Relating Real Crowds with Virtual Crowds

This chapter describes some reflections concerning the challenge of capturing information from real crowds to relate it with virtual crowds. Three parts are discussed here: (i) a study undertaken on the motion and behavior of real crowds, where the goal i

  • PDF / 1,360,042 Bytes
  • 25 Pages / 439.37 x 666.142 pts Page_size
  • 59 Downloads / 218 Views

DOWNLOAD

REPORT


Relating Real Crowds with Virtual Crowds

6.1 Introduction This chapter describes some reflections concerning the challenge of capturing information from real crowds to relate it with virtual crowds. Three parts are discussed here: (i) a study undertaken on the motion and behavior of real crowds, where the goal is to identify some patterns of the behaviors of real people to be used subsequently in virtual crowds, (ii) discussion of a few sociological crowd aspects, and (iii) computer vision methods as automatic ways to capture information from real life to guide virtual crowds.

6.2 Studying the Motion of Real Groups of People In this section, we present a simple and empirical way to observe real crowds, and to use such observation for crowd simulation. We do not employ computer vision algorithms (as it will be presented later) to capture semantic information from videos, but only a visual process, trying to select information from the crowd structure. This process is important for defining which and how the observed information can be used to simulate realistic behaviors in virtual crowds. The observed information in the real crowd relies on two aspects: crowd characteristics and crowd events, as discussed in the next subsections.

6.2.1 Crowd Characteristics Characteristics of the crowd can include: crowd space (if all the space is occupied, if the individuals are close to each other, regions where there is some action, regions where people walk are examples of crowd space questions), crowd size (number of groups and individuals within each group), crowd density (relation between space D. Thalmann, S.R. Musse, Crowd Simulation, DOI 10.1007/978-1-4471-4450-2_6, © Springer-Verlag London 2013

169

170

6

Relating Real Crowds with Virtual Crowds

Fig. 6.1 Group acting on the train station. The drawing in the figure shows the action location

Fig. 6.2 Real crowd scene representing directions of movement (arrows) and interest locations (circle) for crowds

Fig. 6.3 Grouping formation observed in a filmed sequence

and crowd sizes), crowd structure (considering the grouping of individuals), and crowd basic behaviors (walk, grasp, look at some location, apply a posture). Concerning the crowd space, some data are identified: • Space to apply actions (location and size of space). For example, Fig. 6.1 shows a space to apply action identified in a video sequence. • Space to walk (location and size of space), as shown in Fig. 6.2. • The way groups walk (speed, respecting the group formation, occupying all the walking space), as shown in Fig. 6.3.

6.2 Studying the Motion of Real Groups of People

171

Fig. 6.4 Complex scene of crowds where different entities (crowd, groups, and individuals) can be recognized

In order to describe the crowd size, observations are made to identify the grouping existent in the filmed sequences as well as the number of members in each group. For instance, in the filmed sequence of Fig. 6.3 (only shown one picture), approximately 40 persons were observed during some time, whereas the groups a