The laser 3D scanning model reconstruction is an effective technical means for the reconstruction of 3D virtual scenes of old nuclear facilities. However, the existing technology have the following problems: First, the scanning equipment needs to be installed on the triangular fixed bracket, which requires personnel to be installed on site, thereby bringing radiation safety risks to personnel; second, it requires the person to reconstruct and splice the model one face by one face, which is inefficient. Aiming to deal with these problems, this paper proposes a mobile laser 3D scanning method and a rapid model reconstruction technology. These methods include: (1) In order to realize fast scanning of large-scale scenes, and at the same time ensure accurate scanning of local important objects, a laser three-dimensional composite scanning method is proposed. (2) Design and development of a remote control automatic walking and fixing device, which can transport the laser 3D scanner to the position of measurement which is needed. The device has a certain ability to overcome obstacles, and can be reliably fixed to ensure the stability and scanning accuracy of the laser scanner at high speed working station. You can remotely control the laser scanner to start or stop working. The difficulty of this technology is how to make sure the device has the flexibility in motion and the stability on working; (3) An automatic model reconstruction and feature identification method based on a large number of point cloud data is proposed, and the corresponding software is developed. This method can realize the automatic recognition of the standard and common models, and the automatic model reconstruction of the point cloud data in the scene. To achieve this goal, the neural network algorithm is used. These technologies can effectively reduce the radiation safety risk of personnel during laser scanning in high-radioactive places, reduce the intensity of personnel operations, and improve work efficiency.