Event Discovery in Multimedia Reconnaissance Data Using Spatio-Temporal Clustering

Contents

In reconnaissance application scenarios, support for the analysis of important events that happened during a mission is highly desirable. This demands techniques to discover those events from media and sensor data that have been captured during missions. Because reconnaissance missions constitute uncontrolled environments, high-level event detection based purely on media and sensor data analysis methods is difficult. In this paper, we propose the use of spatio-temporal clustering for the discovery of important mission events. We cluster basic events that occurred during a mission — such as the creation of content or basic events detected via media or sensor data analysis — according to time and location of their occurrence. Experiments performed on real-world patrol data show the efficacy of our approach. They indicate the general usefulness of spatio-temporal clustering for event detection in scenarios where media and sensor data analysis methods are not reliable, or practically absent.