Research on Dynamic Decision-making Hybrid Model of Pedestrian Flow Based on Deep Learning

Abstract

With the progress and development of China’s transportation system, pedestrian travel behavior and safety have received increasing attention. At the same time, controlling the population density and population movement in densely populated areas plays an important role in ensuring national security for the public security departments. Therefore, how to properly adjust the switching time of traffic signals has become an urgent problem. In view of the above requirements, this paper proposes a signal switching model based on deep learning for dynamic regulation of pedestrian traffic. The hybrid model is divided into three parts, named the real-time data acquisition part, the historical data analysis and prediction part and the decision model part. Firstly, the LSTM model is used for the analysis and prediction of historical traffic data with time series characteristics. Then, the real-time data acquisition adopts the lightweight and high-performance target detection model MobileNet-SSD proposed by Google. Finally, the signal switching decision model is proposed to analyze and determine the data provided by the above model, and two adjustment factors are defined to adjust the proportion of historical data and real-time data to the impact of decision results.CCS Concepts•Computing methods ➝Artifical intelligence ➝Computer vision ➝computer vision tasks.