Due to tracking errors, outliers occur in some trajectories, defined as points that have too large a distance to the previous one and do not fit into the pattern of the trajectory. It can be useful to have an algorithm in the preprocessing section that detects and corrects those outliers. Another similar case are trajectories of a single person that drop in the middle of the trajectory and continue to track either the wrong person or nothing at all, but there are still some corresponding points to this.

An algorithm that goes through every distance between two following points and checks whether it is larger than expected would find outliers that are set up as described above. Those can be corrected by interpolating points between the last and the first correct point or extrapolating them at the beginning or at the end.

The trajectory needs to be split per person for these methods and to be put together afterwards.