Research on the Method of Analyzing Crowd Behavior Based on Anchors

In the issue of group behavior recognition, a method based on anchor points for analyzing group behaviors was proposed. By using methods such as interaction degree and manifold ranking, the relationships of feature points on the topological structure were calculated. According to the topological relationship with the anchor points, each feature point was assigned a group category, resulting in subgroups. Finally, a continuity merging strategy was proposed to merge subgroups that exhibited continuity in space and movement direction, to obtain the final group. Experiments show that this method fully exploits the topological relationships of feature points at the structural level and can accurately identify group behaviors in complex structured groups.