Section: New Results

Crowds

Participants : Julien Pettré [contact] , Richard Kulpa, Anne-Hélène Olivier, Samuel Lemercier, Yijiang Zhang, Jonathan Perrinet.

Perception of collision in crowds

We designed a level-of-detail (LOD) selection function to determine where collision avoidance constraints in crowd simulation can be relaxed without being perceived by spectators [4] . Collision avoidance is probably the most time consuming parts of crowd simulator. However, when only believable results are required, we argue that visually similar results can be obtained a low computational costs based on macroscopic crowd simulation. Based on a perception study, we determined the conditions for collision to be or not to be detected. We discovered that the camera tilt angle was playing a great effect on perception, whereas distance to camera (usually used in previous works) was only the third most important factor to be considered.

Mixed Reality Crowds

In the task of making virtual crowds and real people interact together, we explore a mixed reality solution [22] . The seamless integration of virtual characters into dynamic scenes captured by video is a challenging problem. In order to achieve consistent composite results, both the virtual and real characters must share the same geometrical space and their interactions must adhere to the physical coherence criteria. One essential question is how to detect the motion of real objects - such as real characters moving in the video - and how to steer virtual characters accordingly to avoid unrealistic collisions. We propose an online solution. First, by analysis of the input video, the motion states of the real objects are recovered into a common world 3D coordinate system. Meanwhile, a simplified accuracy measurement is defined to represent the confidence of the motion estimate. Then, under the constraints imposed by the real dynamic objects, the motion of virtual characters are accommodated by a uniform steering model. The final step is to merge virtual objects back the real video scene by taking into account visibility and occlusion constraints between real foreground objects and virtual ones.

Experiments on crowds

Evaluating crowd simulation models is a difficult task. In the frame of the ANR PEDIGREE project, we put in a lot of effort to perform experiments on groups of walking people in order to dispose of a reference database on groups motion. In order to obtain high-quality data, we measure people locomotion by using optoelectronic motion capture systems. In 2011, we starting obtaining detailed analysis on such motion after large efforts on motion analysis and processing. We had to develop dedicated reconstruction techniques because of the challenging conditions in which we performed our motion capture [12] . We submitted two papers on following modeling and simulation stages (submitted to Eurographics 2012 and Physical Review E).