Abstract：Aiming at the problem of bridge deck segmentation of bridge point cloud, we propose a bridge deck segmentation method based on the fusion of plane elements in adjacent regions. Firstly, the point cloud is voxelized to obtain the voxel point cloud, and the supervoxel space and normal constraint criteria are used to over-segment the bridge point cloud. Then, we use the direction of the patch normal and the model length to perform the patch filter to obtain the candidate patch containing the bridge deck. Finally, we perform the plane element fusion of the adjacent area on the candidate patch and use the statistical distribution method to filter the fusion area to obtain the bridge deck point cloud. Experiments show that the method in this paper can effectively segment the bridge deck of the bridge point cloud, has high stability, and retains the original data before voxelization. Compared with the region growing and the supervoxel region growing methods, the method in this paper integrates across regions. It has a better inhibitory effect and is more suitable for bridge deck segmentation of bridge point clouds.