A depth image camera is used for efficient estimation of walking intention of a pedestrian. Three-Dimensional image coordinates of the pedestrian’s joints are obtained from the image data that includes depth information and are converted into the absolute coordinate values. The absolute coordinate data are classified and matched with all 20 joints of a pedestrian and the 9 joints that are corresponding the lower limbs are finally selected. After calculating each three-dimensional area of a triangle that was formed with the adjacent 3 joints of the 9 lower limb joints, the centroid of all triangles along time is obtained. The walking intention, that includes the direction and the speed of walking, can be estimated with the change rate of this centroid. It is experimentally verified by comparing the distance that is measured with inertia moment unit and the distance that the calculated centroid is moving.