We propose a computer vision method to identify motion pattern changes in human crowds that can be related to an unusual event. The proposed approach can identify global changes, by evaluating 2D motion histograms in time, and also local effects, by identifying clusters that present similar spatial locations and velocity vectors. The method is tested both on publicly available data sets involving crowded scenarios and on synthetic data produced by a crowd simulation algorithm, which allows the creation of controlled environments with known motion patterns that are particularly suitable for multicamera scenarios.
If you use the dataset, please cite the following paper:de ALMEIDA, Igor R. ; CASSOL, Vinicius J. ; BADLER, Norman I. ; MUSSE, SORAIA RAUPP ; JUNG, CLAUDIO ROSITO . Detection of Global and Local Motion Changes in Human Crowds. IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, v. 27, p. 603-612, 2017.
Five different scenarios simulated by CrowdSim in order to provide support information to validate the proposed model.Download Ground Truth and Ground Plane Homography
Two groups, of 60 agents each, start to move independently, and after some time, they gather and move as a single group. After some time, this larger group is dissolved and the agents run to seven exit points in the environment.
Download scene 1.
One group consisting of 60 agents is split into two independent groups (A and B with 30 agents each) during the motion. Each group moves in order to achieve different goals. After some time, both groups increase their velocity and the agents of group A run toward six different directions, while group B remains formed and moves in unison toward the same goal.
Download scene 2.
Two groups A and B, of 10 and 100 agents, respectively, move in opposite directions. After 20 s, agents in group A speed up from 1.0 to 1.4 m/s, keeping their orientation. Agents from group B maintain a constant velocity.
Download scene 3.
Two groups, A and B, of 30 agents each, move in the same direction. The speed of group B (1.0 m/s) is greater than that of group A (0.8 m/s). Agents of group A reach the same goal, while agents of group B, when near to the target point, disperse and move in five different directions.
Download scene 4.
Two groups, A and B, of 30 agents each, move in the same direction, but the speed of group B (1.0 m/s) is greater than that of group A (0.8 m/s). When group A achieves their goal, agents leave the scene. However, when agents from group B achieve their goal, they turn back and move toward the start point.
Download scene 5.