Construction safety early warning method and device based on saliency neural network model

The invention discloses a construction safety early warning method and device based on a significance neural network model, and the method comprises the steps: carrying out the image preprocessing of a construction equipment and personnel image set, generating training sample data, constructing an initial significance neural network model, and training to generate a target significance neural network model. And when a to-be-identified construction site image is received, inputting the to-be-identified construction site image into the target saliency neural network model, generating a target refined feature map, determining a plurality of corresponding target body regions, identifying the target body regions, and identifying the to-be-identified construction site image on the basis of a judgment result whether a preset condition is met or not. Judging whether a construction worker associated with the target ontology data has a construction potential safety hazard or not; the technical problems that the existin