Crowd Counting Method Based on Cross-Source Multi-View and Multi-Granularity Video Fusion

With the rapid growth of the world's population and the rapid development of urbanization, the issue of crowd gathering safety has aroused widespread concern in society. Extensive video surveillance systems provide rich data support for dense crowd management. Video-based crowd counting and density estimation methods are the core technologies to ensure the safety of crowd gathering. Different from single-view video analysis, cross-source multi-view multi-granularity video contains more cross-information. The complementary sharing of information is of great help to solve the problems such as occlusion in the current crowd counting. Therefore, this article proposes a crowd counting method based on cross-source multi-view and multi-granularity video distributed information fusion. By establishing a distributed structure from different cameras that matches low-altitude and high-altitude views, it uses fine-grained from low-altitude images. The high-resolution local information corrects and supplements the global information from high-altitude images, so as to calculate a more accurate and global number and density of people. This method is actually applied to the landmark building of Suzhou Life City Square. The changes in the number of people and the movement situation during the evacuation are analyzed and evaluated, and good results are obtained.