Grant: Engage Plus Grants Program

Detecting and informing users about events of interest is the key function in intelligent video surveillancesystems. Among the most important requirements is that the real-time surveillance system must: 1) operatecontinuously and autonomously in real-world conditions, and 2) deliver real-time awareness of remoteincidents at the appropriate time. These are the two specific problems that we are addressing in the proposedproject, where we aim to develop an efficient video surveillance system that intelligently detects and reportsevents of interest to the user. This is very challenging since the visual characteristics of an unconstrained sceneare unstable, mainly due to large variations in illumination. Currently, most of the existing systems use RGBcameras and operate in constantly and evenly illuminated areas to detect simple events (e.g. intrusion), usingbasic change detection techniques. The novelty of our approach is twofold. The first consists in using the 3Dinformation that can be obtained regardless of the illumination conditions, using recent cost effective RGB-Dsensors. As a consequence, more robust motion models can be constructed by adding 3D position informationto standard color features. The second consists in applying object detection techniques and exploitinginformation about the objects the user is interacting with in addition to the motion recognition algorithm. Theproject objectives include the development of: 1) models for actions and events in the 4D domain (3D spaceand time) and 2) methods for recognizing actions and objects, to be integrated in a real-time solution for eventdetection.