Positurschätzverfahren and robot

An image recognition method according to an embodiment of the invention obtains a camera image generated by capturing a subject using a camera (10) (a three-dimensional sensor). A plurality of coordinates is obtained, corresponding to a plurality of pixels that are included in a predetermined region in the camera image. Subject distance information is acquired that indicates a distance from the subject to the camera (10) to the plurality of image elements. Then, the posture of the subject contained in the subject surface in the predetermined range is estimated based on the plurality of coordinates and the plurality of subject distance information items that were purchased.