Analysis of Gait and Face Biometric Traits from CCTV Streams for Forensics

Abstract

Forensic detectives sometimes face situations where there may be a lack of evidence for a crime. Gait analysis-based identification from CCTV cameras can be used in such situations to help pave the path forward due to the ease of obtaining data and its non-invasive nature. There have been attempts at creating face recognition engines that work from afar; however, the effectiveness of such systems are hampered by the lack of detail due to the distance as well as the image quality from surveillance cameras. Moreover, in situations where the face in the video stream has been obscured by an object such as a mask or sun glasses, the effectiveness of facial recognition systems is greatly hampered. The gait capture technique of key-point generation can work with lower quality source footage as well as facial structure obstruction by considering the entire body for generating a signature. This paper describes the use of this method of capture in a forensics application. The application was then tested on local CCTV footage and showed reasonable success versus face recognition, easily identifying key-points before the face recognition engine could pick out a face.