Crowd grouping detection method based on split-merge strategy

Abstract

A crowd grouping detection method based on a split-merge strategy comprises steps as follows: individuals in a scene are divided into a static group and a dynamic group, and interaction between the individuals in the scene is simulated and quantized on the basis of a social distance module and a social force module; the crowd in the scene is represented with an undirected weight graph, each vertex represents one individual, the weight between adjacent vertexes represents the interaction degree of the two adjacent individuals, the problem about crowd grouping detection is transformed into a problem about undirected weight graph clustering, the undirected weight graph is clustered with a method based on the split-merge strategy, and the crowd grouping in the scene is searched. The crowd grouping detection has important application value in the video monitoring, and the method can have a very good effect in a more general scene, can also perform detection successfully on queuing which is difficult to detect, and has more universality than other methods.

Description

technical field [0001] The invention belongs to the field of image processing technology and pattern recognition, in particular to a crowd grouping detection method based on a split-merge strategy. Background technique [0002] Single-person behavior recognition is a frontier direction that has attracted much attention in the field of computer vision. It has been researched for many years. It has become one of the research hotspots in computer vision and pattern recognition. As the basis of group behavior recognition, the research aspect of crowd group detection has also received more and more attention from researchers. However, due to the complexity of the general scene, the unfixed number of people, and the variability of the interaction between people, the research field of detecting crowd grouping is a great challenge. [0003] Among the existing crowd detection algorithms, there are a series of algorithms based on F-formation. F-formation is a concept proposed by Ken...

AI Technical Summary

Problems solved by technology

However, this method is limited to the detection of several people "talking", which has relatively large limitations.

In addition, although some methods are no longer limited to a certain type of group, the effect is not very ideal for groups with a large number of people

There are also some methods that use the trajectory characteristics of people to perform group detection according to the direction and speed of people's movement, but this method is mainly suitable for people in the state of motion

Images