Crowd Counting for Ancient Architectural Complexes
Haiyan Wang, Lida Huang, Rui Pan, Zhanhui Sun, Tao Chen, Xuehong Gao · 2024
Ancient architectural complexes possess immense historical, cultural, societal, and artistic value, making them highly sought-after tourist destinations that attract a large number of visitors on a daily basis. However, the intricate spatial layout and substantial pedestrian traffic pose considerable difficulties for crowd counting in these settings. While existing research on crowd counting primarily concentrates on public areas like scenic spots, commercial streets, and gathering places, there is a notable scarcity of studies specifically addressing crowd counting in ancient architectural complexes. This study proposes a crowd counting method for such complexes. To address the difficult problem of crowd counting in ancient building complexes as scenic spots, the improved YOLOv5 crowd counting algorithm is used to calculate the crowd density. The algorithm was proposed and trained using video surveillance data collected from ancient architectural complexes. It was compared to the official lightweight version of YOLOv5s, resulting in a 90% increase in computational efficiency and improved crowd detection accuracy. This indicates that the algorithm proposed in this paper has faster computational efficiency and higher accuracy for crowd counting in ancient building scenes.