Rapid migration method of industrial safety intelligent system for personnel detection

The invention discloses a personnel detection-oriented industrial safety intelligent system rapid migration method, which comprises the following steps of: mixing a public data set and a historical database to obtain a natural data set; training the natural data set to obtain a basic model; acquiring field data of a new scene; deploying natural data in the natural data set to the field data to generate artificial data; mixing the field data, the artificial data and the natural data in proportion to obtain application data; and training application data based on the basic model, obtaining a scene model, and completing migration deployment. According to the method, the character target elements are extracted by utilizing the segmentation network and are added into the generation of the artificial data, so that effective data are efficiently generated, and the model generalization ability is improved; automation of intelligent system migration is realized, and the migration efficiency of the whole process is gre