针对带有强噪声离散点云数据曲率计算问题,提出一种基于稳健统计的曲率估计方法.首先,用一个二次曲面拟合三维空间采样点处的局部形状；其次,随机地选择该采样点邻域内的子集,多次执行这样的拟合过程,通过变窗宽的最大核密度估计,就得到了最优拟合曲面；最后,将采样点投影到该曲面上,计算投影点曲率信息,就得到采样点曲率.实验结果表明,所提方法对噪声和离群点是稳健的,特别是随着噪声方差的增大,要明显好于传统的抛物拟合方法.

For the low sampling efficiency and poor sampling efficiency in the traditional ray casting algorithm,a volume rendering algorithm with a new sample synthesis function is proposed.This algorithm applies Blinn-Phong illumination model and uses early opacity termination.It's implemented by vertex programs and fragment programs in Cg language.The results of experiment show that this algorithm can not only enhance the effect of traditional ray-casting algorithm,but also accelerate the speed of the algorithm.

A robust dual digital blind watermark is proposed based on three-dimensional mesh model in order to solve the weakness of the single watermark's difficulty to resist against various attacks.The scheme embeds the watermark by modifying the location of vertices in their related space and topological information of 3D mesh,which makes the model resist not only large scaled cropping operation but also high strong noising.With this approach,the original model is not required during detecting watermarks.Experimental results show that the algorithm has good visualizing quality and robustness which can be against the attacks such as translation,rotation,scaling,model cropping and vertex permutation etc.

In order to establish the forecasting model of SaaS(Software as a Service)operation,the combined fore-casting model based on principal component analysis is proposed,and is applied in the forecasting of SaaS opera-tion.A novel combined forecasting model based on principal component analysis is established,which uses three different forecasting models—phase space reconstruction model,grey model,cubic exponential smoothing model.The simulation results show that the combined forecasting model based on principal component analysis,which takes advantages of the unique strength of each model,can provide more precise forecasting than that of each indi-vidual model.This combined forecasting model is demonstrated to be efficient for the forecasting of SaaS operation.

Aim To improve the surface reconstruction algorithm based on radical basis function(RBF).Methods Due to the amount of data is too large,surface reconstruction can not be done in the low-end computers,hence by following dividing and conquering thoughts,the point cloud surfaces are firstly split,and then rebuilt.Results The presented method can obviously improve the efficiency of surface reconstruction and realize the surface reconstruction on small and medium size computers.Conclusion For the irregular object to be reconstructed,the use of the bounding box of any direction can better reduce the number of operations than the use of axial bounding box in the reconstruction process.

Aim For the low sampling efficiency and poor drawing efficiency in the traditional ray casting algorithm based on GPU,the algorithm is used to optimize and acceleratien in CUDA structure.Methods Firstly,the volume data are encoded by octree,it can eliminate the parts which do not contribute to reconstructed image efficiently.Secondly the data is bound to 3D texture and the thread is distributed according to the size of volume data,parallel computing is used when finding out the intersection of each of the light and the volume data.Finally the ray casting algorithm is realized in CUDA kernel.Results The experimental results show that the algorithm can increase the speed by 7~15 times,and enhance the algorithm.Conclusion The algorithm can improve the speed and effect of traditional ray-casting algorithm based on GPU.

This paper proposes an algorithm to generate and render fireworks in real-time based on programmable graphics hardware, which divides the queue into several regions to save the time when waiting for CPU, and uses the XML external data and XNA framework to achieve the fast simulation of fireworks. The experimental results show that the algorithm can draw the fireworks vividly, meets the demand of real-timing very well, and has a great scalability.

In the spherical coordinate system, multi-layered geometry image which represent the point cloud surface exist shortcoming.This paper proposes using cylindrical coordinate system, which is the approach of multi-layered geometry image representation of point cloud surfaces.The approach created can precisely represent point cloud surfaces structure with high complexity, and it is much simpler to implement in the compression, editing, simplification, model retrieval and progressive transmission of point cloud surfaces.