Abstract：Aiming at the scarcity of real label data caused by equipment occlusion in the construction of

substation 3D map，a 3D map construction method was proposed based on semi-supervised deep closest point

（SEMI-DCP）. Firstly，the iterative closest point algorithm was used to generate pseudo-labels for the registered

point cloud data，and the complex occlusion scenes were effectively dealt with by selecting appropriate pseudolabels.Secondly，combined with a small amount of real label and pseudo-label data，the accuracy and convergence of the point cloud registration model were gradually improved through the alternating iteration method. Finally，the model was trained and tested on the ModelNet40 dataset. The experimental results show that the proposed method based on semi-supervised deep point cloud registration is superior to the traditional method，especially in the case of scarce real label data.

      Key words：substation 3D map；semi-supervised learning；deep closest poin（t DCP）；pseudo labelling；iterative closest point（ICP）algorithm