Determination of the degree of individuals acting as a union in collective motions is an interesting research topic, while defining and quantifying such a structure and its motion is a challenging task. This paper proposes a method for such an effort by providing a descriptor of collectiveness, its constituent individuals, and its computation, with the proposal of a collective merging algorithm for the detection of collective motions from random ones. Various datasets, for example self-driven particles, pedestrian crows, and bacteria colonies, are used in evaluation of the method’s accuracy and robustness; in addition, some are compared with those of some human subjects’ perceptions to demonstrate consistency.

The main contributions of this paper are: (1) the presentation of a method for crowd collectiveness measurement, which includes descriptors of collectiveness, its constituent individuals, efficient computation, and a collective merging algorithm; (2) a theoretical analysis of the method; and (3) potential applications/research on the crowd system as a new universal collectiveness descriptor.