Trajectory Similarity Measures Using Minimal Paths

Dealing with surveillance systems, large amount of distance measures are presented in order to classify both normal and abnormal behavior. Typically, techniques based in point-to-point distances are used. However, these techniques do not take into account information about the environment, like pits or restricted areas, for instance. Using a minimal path algorithm to model the usual paths, we develop new trajectory distance measures that are able to introduce information about the scene. The results obtained show promising results.