The invention discloses a hoisting process danger identification method and system based on deep learning, and the method comprises the steps: S1, training to generate a Faster R-CNN worker and lifting hook detection network; S2, detecting a worker who correctly wears the safety helmet, a worker who does not correctly wear the safety helmet and a lifting hook in the current frame image in the video shot by the lifting site by utilizing the detection network; S3, judging whether a worker is detected in the current frame image or not, and if so, executing the step S4; if not, extracting the nextframe of image of the shot video, and continuing to execute the step S2; s4, judging whether a person is allowed to enter the hoisting site or not, and if not, triggering an alarm; if so, executing the step S5; s5, judging whether a worker who does not correctly wear the safety helmet is detected in the current frame image or not, and if so, triggering an alarm; if not, executing the step S6; s6,generating paths of the workers and the lifting hooks based on the detected positions of the workers and the lifting hooks; and S7, judging whether the distance between the worker and the lifting hook is less than a preset distance at a certain moment in the worker and lifting hook prediction path, and if so, triggering an alarm. The danger in the hoisting process is comprehensively and timely recognized, and hoisting accidents can be effectively prevented.