In order to solve the problem of action localization in video, a weakly supervised method based on template matching was proposed. Firstly, several candidate bounding boxes of the action subject position were given on each frame of the video, and then these candidate bounding boxes were connected in chronological order to form action proposals. Secondly, action templates were obtained from some frames of the training set video. Finally, the optimal model parameters were obtained after model training by using action proposals and action templates. In the experiments on UCF-sports dataset, the method has the accuracy of the action classification increased by 0.3 percentage points compared with TLSVM (Transfer Latent Support Vector Machine) method; when the overlapping threshold is 0.2, the method has the accuracy of action localization increased by 28.21 percentage points compared with CRANE method. Experimental results show that the proposed method can not only reduce the workload of dataset annotation, but also improve the accuracy of action classification and action localization.

Focusing on the energy efficiency optimization of network traffic transmission, a Dynamic Weight of Weight Fair Queue algorithm (DW_WFQ) was proposed. In this algorithm, weights for various types of flows were dynamically allocated based on Weight Fair Queue (WFQ) algorithm, and service rates for traffic flows were assigned in a more flexible way. Then the energy consumption model of the scheduling algorithm was derived by combining with efficiency function of continuous flow speed scaling model, and the optimization of energy efficiency was carried out. At last, simulation test and comparison on DW_WFQ, First Come First Server (FCFS) and WFQ were given using Matlab. The simulation results show that the proposed traffic scheduling algorithm can effectively reduce system energy consumption, and meet the Quality of Service (QoS) requirement at the same time.

To resolve the problems on description and proof of semantic property preservation in Model-Driven Software Development (MDSD), a formal approach was proposed for software architecture model on the basis of type category theory and process algebra. The semantic constraints of component specifications which should be kept through model transformation, were deeply analyzed and discussed. From the view of diagram structure, port and configuration constraints, external behavior and component substitutability, the problem of property preservation was described, and the corresponding criteria was built at the same time. The framework provides a guidance for the definition of model transformation rules, and provides the basis to verify the correctness of model transformation as well as to analyze the effect of model transformation. The application research shows that, the approach enhances semantic description capabilities of component model, and can be used as an effective supplement for existing software modeling method.

For the scheduling optimization problem of cross-over twin Automated Stacking Crane (ASC), a multi-objective mathematical formulation model was proposed with considering the conflicts between two ASCs while reaching the same bay. The operation sequences of ASC was optimized, the optimal task sequence, task finish time and ASC's idle time after avoiding the conflict were gained and the practicality of optimized model was proved. Four experiment scenarios were performed to further analyze the efficiency difference between twin ASC and single ASC, and the effects of parameter changes. The results illustrate that the utilization rate of equipment of twin ASC is 107% lower, and the operation efficiency is 35% higher than single ASC. The ASV's total operation time will be decreased while the number of containers decreases and the travel speed of ASC increases. While the ratio of storages and retrieves of containers is 1, the optimized tasks' total operation time and ASC's idle time will be gained. It shows that the container terminals can improve their operation through adjusting the ratio of storages and retrieves during a short period or optimizing ASC travel speed.

To solve the problems of Web service composition and verification, a formal model was proposed based on the framework of category theory. Process Algebra was introduced into the framework to describe the external behavior of service component, establishing a formal semantic model for the architecture of Web service system. The service network was described with category diagrams, in which Web services were used as categorical objects, and the interactive and composition relationships between services were used as morphisms. On the basis of the formal definitions of service interface, Web service and service composition, a further analysis and discussion about the semantics of service composition and interaction was undertaken. The concepts on Web service substitutability and service request satisfiability were formally defined. The application research shows that the proposed framework enhances semantic description capabilities of Web service architecture.

To solve the low running speed problem of Knuth39 random number generator, a Knuth39 parallelization method based on Many Integrated Core (MIC) platform was proposed. Firstly, the random number sequence of Knuth39 generator was divided into subsequences by regular interval. Then, the random numbers were generated by every thread from the corresponding subsequence's starting point. Finally, the random number sequences generated by all threads were combined into the final sequence. The experimental results show that the parallelized Knuth39 generator successfully passed 452 tests of TestU01, the results are the same as those of Knuth39 generator without parallelization. Compared with single thread on Central Processing Unit (CPU), the optimal speed-up ratio on MIC platform is 15.69 times. The proposed method improves the running speed of Knuth39 generator effectively, ensures the randomness of the generated sequences, and it is more suitable for high performance computing.

