Host: The Institute of Image Information and Television Engineers
Research on crowd movement started with addressing human perception issues and has been developing in a diversified way up to the present. Demands for models to automatically generate crowd movements have a wide spectrum. Taking a closer look at entertainment models, it has become a prime virtue to maintain the plausibility of crowd movements for a variety of scenes without loss of computation speed. In this article, we place more focus on the pleasing quality of human movements under the condition of real-time scene representation, and propose a robust and efficient model to trace movements of a crowd whose density distribution changes drastically, without camera blockings. We take into account the movements of individuals as well as small pedestrian groups consisting of a few people, to gain the plausibility of the model.