A Framework for people Tracking and Stituation Evaluation in Multi-Camera Outdoor Environments

Kurzfassung

People tracking has been a major topic of research in computer vision for a long time. This is why we choose to combine several established approaches and algorithms in order to achieve robust tracking of objects in general and both detection and tracking of people in particular. We aim to establish a framework, that handles different situations (both indoor and outdoor) and can combine camera input from many cameras, but also works well with just a single camera. Our framework allows computation of information about the environment, in order to support tracking algorithms and allow the assessment of situations. Within the development of the framework, we compare different approaches (such as SURF-features and BRIEF-features) on both performance and how they contribute to the correct analysis of the images. This allows us to achieve good results while retaining near real-time computation on normal consumer hardware.