With the rapid development of the sports industry and the orderly advancement of smart city construction, outdoor sports venues and indoor stadiums, as densely populated public activity spaces, have increasingly demanding requirements for safety management and operational efficiency. However, the data value contained in current sports venue surveillance data has not been effectively exploited, as it is mostly used as a basis for tracing the accident process after an incident occurs. In fact, through in-depth learning of surveillance data, we can effectively prevent the occurrence of sports accidents. Based on this, this paper designs a crowd motion prediction and anomaly monitoring system based on sports venue surveillance data. The core idea of the system is to preprocess crowd motion trajectories, implement trajectory clustering using the LDA topic model, perform trajectory prediction integrated with convolutional long short-term memory networks, and conduct abnormal trajectory monitoring, so as to realize the classification, prediction and monitoring of crowd behavior patterns in sports venues. The system can effectively realize the early prediction of sports accidents and remind the activists on the sports venues through audio equipment.
Hu et al. (Thu,) studied this question.