The invention discloses a human body height estimation method based on characteristic inflexions, and mainly solves the problems of high complexity of traditional binocular vision and resulting failure to meet the real-time requirement. The method includes the steps of firstly, performing motion detection for a video frame to obtain a motion pedestrian foreground; then, performing characteristic inflexion detection on a foreground region to obtain head and shoulder characteristic inflexions; next, performing segmentation to obtain the pedestrian head according to the inflexions and calculating the current head to body ratio; comparing the head to body ratio of the current frame with a head to body ratio model value; and determining whether occlusion occurs, if not, updating the head to body ratio model value, and if so, completing the human body height estimation with the head to body ratio model value and the head height. The method of the invention completes the human body height estimation through human body head to body ratio characteristics and obtains positions of pedestrian feet, has the advantages of low cost and low algorithm time complexity, and can be used for intelligent monitoring and visual positioning systems and so forth.