Automated Entity Classification in Video Using Soft Biometrics

With the growing concerns surrounding security and terrorism around the world, biometrics has become one of the premier solutions to combat these problems. Traditionally, biometrics has been an academic problem that has been studied from the perspective of optimal environments (good lighting, cooperative subjects, single-frontal-2D / 3D photographs, etc.) and unlimited time and processing power. In the real world, this is not the case. Today, even the most sophisticated identification algorithms can take up to 10 seconds to accurately identify a single subject due to the massive size of realistic databases (~1 million subjects). In this proposal, we present a variety of methods that will lower identification time by reducing the overall number of “possible subjects” through the use of Indexing and Soft Biometrics. Based on prior published experience in this area, we will evaluate the features that can be best used to categorize individuals into predefined bins (based on gender, skin color, height, weight, anatomical proportions, geometrical facial features, etc). Once identified, these features can be extracted from a video source and entered into a database for later use or can be specified a priori to locate subjects that match the given parameters.