In order to avoid the complicated computer aided design（CAD） model construction from point cloud and realize the automatic welding of freely placed vertical plates,an automatic extraction method of welding trajectory from point cloud for robotic laser welding was proposed.The uniform sampling method was used to simplify the point cloud,the statistical filtering was applied to eliminate the measuring noise,and the overall tilt correction of the point cloud was performed at the same time.Then a data scanning strategy was established to realize the rapid segmentation of the top and bottom surfaces of the vertical plates.After that,based on the boundary points on the top surface identified by the Alpha Shape algorithm,distance constraints were used to improve the RANSAC algorithm to quickly fit the top boundary of the vertical plates.Subsequently,the key corners of combined vertical plates were identified through quick sorting and restoration of topological relationships of top boundary segments.On basis of these,the automatic extraction of welding trajectory for robotic laser welding was realized.The experimental results show that the proposed method can directly extract the welding trajectory from the point cloud without CAD model reconstruction,realizing automatic welding of freely placed vertical plates.