Ubiquitous criminal identification system

Abstract

The main idea of my thesis is to build a system that helps to identify criminals on a crime scene. My thesis assumes that there are IP cameras installed in the street where the crime is to take place. IP camera information, such as its location is stored in a database. The database also consists of information related to individuals with past criminal records. The IP cameras are then accessed by the time of the crime and length of video required. The requested video is then returned to the system. The video is then converted into frames which are later used by feature detection and recognition procedures to identify the criminal. The features used for recognition are frontal face, profile face and ear, all of which are supported by the individual's height. Identification is done by means of comparing the people present on the crime scene to the ones in the criminals' database. Since there may be erroneous results, the disadvantages brought about by each feature recognizer are overwhelmed by the other classifiers. All recognition results are then merged into a particular individual. The process is then repeated for neighbourhood IP cameras. The recognized individual is therefore tracked from one camera to the next. The individual's ID together with the location of the IP cameras are present as output by the system, hence providing an identification of the criminal together with the route followed.