from the local harbour. This makes it particularly easy for ticket dodgers to avoid having to pay the entrance fee for the park – when they took a close look at the beach security guards found that some people would simply hire a boat from the nearby harbour and land on the beach within the park, thereby avoiding the official entrance.
The problem facing the management of the park was this: how is it possible to prevent people landing on the beach without a permanent manned security presence? A traditional motion-sensing camera would not be able to distinguish between people seeking illegal access to the park and law-abiding visitors enjoying the atmosphere of the beach. The solution came in the form of an intelligent vision system, based on Perceptrak software from Smart CCTV through Cernium. This system monitors the video feed from standard CCTV cameras for unusual behaviour, in this case people attempting to land on the beach. If such behaviour is detected the system can automatically alert the appropriate member of staff to the potential problem, meaning that a permanent security presence is not required.
The Perceptrak software can detect many different kinds of suspicious behaviour in many different applications, ranging from vehicles stopping suddenly in suspicious locations, to seeing a person running in a crowded place. The system can be configured to effectively disregard the background of its viewing field, ignoring falling leaves, moving shadows or people and vehicles behaving normally. Perceptrak can also detect suspect packages and left luggage in busy airports.
So far Perceptrak is being used in a wide variety of applications. As well as SeaWorld, it is being used to protect historic sites such as the Mount Vernon Estate – George Washington’s family home. The system is being used to ensure the safety of students at Wayne State University, and to prevent people gaining unauthorised access to Boeing’s airplane manufacturing facilities.
