The work aims to propose a fast boundary extraction algorithm applicable to stepped stack point cloud, so as to solve the problem of low efficiency of existing boundary extraction algorithm in stepped stack point cloud boundary detection. Firstly, Kd-Tree was used to construct the topological relationship of point cloud data. Then, the sampling point and its adjacent point were searched to construct the unit plane vector and perform vector superposition. The candidate boundary point was judged according to the vector modulus length after superposition. Finally, the maximum angle method of adjacent points was used to extract the boundary point accurately. Kinect camera was applied to obtain the stepped stack point cloud data to verify the algorithm. The experimental results showed that the proposed algorithm could shorten the running time by 75.14% compared with the classical boundary extraction algorithm when ensuring the extraction accuracy of stack boundary, and shorten the running time by 11.06% compared with the boundary extraction algorithm before optimization. The proposed boundary extraction algorithm can quickly and accurately extract the stepped stack point cloud boundary, which can provide reference for the design of automatic stack system.