However, a conventional particle filter is not suitable for object pose tracking because the search space is considerably large and a long computational time is required to find the solution. Consequently, back projection-based sampling is introduced, which has a notable advantage in that the search space is reduced. Back projection-based sampling reduces the search space using a depth map; if the depth information is known, one point on the image corresponds to one point in 3D space. The back projection method is shown in Fig. 9. One point on the depth map corresponds to one point on the surface patches in 3D space. Therefore, 3D space volumes are diminished to the 3D surface patches, resulting in the reduction of the search space.