Institute of Cartography and Geoinformatics Research Big Data and Machine Learning
Scene analysis - pattern recognition in person tracks

Scene analysis - pattern recognition in person tracks

Team:  Fischer, Sester
Year:  2017

Content

The project deals with the automatic recognition of conspicuous movement patterns from given trajectories as (x, y, t) sequences of objects. For this purpose, patterns are defined that indicate conspicuous behavior for which automatic extraction methods are to be researched. This includes, among others, the qualitative description of the recognition rates of patterns.

Here, patterns can be defined in different contexts: a distinction is made between individual patterns and group patterns resulting from multiple individual types of behavior in a common space-time context. It is important that patterns typically do exist undependently, but are usually influenced by external factors (for example stationary or mobile physical obstacles when crossing a room or different behavior patterns depending on the time of day).

To recognize safety-critical, person-induced behavior, scene-typical behavioral patterns are learned in a spatio-temporal context. Based on this, deviating behavior can be detected on-line in order to be able to provide immediate indications of a possible safety hazard.