A pedestrian statistical method and apparatus based on the recognition of a parietal region of a human body, being applied to the technical field of image recognition. The method comprises: reading a pedestrian depth image photographed by means of a depth camera and comparing the pedestrian depth image with a pre-acquired environmental mean image to obtain a foreground image (S101); segmenting the foreground image into a plurality of regions, detecting whether there is a step for edge pixel points of each of the regions, and detecting whether a region curved-surface formed by each region is coincident with a curved-surface of the parietal region of a human body (S102); determining that the currently detected region is the parietal region of a human body when there is a step for edge pixel points of the currently detected region and when the region curved-surface formed by the currently detected region is coincident with the curved-surface of the parietal region of a human body (S103); and counting and outputting the number of pedestrians according to the parietal region of a human body determined from the pedestrian depth image and the parietal region of a human body determined from an adjacent pedestrian depth image (S104). This method improves the recognition accuracy of the parietal region of a human body and realizes accurate statistics of pedestrians.