تکامل تعاملی کامپوزیت های صورت

Facial composite is a tool used in criminal investigation to identify an unknown person, usually a suspect of a crime. Traditionally, it is created by a forensic artist on the basis of a detailed description provided by one or several eye witnesses. Due to the subjectivity and low efficiency of the composition process, facial composite has not been considered to be a reliable evidence until recently. Despite expectations, replacing traditional drawing by computer-driven systems did not result in improved identification rates. Therefore, the research of the last decades has been oriented on developing a method more compatible with the human way of perceiving faces. In this paper, we present a system developed as a result of deep psychological research on human face perception and recall. The proposed approach enables automatic generation of facial images and their presenting to a witness, resembling the process of line-up identification. Besides providing information on computational model of a face, the primary goal of the paper is to discuss the applied evolutionary strategies in terms of individual parameter settings and evolutionary operator choice and to evaluate the strategies with regard to the construction time, operator usability and composite accuracy. Besides, we summarize the procedures recommended for further improving the efficiency of the composition process.