Aiming at the precise positioning issue of the battery pack locking mechanism during the battery swapping process of new energy electric vehicles, a pose estimation method based on point cloud segmentation and principal component registration is proposed. The method first utilizes an instance segmentation network to extract the locking mechanism instances from the scene images. The depth data corresponding to the locking mechanism instances are projected into point clouds, and statistical filtering and voxel filtering are applied to denoise and downsample the point clouds. Secondly, by embedding the SGE attention mechanism module into the feature extraction layer of PointNet++ network, the spatial semantic features of points in the point cloud are enhanced, and the lock head point cloud is segmented. Finally, the spatial pose for locking mechanism is obtained by aligning the locking mechanism head point cloud with the target point cloud through principal component registration. Experimental results show that the pose estimation algorithm proposed in this paper has high accuracy and certain anti-interference capability, achieving a locking mechanism point cloud segmentation accuracy of 98. 02% , a translation error of 2. 239 mm, an angular error of 1. 822°, and an RMSE of 1. 495 mm, meeting the positioning accuracy requirements for battery swapping robots.