Overview of crowd group classification methods

Crowd grouping is the process of dividing people into different groups according to the classification rules, which is the basis for subsequent research on crowd behavior. This article provides a review of various methods for crowd grouping, discusses in depth the advantages and disadvantages of each method, and points out some shortcomings of current methods, providing useful guidance for future research and promoting further development and innovation in the field of group partitioning. Firstly, based on the way individual motion information is used in group classification methods, this article elaborates on the current methods from three perspectives: using position and direction proximity methods, using position, velocity and direction proximity methods, and using trajectory similarity methods. Then, this article compares and analyzes the effectiveness of different methods on various datasets. Finally, a summary of the various methods described in this article is provided, and some challenging questions and prospects for future research are proposed to address the shortcomings of current methods.