Den här sidan är utskriven från Högskolan i Halmstads webbplats (www.hh.se). Texten uppdaterades senast den 2015-06-12. Besök webbplatsen om du vill vara säker på att läsa den senaste versionen.

2015-06-12

Next generation fingerprint analysis

Fingerprints found at crime scenes, left in a speck of blood or dust, are often incomplete. Criminals might go unpunished and there is a risk of wrongful convictions. New research from Halmstad University aims to develop a more reliant method for fingerprint identification. Prints that cannot be used today might be used tomorrow – advanced image processing is the next generation tool for forensic fingerprint analysis.

– Today, identification of fingerprints is mainly based on the visual skills of forensic examiners. They can work for hours or days, trying to extract fine details from images manually, which they later insert in a matching algorithm. The human factor is crucial – if the examiner misses something, the result might be devastating. Our goal is to provide an efficient tool for image analysis to help the forensic examiners in their work, says Anna Mikaelyan, a researcher in mathematical methods of image processing who recently defended her PhD thesis in image analysis at Halmstad University.

Recycling garbage

The average human finger has 75 to 150 ridge characteristics creating a unique pattern that differs even in identical twins, unlike DNA. Fingerprints, which are impressions of the pattern, have been used for identification and evidence in criminal investigations since the 19th century. Nevertheless, fingerprints found at crime scenes are often only partial and of poor quality. In order to find a matching pair of fingerprints in a forensic database, the images need to have around 15 fine details in common.

– Only fine details from the fingerprints, extracted by examiners, are used, the rest is thrown away. There is information valuable for identity establishment in the discarded material, and our research has made it possible to use this material. Fewer details, for example five, can be as powerful as 15 if all available information is utilised properly, says Anna Mikaelyan.

Halmstad researchers have therefore developed a tool for forensic examiners, an image descriptor called SAFE (Symmetry Assessment by Finite Expansion). SAFE can use pieces of a fingerprint that is thrown away today –recycling useful image material. Anna Mikaelyan explains:

– The image descriptor SAFE gives an additional identity to a fine feature by using the surrounding images that, until now, have been regarded as unusable. We hope that this image analysis tool, combined with the examiners’ manual work, will increase the chance of correct identification.

Iris identification

The research has so far been focused and tested on forensic fingerprint images. However, there are other application areas within biometrics for SAFE, for example regions of the face and of the iris.

– The image descriptor is easily generalisable and can therefore be used for any image processing application, but this requires additional experiments, says Anna Mikaelyan.

Text: LOUISE WANDEL

Anna Mikaelyan recently defended her PhD thesis in image analysis at School of Information Technology. Photo: ROLAND THÖRNER

Fingerprint analysis is difficult and time consuming. The research results will hopefully lead to a more efficient finger print analysis that can increase the chance of correct identification.

About Anna Mikaelyan

Anna Mikaelyan received her MSc degree in Applied Mathematics from Southern Federal University in Russia in 2006, and her MSc degree in Financial Mathematics from Halmstad University in 2008. Since 2010 she has been a doctoral student in the image analysis group at Halmstad University. The main field of research is mathematical methods of image processing in the fields of biometrics and forensics. Anna Mikaelyan successfully defended her PhD thesis on April 17 april, 2015: Compact orientation and frequency estimation with applications in biometrics: Biometrics on the orientation express.external link

Updated 2015-06-12