The invention provides a method and device for identifying smoking based on a video image, a server and a storage medium, and the method comprises the steps: obtaining a to-be-detected monitoring video, extracting a key image frame, and carrying out the human body posture feature extraction of the key image frame; identifying the to-be-detected monitoring video by utilizing the smoke identification model to judge whether smoke exists or not; utilizing a posture recognition model to recognize the extracted human body posture so as to judge whether a specific human body posture matched with theextracted human body posture exists in a posture library or not; if the smoke recognition model outputs that smoke exists and the posture recognition model outputs a recognition result that a specifichuman body posture exists, determining that a smoking event exists. According to the method and device, smoking events are monitored and recognized in time, the method and device can be widely applied to no-smoking areas such as schools, shopping malls, factories and urban public transportation places, rapid, real-time and online intelligent recognition and monitoring of illegal smoking behaviorsare achieved, public health maintenance is facilitated, public safety is guaranteed, and civilized cities are established.