Automated Video Activity Analysis for Military Operations

The increasing use of video in force protection, autonomous vehicles reconnaissance, and surveillance in general has created a great demand for automated analysis and monitoring. It is infeasible to have humans monitoring and analyzing the vast amount of video used in such surveillance, and automated analysis yields hope to ease the burden and increase the amount of information extracted from collected video. Recent research on automated analysis has focused on detection of a set of simple activities in a given context. However further development is needed to generalize this analysis to be truly useful in effectively using collected video. For example, it is not generally possible to know ahead of time what an image analyst will be looking for. Additionally in many cases human attention is required for decision making, but this attention needs to be focused on a very limited set of collected video. We propose to develop a system for extracting useful information from large-scale video collections. To do this we will develop effective tools applicable to several scenarios: finding activities in a video collection matching an example activity, notification and highlighting of abnormal activity, and recognition of known activities.