The invention discloses an escalator intelligent safety monitoring method based video analysis. The method comprises the followings steps of firstly obtaining a real-time video image sequence in a camera installed in a monitor area; then establishing a hybrid gaussian background model, and inhibiting shadows in a video image to extract a foreground; then carrying out human body object identification of the foreground via a classifier, and calculating the angular point light stream of a human body object, so as to determine whether abnormal behaviors such as retrograde motion or falling of passengers exist or not and to timely take measures of sudden stop of an escalator and the like, thereby ensuring the personal safety of passengers on the escalator. With the method provided by the invention, the precision of human body object detection is high, the detection speed is fast, by determining abnormal behaviors of the human body object with the angular point light stream, the probability of trampling events due to retrograde motion or falling of persons is effectively prevented, and the personal safety of passengers is protected to the greatest extent.