Introduction

Figure 5.1: Overview of the components of the model. This chapter explains how the trajectory in the appearance parameter space $P$ is transformed into a sequence of pathlet models. The trajectory is split into pathlets and the pathlets are grouped to form pathlet models.

This chapter describes how we break trajectories into sequences of pathlets, which are then grouped and used to train a set of statistical pathlet models. This is step 3 of our framework (see figure 5.1). The first section of this chapter describes how we extract pathlets from the full trajectory. In the second section, we describe how we model the groups of pathlets. In the next section, we present our algorithm for grouping a small set of extracted pathlets, as well as an alternative for a larger number of extracted pathlets. Finally we describe how to deduce the sequence of pathlet models given the algorithms in the previous sections.