Heightened airport security in the wake of 9/11 and other subsequent security alerts – one of the most recent being the failed attempt to detonate a bomb on a flight from Amsterdam to Detroit on Christmas Day 2009 – is something facing most travellers. Airports will now restrict any liquids carried onto the plane to less than 100ml. New full-body scanners are being trialled at Manchester airport in the UK, which could add another layer of security to pass through. The scanners use passive millimetre wave technology to identify potential security risks concealed on the human body, although issues over the effectiveness of the technology as well as privacy concerns mean there is still fierce discussion surrounding their use.
Tightening of security might be necessary, but it can lead to longer queues and greater disruption for travellers. A Dutch airport has installed a smart surveillance system to monitor passengers passing through arrival security filters in an attempt to speed up the process. The system, developed by Abstract Computing International and using smart cameras from Sony Europe’s Imaging Sensing Solutions (ISS) division, uses face detection to track individuals as they pass through the airport and determine how long they remain inside the security filter areas. ‘There is pressure on airport staff to process passengers quickly, especially those catching connecting flights,’ explains Allard Blom, chief operating officer of Abstract Computing International. The system was designed to provide airport personnel with real-time information on the volume of passengers moving into the arrival hall allowing faster processing times.
The area that had to be covered was large (approximately 25 x 50m) with 12 lanes of people to monitor. Queues could be up to 50m long, extending outside the main hall. Sony’s smart cameras were installed above each of the lanes looking horizontally into the faces of the queuing passengers. The cameras conduct face detection to count people passing through an area and also to monitor how long individuals remain in a queue.
Abstract Computing’s ScanaFace software was used to detect and track faces in the images (the complete solution is called ScanaLane). The system provides real-time information on the position of passengers in the queue, which is then fed to a central database to give updated information every five minutes to operational staff on the airport floor. ‘If there is an influx of people from an incoming flight, airport staff can act on the data provided by the system,’ says Blom. It also allows immigration and customs officials to communicate more effectively and to staff their respective areas according to the volume of people arriving.
‘Standard video analytic techniques cannot be used for counting people in this instance,’ notes Blom. ‘The cameras are looking horizontally and objects such as baggage would cause a lot of false-positives and disturbance in the data. ScanaFace not only counts each face, but it tracks faces within the image, meaning the analysis provides information on the number of individuals as well as how much time they spend in the queue.’ The information is transposed onto a 3D model of the area to establish where people are walking and how long it takes to move from one point to another.
A colour camera was needed to discern faces and the Sony smart cameras were used as they provide megapixel resolution and are stable in terms of colour and quality of image. In addition, the camera could export raw images, which is a stable way of transferring data and ensuring the colours remain true – again, important for video analytics in surveillance applications. The cameras also have high sensitivity, providing colour images in low-light intensities.
ScanaLane uses feature extraction to characterise a face. ‘It is taught what is and what isn’t a face from thousands of examples,’ Blom says. The system installed in the Dutch airport tracks faces throughout the scene covered by each camera, but Abstract Computing is working on a system for handing over the tracking from camera to camera, technology made possible due to the reproducibility of the algorithms.
Spanish company InfoGlobal has been involved in a large-scale project to install a surveillance network in Mexico City’s subway system. The scope of the project involved setting up a broadband IP-Ethernet network backbone covering all the stations on metro lines 1 to 11, installing up to 3,500 analogue video surveillance cameras throughout the subway system, including onboard trains, and using InfoGlobal’s IG-Monitor video encoder-recorders deployed in technical rooms to capture the video streams. The system also uses Euresys Picolo H.264 frame grabbers to digitise the incoming video feeds.
According to Jose Ignacio Guzman Crespo, associate director at InfoGlobal, the surveillance system has had a critical impact in detection and prevention of criminal acts on the metro system. Crespo also pinpoints installing cameras on the trains themselves as being important to improving surveillance, and InfoGlobal has developed its IGMON6-SE video encoder-recorder for just such an application, with the system being deployed in new trains on the Madrid metro.
