A step ahead of crime

Detectives will soon be able to spot criminal suspects simply by examining their feet using a system that automatically identifies shoes from the prints left at crime scenes.

'An officer at a crime scene could photograph a shoeprint and know within minutes which make and model of footwear made that print,' said Dr Ahmed Bouridane, reader in computer science in the school of electronics, electrical engineering and computer science at the Queen's University, Belfast.

He is leading research into how image processing can boost the use of shoeprints in crime fighting. They are more common than fingerprints at crime scenes but currently are not always helpful as evidence. One study shows that of 14,000 marks collected from crime scenes in a year, only 2,000 were compared against known patterns and, of those, only 500 were positively matched.

One of the main reasons is that feet move, so the prints are rarely perfect. The print can be blurred or distorted by foot movement on a soft surface, and only part of the sole might touch the ground if the wearer is running on a hard surface. And these imperfections are compounded if other footprints cross the same tracks or if there are objects such as stones within the marks.

So although shoeprints are one of the hardest types of evidence for a criminal to remove from a crime scene, their complexity and subtlety has meant that the task of recording them is usually given to forensic scientists. Bouridane is working on a system that will mean the job can be done by a police officer. It will also speed up the analysis and identification.

Currently, 3D prints in soft surfaces are photographed and then cast using dental stone. Two dimensional prints on hard surfaces or textiles such as carpets are photographed or 'lifted' using electrostatic sheets and powder, a technique akin to photocopying. Only then can the marks be compared with the 20,000 that are stored in databases.

'This checking is done manually,' said Bouridane, 'It's basically tedious and not reliable. Officers have to match by eye the print of interest to one of thousands that are in the databases.'

There are three kinds of database used for comparison. One is a book showing the design of every shoe on the market, which is kept up to date by manufacturers volunteering the information. The second is a database of shoeprints lifted from other crime scenes and which helps to relate one crime to another. The third consists of a database of shoeprint profiles from suspects. Matching the print of interest to one in the database is both a skill and an art.

'To increase the reliability of decision-making in such forensic settings and assist the duties of the forensic specialist, there must be an effective and precise framework for searching, browsing, querying and interacting with these important collections,' said Bouridane, 'And such processes are expected to be carried out in a timely manner.'

So he is applying his expertise in computer vision to develop robust, reliable and repeatable image processing algorithms. 'We are using image transformations that grab the features of a shoeprint at different resolutions, scales and orientations,' he said. 'There's no need for any pre-processing of the image.' The transformations automatically 'clean up' the print, 'correcting' for blurred or scuffed marks and spotting when there are inclusions such as stones.

'Then we take some of the discriminating characteristics of the print's pattern, such as circles, rectangles or zigzags. These can be compared automatically with those stored in the databases to come up with an answer.'

In preliminary trials the image processing and database matching system has proved to be 95 per cent reliable. What's more, it completed each match in less than 15 minutes compared to the hours it can take a person to do the same job.

Bouridane's team has been working on the project for almost a year in association with the government-owned Forensic Science Service and expects to deliver a workable version this summer. One of the next steps is to ensure that scene-of-crime officers have the equipment and skills to be able to take digital images of shoeprints that are of sufficient quality to be useful to the new system.

'An officer will be able to take a picture with a digital camera, transfer it to his or her PDA, upload it wirelessly to the central computer where the transformation will take place followed by database matching,' said Bouridane. 'While still at the crime scene, the officer will receive a picture of the suspect's shoe.'