Research and Development

The technology we are going to develop is the inter-camera correspondence resolver that enhances human tracking accuracy based on intelligent fusion of several cues including, but not limited to, the face recognition score for the human faces captured from multiple cameras, the estimated probabilities of a human entering the coverage areas of particular cameras based on the space-time constraints, as well as the general appearance model of a human object. Another technology that we are going to develop is the self calibration and site modeling that enables a tracking system to learn and refine incrementally about the actual camera topology and path probabilities based the online tracking results.

Other technologies that we shall be using in this project includes our formerly developed face reconstruction and recognition, that could provide robust face recognition score to establish a more reliable cue for human tracking; and single camera human tracking that works with the face reconstruction and recognition technology to tackle the human tracking problems in an environment where severe occlusions are unavoidable.

With the innovative use of these developed and existing technologies, it is expected that the accuracy and usability of human tracking systems in a multiple cameras system can be substantially improved.