How to ensure the safe evacuation of personnel in port accidents has become a key research topic in port planning and operation. The key to solving the problem of crowd evacuation is to understand the movement patterns of the crowd. The key to solve the problem of crowd evacuation is to understand the movement law of the crowd. Crowd movement is a complex system, involving many factors such as crowd interaction behavior, urban spatial form, and architectural environ-ment. In order to provide accurate and reliable research basis for the study of crowd evacuation, this paper proposes a crowd motion simulation model based on deep convolutional neural network. In order to obtain the data required for neural net-work training, this paper uses CSRNet neural network and DBSCAN algorithm to extract real crowd trajectory data from surveillance video. Through the training of deep convolutional neural network, the real crowd behavior pattern is deeply studied, and the crowd motion simulation model is established by using the trained deep convolutional neural network. The experimental results show that the model can accurately predict the movement behavior of the crowd and truly simulate the movement trajectory of the crowd, which can provide a basis for the formulation of emergency evacuation strategies and the design of evacuation channels in public places.