Dangerous action identification method and system applied to station passengers

The invention discloses a dangerous action identification method and system applied to station passengers, and the method comprises the steps: collecting the dangerous action video data of a station platform according to the dangerous action characteristics of the station passengers, and building a dangerous action data set; the collected video data are converted into picture frames, and an AlphaPose model is adopted to extract human body joint skeleton feature values; a zebra optimization algorithm is improved, that is, a lens imaging reverse learning strategy is introduced into the zebra optimization algorithm, population diversity is improved, and the improved zebra optimization algorithm is adopted to optimize an AlphaPose model; the extracted human body joint skeleton feature data are sent to a constructed TCN-transformer network model for training, and the trained model is used for identifying dangerous actions of passengers in the station; and carrying out front-end interface display on the obtained da