Marc Damhaut, VP product management at Euresys, also sees increased demand for mobile security solutions – for use in buses or trains, for instance – which presents its own set of problems. Systems have to be rugged and reliable, as well as compact: ‘There are design challenges in terms of managing heat generation, environment constraints and shock and vibration – but also in terms of transmission of the images, which need to be transferred over a wireless network.’
Euresys has released the Picolo U4 H.264 PCI-104, which follows the PCI-104 standard (a form factor suitable for mobile and ruggedised applications) specifically designed for mobile security applications.
In mobile applications, there might be different types of wireless connection between the vehicle and the control centre with different bandwidths. 3G communication, for instance, which is a common wireless communication technology, has quite a low bandwidth of 300kb/s. The challenge for video capture cards is to compress the image faithfully so that a high-quality image can be transferred at a low bandwidth. ‘Even when transferring images over wireless bandwidth, it is still important to record the images at a much higher resolution,’ explains Damhaut.
Euresys’ cards can compress the images at two different programmable rates. From a single camera, the system can capture a high-quality, high frame-rate, low-compression signal while also capturing a highly compressed, lower-resolution signal that can be transmitted wirelessly. ‘For PC-based applications there typically aren’t limitations on bandwidth, but with mobile applications this is a real challenge,’ Damhaut says.
Euresys has supplied frame grabbers for video surveillance on buses, whereby the cameras record high-resolution images locally during the day and on returning to the depot in the evening, the images are downloaded wirelessly.
With higher levels of security and more cameras covering wider areas there is a need for intelligence to filter the images and generate meaningful security data. Smart cameras excel at adding intelligence to a surveillance application, as complex algorithms can be ported onto the cameras for onboard processing.
Myriam Beraneck, product manager at Sony ISS, adds: ‘Adding intelligence increases the level of automation of certain tasks. Incident detection, for instance, becomes more reliable, with alarms flagged up by the system alerting operators to potential security risks. This also reduces the bandwidth placed on a network.’ Adding intelligence in the form of image processing to look for security threats in the masses of video data available, is crucial to providing an effective surveillance solution and potentially averting a terrorist attack.
Thermal imaging is becoming increasingly popular for surveillance of large sites, such as banks or power stations, due to the ease with which it can detect people. ‘The problem with using visible imaging for security applications is that there is too much information in the image for automatic detection algorithms,’ comments Jean-Luc Tissot, technical director at Ulis, a manufacturer of infrared detectors.
Thermal imaging only detects a temperature gradient and so any hot objects, like people, will show up very clearly, irrespective of day/night or environmental conditions. This means the image data can be extracted very easily in software, providing an effective intruder warning system. ‘It is easier to detect an intruder using thermal imaging, but once detected a high-definition visible camera can be used to acquire a more detailed image of the person,’ Tissot says.
Ulis’ infrared sensor technology is based around amorphous silicon, which operates in the thermal imaging range (8-14μm). The advantage of using amorphous silicon for thermal imaging is that the material is fully compatible with silicon foundry lines and so thermal sensors can be produced in high volumes, lowering the overall price of a system.
Humans show up very clearly using a thermal camera, and the technology can be an effective addition to a surveillance system. Image courtesy of Ulis.
Piet van Riel, sales manager at Belgian company Xenics, comments that, very often, security applications will require a combination of wavelength bands. A two-sensor system, such as Xenics’ Meerkat PTZ, means end users can add visible imaging capability to long wave infrared (LWIR) to provide more effective day/night imaging. Other systems might combine imaging with radar to sweep the area for motion and then focus the cameras on the area to identify the movement.
Van Riel explains that heavy fog or rain reduce imaging capabilities in the LWIR region, but SWIR will be less affected. High temperatures can be viewed most effectively with mid-wave infrared (MWIR). In certain locations, such as in South Africa, which has the warm Pacific on one side and the cold Atlantic on the other, two different wavelengths will be needed to carry out the same application on different coasts. ‘The background temperature, atmospheric temperature, aerosols, humidity, distance to the object dust and pollution, can all affect the quality of thermal reading and the wavelength used,’ he says. A further complication that van Riel notes is that the air over the sea is different to the air over land. Xenics’ cameras are equipped with algorithms to enable thermal imaging to account for these aberrations involved in coastal imaging.