In the space-based surveillance technology, the 3D reconstruction technology to obtain 3D information of space targets is the most important. In this paper, we use the time-of-flight camera for data acquisition of satellite model, and solve the problems of obtaining the target object obscured, reflection, noise and field of view in the point cloud by coarse and fine registration of consecutive multi-frame point clouds, and propose an improved ICP (Iterative Closest Point) algorithm based on ISS (Intrinsic Shape Signatures) feature points. The error accumulation and transmission problems in the registration process of consecutive multi-frame point clouds are effectively solved by using loop closure detection and pose graph optimization. The model reconstruction is completed using the truncated signed distance function for the completed aligned point clouds. Finally, the reconstruction results are evaluated at two levels: point position error and model reconstruction accuracy.