Research On Real-time Monitoring Framework For Crowd Flow Based On YOLOv5 And Spark Streaming

In order to address the issue of real-time monitoring of crowd flow in everyday public and commercial areas, this paper proposes a real-time monitoring framework for crowd flow (RTMFFCF) using YOLOv5+Spark Streaming, and explores the characteristics of crowd flow. Firstly, the framework utilizes the Market-1501 dataset to train a video-based person detection and tracking model using YOLOv5+deep sort algorithm. Then, the recognition results are fed into Spark Streaming through Kafka for real-time analysis and processing. Finally, the analysis results are presented in real-time visualization. Experimental results demonstrate that the proposed framework effectively achieves real-time monitoring of crowd flow.