Aiming at problems such as low efficiency and over-segmentation in complex background segmentation, a 3D point cloud segmentation method based on normal estimation is proposed. First, the RANSAC algorithm is improved, and normal constraints are introduced to remove most of the plane point clouds in the background. As a result, the target point cloud is separated from a large number of point clouds. Then European clustering is used to extract the target point cloud from it. This paper uses the zokor skull point cloud data for experiments, and experimental results show that the merged point cloud segmentation algorithm can accurately segment the target Zokor Skull point cloud, completely remove the redundant point cloud, with fast speed, high accuracy, which shows better performance than the same type of method.