Yingsong Hu

3D Simulation of Leakage and Diffusion of Gas in Complex and Sealed Space

The leakage and diffusion accident of inflammable and explosive toxic gas has become a serious problem in the field of environment and security. Based on the optimized three-dimensional finite element model (FEM3) and the actual condition in complex and sealed space, a three-dimensional simulation system is developed to simulate the dynamic process of leakage and diffusion of gas in complex and sealed space. Changes of gas concentration in dynamic diffusion process are accurately simulated in the meanwhile. Experiment results show that this system can accurately simulate the changing condition of gas concentration in leakage and diffusion process. It provides a new method for simulation of gas leakage and diffusion.

Research on Samples Self-learning of BP Neural Network Based on Clustering

The generalization ability of neural network is an important aspect affecting its application. Meanwhile, the selection of training samples has a great impact on this ability. In order to improve the completeness of training samples, a method of samples self-learning of BP neural network based on clustering is put forward in this paper. By using the method of clustering, new samples can be collected in networks practical application. Network will be more suitable for practical situation after retraining. This method is proved to be simple and creditable. Furthermore, concerned experiments show that this method has obvious effect in improving generalization ability of neural network.

A Priority-Based Hierarchical Application Layer Multicast Model

Application level multicast is attracting more and more attention. To deal with the fact that different end hosts in real network have different network interface bandwidth, most of the present researches cannot take both bandwidth and latency into consideration well and keep the system robust simultaneously. In this paper, a priority-based hierarchical application layer multicast model (PH-ALM) is proposed. At the foundation of maintaining the superiority of the hierarchical ALM scheme, Both the latency and available bandwidth of peers are taken into account, and the idea of building local ALM tree in cluster and using priority to build ALM tree is proposed in PHALM. Periodical position adjusting strategy is also adopted to optimize and maintain local ALM trees. The simulation demonstrates that PH-ALM can well adapt to large scale multicast with various network bandwidth of end hosts.

Light re-rendering technology based on image brightness ratio

Illumination uniformity in the augmented reality has played a vital role, but most of the existing algorithm using hardware or fixed object to realize the corresponding function, has a great deal of binding. To this, this article proposes the illumination rendering technique based on image brightness ratio study the scene of the image, realize the simulation of light intensity, then access to the light source of information for modeling and virtual object import scenario, thus realizes the virtual objects and real scene illumination uniformity. Illumination rendering technique based on image by adapting to a variety of images of objects detection algorithm, implementation is a complex object in the environment of the light source, and accepts the influence of the ambient light shadow on this algorithm.

Rapid Skin Deformation Algorithm of the Three-Dimensional Human Body

The real-time skin deformation of 3D virtual human is one of the hot spots of research fields in 3D virtual human. On the basis of smooth skinning deformation algorithm, this paper proposes a whole body skin deformation method of real-time in motion state, which uses the simple polygon surface model to virtualize the human skin geometry model and uses the bone skin algorithm to drive the skin deformation. This method can simplify the bone level, decrease the complexity of skeleton subspace, simplify deformation algorithm, divide the skin into rigid areas and flexible areas to reduce the amount of calculation, and improve the efficiency of the deformation. At the same time, combining with the quaternion interpolation method, it further improves the effect of skin deformation of large-range movement area such as shoulder and hip. This method is applicable to any virtual human model of grid skin organization, and could generate real-time and realistic skin deformation, on the condition of any motion.

A path planning algorithm based on genetic and ant colony dynamic integration

Genetic algorithm has strong global search ability and robustness, but can not use the feedback of system and easy to do a lot of unnecessary redundancy iteration in the latter, resulting in the decline of convergence rate. Ant colony algorithm has the good characteristic of feedback, but due to lacking of initial pheromone, solving speed is slow. A method of adaptive dynamic integration is presented in the paper based on two algorithms, it uses genetic algorithm to generate initial pheromone distribution, and uses ant colony algorithm for dynamic integration of genetic operators in the latter. It not only improves the convergence rate of solutions, but also solves the problem of precocious and poor global search ability caused by the excessive convergence of ant colony algorithm in the late.

Multiple linear parameters algorithm in illumination consistency

According to outdoor scene light estimation, this paper gives a solution based on multiple linear parameters. Broadly, the programs can be divided into three different stages. They are image preprocessing stage, image decomposition stage and light parameter estimation stage. Image preprocessing stage is to guarantee the authenticity and validity of the image data. After analyzing the characteristics of the outdoor lighting, we can decompose the image. It is a linear model between sunlight and skylight. Then we get the multiple linear parameters based on the multiple linear illumination models. We achieve the simulation of each module in the system on MATLAB platform. And we can estimate all the parameters for red, green and black channel. At the same time the estimation of correlation, confidence interval and hypothesis test are achieved. Finally based on the results of the parameter estimation, the image is mixed. The process is real-time and highly efficient and the image mixed is highly lifelike to the original image. This linear model can effectively estimate the outdoor scene illumination parameters.

A Fast Accurate Interpretive Simulator Based on Shared Basic Block Cache Technique

Instruction set simulator plays a very important role in new micro-architecture design domain and cross-platform software design and development. The simulation speed performance and the simulation accuracy are the two most important factors that are used to evaluate simulators. Interpretive instruction simulation technique and compiled instruction simulation technique are the most common used and mature simulation technique. Interpretive instruction simulation technique is flexible and accurate, however, due to its simulation mechanism it has a slow simulation speed. In this paper, a new fast accurate interpretive simulator based on shared basic block cache technique is proposed, which takes advantages of the temporal and spatial locality principle to extremely improve the efficiency of time-consuming fetch-decoding phases of interpretive simulation technique. And the experimental tests show that the shared basic block interpretive simulation technique is especially superior to handle large target machine applications due it mechanism.

A Historical Connection Information Based Inter-Overlay Optimization Algorithm

Peer-to-Peer streaming system has increasingly wide applications. To improve the performance of the system, most researches are based on intra-overlay optimization at present. However this method has some disadvantages, such as the structure of the overlay is not globally optimal, the efficiency of resources utilization is low, and all of the leaf nodes cannot contribute any bandwidth to other nodes in multicast trees. In this paper, a historical connection information based inter-overlay optimization algorithm (HCI) is proposed, where each peer maintains multiple inter-tree connection tables as well as one intra-tree connection table. The algorithm makes full use of the original historical connection information. This results in that the nodes participating in multicast can use their nearest neighbour nodes so that the quality of streaming service is guaranteed. This algorithm can balance the load among the multicast group members to improve global resource utilization of P2P streaming system. The simulation results show that the HCI algorithm is flexible, easy for implementation, and able to efficiently improve the performance of the system.