Locating people in surveillance video using soft biometric traits

Abstract

Semantic descriptions are a commonly used and very natural way for people to describe one another. Descriptions comprising details of clothing types and colours, skin and hair colour, gender and build are very effective ways to communicating an approximate appearance; however such descriptions are not easily utilised within intelligent video surveillance systems, as they are difficult to transform into a representation that can be utilised by computer vision algorithms. In this chapter, we will investigate two recent approaches to using these semantic, soft biometric descriptions to automatically locate people in surveillance imagery. We present the strengths and weaknesses of each, and discuss their suitability for real-world deployment and how they may be further improved.