In Low Earth orbit （LEO）satellite multi-beam communication scenario， the traditional fixed resource allocation algorithm can not meet the differences in channel capacity requirements of different users. In order to meet the requirements of users， the optimization model of minimum supply-demand difference of combining channel allocation， bandwidth allocation and power allocation was established， and Pattern Division Multiple Access technology （PDMA）was introduced to improve the utilization of channel resources. In view of the non-convex characteristic of the model， the optimal resource allocation strategy learned by the Q-learning algorithm was used to allocate the channel capacity suitable for each user， and a reward threshold was introduced to further improve the algorithm， speeding up the convergence and minimizing the difference between supply and demand when the algorithm converged. The simulation results show that the convergence speed of the improved algorithm is about 3.33 times that before improvement； the improved algorithm can meet larger user requirement， about 14% higher than the Q-learning algorithm before improvement， about 2.14 times that of the traditional fixed algorithm.

In order to enrich and improve the ability of the existing models for reasoning and predicting with 3D cardinal direction relations and enhance the usability of the existing models, and then better meet the demands of real applications for complex 3D spatial data, the inverse reasoning of 3D cardinal direction relations was studied. After deeply studying the theory of n-dimensional block algebra, an algorithm for computing the inverse of the basic 3D cardinal direction relations on the basis of 3D block algebra was devised. Theoretical analysis and the results of the example show that the proposed algorithm is correct and complete. This work can better enhance the power of intelligent analysis and processing for the complex 3D direction relations of the spatial database.

An octree based on the rapid k nearest search was presented for scattered points from 3D scanner. Through the establishment of bounding box on point sets, octree was used to record segmentation process, so that the search of point neighbors was limited to the bounding box of sample points and its neighbor, and through pruning strategies to further narrow the scope of the search.The experiments on a great deal of real data show that the search speed can be well improved.

On the basis of the analysis of the existing reconstruction methods limitations, a fast neural network based algorithm for 3D surface reconstruction from cloud data was presented. Firstly, a unitary processing for the cloud data was made. And then the contour lines were extracted and the surface based contour lines were segmented. The method can directly from the neural network value matrix get the right curve control points or surface control mesh, and through neural network values restriction achieve the right curve or patch of smooth connection. Experimental results show that this method can quickly obtain good shape Mesh.