Facial Recall and Computer Composites

Abstract

Imagine, if you will, that you are sitting quietly outside a café sipping your favourite hot beverage when someone rushes past and snatches your mobile phone, which you left on the table, as you often do. You were able to get a good look at the person’s face, albeit for a short time. Your next hour is spent speaking with a police officer giving a description of what happened and what the offender looked like.
It is likely that you could describe accurately what happened. You will probably also be able to describe the perpetrator’s build and clothing. There should be no trouble in saying what was the sex of the person and his or her ethnicity; you should be reasonably accurate at estimating the age, height and weight. You could probably remember some details of the person’s face.
Incidents such as these are known as “volume” crime. They occur frequently, often without physical assault to the victim, and their seriousness, at least from a legal perspective, is fairly minor. Due to limitations in police resources, many perpetrators of volume crime are never caught, although time spent locating particularly prolific offenders can be worthwhile.
Crimes involving these repeat offenders, and other crimes of a more serious nature including murder, arson and rape, are generally given higher priority in police investigations. It is in these cases that eyewitnesses (witnesses or victims) may be asked to engage in a range of tasks to assist in the detection and later conviction of the offender. When the police have a suspect, they may be asked to take part in an identification parade. (Further details about this are the focus of a separate chapter.) Alternatively, eyewitnesses may be shown photographs of previously arrested criminals for identification, sometimes referred to as mugshots. In the absence of a suspect, CCTV footage or other evidence, witnesses may be called upon to externalise an offender’s face. The aim is to create a visual image based on remembered information, so that it can be shown to other people for identification. Such images are known as facial composites and are seen in the newspapers and on TV, for example, BBC CrimeWatch. The idea is that someone who is familiar with the face will name it to the police and, in doing so, will provide new lines of enquiry.

The focus of the current chapter is on the construction and the recognition of facial composites produced by modern software systems. A separate chapter in this volume details how composites are created by sketch artists. This chapter describes and evaluates typical software programs that the police use to construct faces. It will be demonstrated that the traditional approach used with eyewitnesses is generally ineffective for producing identifiable images, and that alternatives are required if composites are to be effective in the battle against crime. Several successful developments are described. The final section looks to the future and asks what might be on the horizon for producing even more effective faces.