For multi-robot assembly of two large cabins, the pose measurement of cabin is indispensable. Traditionally, this is achieved by installing targets or markers on the cabins, making the process inefficient. To address this, a pose measurement method based on point cloud which can be readily collected by 3D vision sensor is proposed in this paper. The cylinder model of the cabin is firstly established according to its geometry. The
point clouds of cabins are segmented from the point cloud scene using RANSAC model fitting. Then the relative pose between two cabins can be calculated using the segmented point clouds and model coefficients. It is validated in the experiment that the proposed method is accurate and robust for pose measurement of large cabins.