Point cloud registration is an important step in the processing of terrestrial three-dimensional laser scanning data. Aiming at the scene with small terrain fluctuation, we propose an automatic point cloud registration method based on projection distribution entropy. Initially, information entropy is used to describe the intensity of point cloud projection distribution. Following this, a coarse registration is achieved by seeking an optimal point cloud distribution between two point clouds. Consequently, the transformation parameters are determined between the two point clouds with different distributions and supplied as an input to the iterative closest point algorithm to achieve a fine registration. Compared with the automatic point cloud registration method based on features, the proposed method's main concern is the consistency of the overall distributions of the clouds. Results show that the proposed method shows a robust and accurate registration outcome, especially for the point cloud scene with great change of perspective and multiple repetitive symmetrical structures.