add a normalization method with "random spatial distribution" in pdf #327
Description
In the current method, the probability of measuring distance for pedestrians who do not interact is computed using time randomisation of the original trajectory file (i.e. shuffling the frames in the trajectory database).
This technique enables the method to account for areas that would not be used by pedestrians, which are not represented as obstacles in the walkable area. In other words, the method assumes that, while all pedestrians are free to walk everywhere, they will use preferred trajectories.
This assumption is true in most scenarios, but not all. In a dense and static or semi-static crowd, the positions of all pedestrians will not vary over time. However, if the crowd were not present, each pedestrian could walk freely within the considered space.
In this situation, time randomisation would result in all pedestrians occupying the same position. Therefore, a 'random spatial distribution' is required to represent the distances between pedestrians who do not interact in this case.