Big Brother has got bags of potential
Utilising biometrics and facial recognition to harness the power of CCTV - Orly Golding, senior forensic analyst for the Facial Verification Bureau (FVB), explains the link between biometrics and facial analysis and the ever present CCTV
When identifying people from their unique physical characteristics or behavioural traits using automated systems, biometrics has become increasingly common in everyday life. Established examples include fingerprint, iris recognition and face recognition, however the applications of biometric technologies are far reaching, ranging from identifying criminals to controlling access into a building.
Transforming forensic science
Biometrics can be split into two categories. The first, verification, involves a one to one comparison, while identification sees the comparison of one set of data against an entire database.
Biometrics has transformed some areas of forensic science. The introduction of automated fingerprint identification systems, holding millions of prints, is one such example. When a new fingerprint is entered, it runs that print against others on the system, picking out the closest matches. Human analysts then compare the results to verify accuracy, a practice widely accepted when using automated systems.
If the entered print is from a crime scene this would fall into the identification category. Alternatively, the database can be used to verify when the identity of a suspect is questioned.
Facial biometrics in the area of law enforcement will eventually follow a similar process.
Here's looking at you!
An area of policing that would benefit from automated facial recognition systems is in the use of CCTV footage. Due to the ubiquitous nature of CCTV in the UK, many crimes, or at least the actions of individuals before or after a crime, are captured on CCTV. Searching through hours of footage can, however, be a time consuming and painstaking process for investigating officers. When the perpetrators are located, the next task is identifying them.
As well as searching CCTV footage for a particular person, such as a repeat shoplifter, automated systems compare CCTV images with images on a database comprising police mug shot images and other crime scene images to identify a potential match. In this way, crimes committed by the same perpetrator which are already on the system can be linked.
While automated systems are an excellent solution to the laborious and fallible task of scanning surveillance footage for identification purposes, CCTV quality is usually very poor. With little control over lighting conditions, viewpoints or distance of the subject from the camera, identification can be problematic. Biometric analysis of CCTV footage will therefore never be as accurate as that used in controlled situations, for example when used as an access check, and the output will always require further verification by human analysts. The forensic verification of a suspect's identity has for many years been carried out by a number of individuals and small companies, such as the FVB.
The independent company specialises in analysing forensic imagery and confirming or disproving the various hypotheses proposed by the police or defence. One of the services provided by the FVB is facial image comparisons, more commonly known as 'facial mapping', which involves comparing images of a crime perpetrator, usually from CCTV footage, with images of a suspect.
Even if at a first glance the faces look similar, one observable significant difference which cannot be attributed to time lapse, lighting conditions, image quality etc, means that the suspect can be eliminated. If no significant differences are identified, a level of support is given towards the contention that the crime perpetrator and suspect are one and the same person. The level of support will depend on the number of observable facial features and their distinctiveness.
When applied properly, biometrics such as DNA or fingerprints, are virtually indisputable evidential tools. Facial image comparisons complement automated biometric systems, they can accommodate for varying lighting conditions, viewpoints etc. In the first instance, the biometric search may result in several matches from a system. The FVB would then come in to verify the facial information by comparing the facial features and either eliminate or provide evidential support that the two people in the images are one and the same. In addition, the facial evidence may also be corroborated with additional data such as the height of the crime perpetrator or clothing continuity.
By analysing this data, the FVB can provide the police, or the defence, with expert evidence regarding the identity of the person in the footage. This may strengthen a case or eliminate innocent persons as suspects.
The move to mobile biometric devices
The rate of adoption of mobile technology by police services is steadily increasing, opening the way for biometric handheld recorders. Considering that many frontline officers already use head cameras, the potential for identifying suspects against facial databases could well be revolutionised.
When out on the beat, these real-time, browser-based devices would enable verification and identification of a suspect's identity without the need to return to police headquarters. In a short amount of time, the suspect's facial image could be compared against the database and the officer will know whether to let the suspect go, or to make an arrest. Handheld biometrics devices have the potential to reduce the time needed to make accurate comparisons for identifying and verifying, provided the experienced human analysts are there as a second confirmation. Mobile biometric deployment could make a major contribution in helping reduce police bureaucracy, improve service levels and meet operational and financial objectives.
Fulfilling the potential of CCTV
The combination of automated facial recognition systems and forensic analysis by experts in the field of imagery comparisons would help to utilise the millions of CCTV cameras throughout the UK and fulfil their potential in establishing the facts of cases and assisting in convictions. As this potential is realised it may also increase the effectiveness of CCTV cameras as a deterrent.
When identifying people from their unique physical characteristics or behavioural traits using automated systems, biometrics has become increasingly common in everyday life. Established examples include fingerprint, iris recognition and face recognition, however the applications of biometric technologies are far reaching, ranging from identifying criminals to controlling access into a building.
Orly Golding, senior forensic analyst for the Facial Verification Bureau (FVB), which provides an independent imagery analysis service from CCTV images for the purpose of forensic identification
