Population mobility chat based on multi-sensor data fusion

The human mobility of urban citizens is pivotal in urban planning and management and especially plays a vital role in security and emergency control, which has drawn substantial attention. While most of the study about human mobility found are based on one single computer vision technique using data from surveillance cameras. Since the effect of outside environment like light variation, single computer vision technique is revealed with lack of reliability and accuracy. With the popularity of mobile devices like smartphones and WIFI techniques, the vast scale of real-time data containing urban citizens' location can be achieved. Thus, to overcome the limitation of using single computer vision technique, we propose a system combining data from various sources including surveillance cameras and WIFI detectors in supervised areas to gain real-time population density and citizens' inflow and outflow. Once the population density is higher than the threshold, the system can automatically set the alarm. To implement the system, observation points are placed in Beijing Institute of Technology, Zhuhai campus and a large amount of distributed data is collected, processed and analyzed using big data technologies (like flume, spark and Hadoop) and data fusion algorithm to obtain real-time population density. Human mobility is displayed in Baidu map after data process, which enables users to monitor and predict emergency situations happened in the university campus. This system also provides a solution for spatial and temporal emergency control in densely-populated areas in the city like train stations and airports without a large investment of human resources, which can serve as a crucial tool in urban management.