This talk will focus on identifying subjects in surveillance imagery. This is needed by society to solve crimes where imagery is at low resolution and faces and other traditional discriminatory information is not available. Soft Biometrics is a new field wherein the majority of new techniques augment traditional ‘hard’ biometrics data with soft measures. These measures are perceived to have low discriminatory ability, though can be used to increase the performance of a traditional biometric system, e.g. by augmenting data with estimates of age and gender. At Southampton we have been working on using these measures for identification alone. It required a totally new approach using human derived labels. These are then used for recognition. They can also be used to prime computer vision systems to estimate the labels a human would describe. It’s a totally new area encompassing computer science and psychology, bridging the semantic gap between computer and human vision. The labels ae used to describe the body, he face and the clothing for recognition. I will describe the background, motivation and accomplishments of these fascinating new approaches.
Mark S. Nixon is currently a Professor of Computer Vision with the University of Southampton, U.K. His research interests are in image processing and computer vision. His team were early workers in face recognition and later pioneers of gait recognition. He has chaired/program chaired many conferences (BMVC 98, AVBPA ’03, FG ’06, ICPR ’04, ICB ’09/15, and BTAS ’10). He is a member of IAPR TC4 Biometrics, the IEEE Biometrics Council and Fellow of IET, IAPR, and BMVA. His textbook on Feature Extraction and Image Processing for Computer Vision is currently in its Third Edition, published by Academic Press/ Elsevier in 2012. His first book, Introductory Digital Design - a programmable approach , was published by MacMillan, July 1995 and there's a new version Digital Electronics: a Primer, published by Imperial College Press 2015. With Tieniu Tan and Rama Chellappa, he wrote Human ID based on Gait which is part of the Springer Series on Biometrics, and was published late 2005. He recently co-edited the first book on Biometrics Spoofing Handbook of Biometric Anti-Spoofing, with Springer 2014 and its second Edition 2018; We wrote the survey on Gait Biometrics in the first text on Biometrics: Personal ID in Networked Society, and the Ear Biometrics Chapter in The Handbook of Biometrics. He’s written a heck of a lot of papers and supervised a heck of a lot of (great) PhD students.