A rough set based reasoning approach for criminal identification

Abstract

As a supplement to mugshot detection, a new approach is proposed to capture the eyewitness’s visual perception in the form of symbolic representation. It reveals physiological and facial characteristics of criminal which help in their identification. A rough set theory based technique is introduced to model those symbolic representations. This approach provides an intuitive insight to process criminal’s imprecise and imperfect knowledge. We used a benchmark mug-shot dataset consisting of 300 criminals faces from the Chinese University of Hong Kong (CUHK) to study the correctness of our proposed model. We took the help of 105 students of Indian Institute of Information Technology, Allahabad, who were treated as eyewitness to depict the visual perception about 300 criminal faces of CUHK. The experimental verification is composed of two modes which are analogous to viewed sketches and forensic sketches. Like viewed sketches we have generated test case-I, where perception is given while looking at the photo whereas test case-II is like the forensic sketches where the description is given by recalling the memory. We have achieved encouraging results on the viewed sketch database as well as forensic sketch database.