A Framework for Trajectory Segmentation by Stable Criteria and Brownian Bridge Movement Model Master’s Thesis

Criterion-based segmentation is the problem of subdividing a trajectory into a small number of parts such that each part satisfies a global criterion. We present an algorithmic framework for criterion-based segmentation of trajectories that can efficiently process a large class of criteria. Our framework can handle criteria that are stable, in the sense that these do not change their validity along the trajectory very often. Our framework takes O(n log n) time for preprocessing and computation, where n is the number of data points. It improves upon the two previous algorithmic frameworks on criterion-based segmentation, which could only handle decreasing monotone criteria, or had a quadratic running time, respectively. Furthermore, we propose a new segmentation method based on the dynamic Brownian Bridge Movement Model. This segmentation method has only one parameter: the segment penalty factor, which can be chosen automatically or interactively using a so-called stability diagram. We finally show how to combine the dBBMM-based method with criteria.