For the problem of that the particle easily falls into local optimum when avoiding obstacles and cannot walk forward, the method of retreating the position of the local optimal particle to its historical best position and searching the feasible and optimal location in the 8-neighborhood of this position was proposed. The position which is nearest to the target and is not a barrier could be found by this method, then the particle was moved to it. Meanwhile, the global optimal position of current generation particle swarm could be found, and the location of each particle was set for this position to continue iteration. In the experiments for obstacle avoidance in grid map, the fact was found that when encountering obstacles, the particles fell into local optimum by the traditional methods, but particles could successfully avoid obstacles and arrive the end by using the improved algorithm. The improved algorithm was introduced into the 3D visual simulation system, in large map with many obstacles scenarios, the probability of falling into local optimum of the particle was reached to 50% which resulted in routing failure. After adding a circular slope to the obstacles for the further improvement of the algorithm, the probability of successful path-finding of the particle was increased to 83%. The experimental results show that the search ability of the improved algorithm is increased and it can effectively plan path in a complex scene.

Pre-positioning facilities for collecting tents held by communities is helpful to improve the efficacy of tents collection in post-earthquake relief. A problem for pre-positioning facilities that collect the tents held by commnities considering various earthquake scenarios was studied. A bi-stage stochastic programming model was formulated for the facility location problem. The model aimed at minimizing economic costs and time-related costs. Considering the effects of weights and other parameters on the solutions, five experiments were performed and analyzed. A case was used to verify the practicability of the developed model that located the facilities of tents collection from communities in Pudong New District, Shanghai. Finally, a scenario based on Ya'an earthquake was used to analyze the influence of parameters on the solution under such a certain scenario.

With the growth of cloud technology, distributed application platform develops towards elasticity resources and dynamic migration environment. The refinement of distributed application access control policies was associated with resources and environment, which also needs to improve performance to adapt to the dynamics. Although present access control space policies conflict analysis methods could be used in the conflict analysis of distributed application access control policies refinement. The granularity of its calculating unit is too fine to make batter performance. In this article, the authors designed a conflict analysis algorithm used in distributed application access control policies refinement, the conflict analysis algorithm was based on recursive calculation the intersection of sets and the calculation unit of the algorithm was permission assignment unit which improved computing granularity. The experimental results and analysis show that the proposed algorithm has better performance, and fits the needs of improving computing performance of cloud platform access control policies refinement.

Aiming at solving the misplaced and mismatch problems when 3D hair attaches different head models in the reuse process, a 3D hair reusable method based on scalp layer feature points was proposed. Firstly, according to the data storage structure of the model file, the scalp layer was isolated from the hair model and the feature points were extracted. Secondly, combined with the 2D face image detection method, feature points of the range of hair root on head cortex were extracted. Then shift and zoom coefficients were calculated by the above mentioned feature points. Finally, the fitting process of scalp layer and head model was handled individually. Eventually the 3D hair was adapted to the target head model,which could keep hair styling information without loss. The effect of a close fit between the scalp layer and the head model was achieved. The experimental results show that the proposed method can effectively improve the reusability of 3D hair model, and it is not influenced by the restrictions of the hair model personality part and the distribution area.

To resolve the problems of system composition and semantic verification in the construction process of distributed computing model, a semantic description framework for Agent-based distributed computing system was proposed based on category theory and process algebra. The structural semantics of the system model was described within category diagrams, and the relations between components were formally described by morphisms. On this basis, the semantic properties that should be preserved during the process of system modeling, refinement and migration were further analyzed and discussed. The application research shows that the proposed framework can not only be used for distributed system modeling, but also be used for the correctness analysis of system decomposition and composition.

Large sample size and characteristics of high dimension affects the K-Nearest Neighbor (KNN) classification algorithm in classification performance. Therefore, an improved KNN method named AMI&CD-KNN was put forward, which had feature dimension reduction and pruning mechanism based on average mutual information and class distinction of the feature parameters. Firstly, it used the concept of average mutual information in entropy to measure the accuracy of the feature parameters reflecting the class characteristic information. Secondly, it described the class distinction through the feature parameters odd ratio relative class and its distribution probability in the dataset, measured the size of the amount of class information provided by feature parameters. Finally, the relation was established between average mutual information and class distinction, and the sample pruning method was designed. Therefore, the classification was speeded up while ensuring the classification accuracy. The theoretical analysis and simulation experiment show that compared with the KNN and other algorithms with pruning mechanism, the improved algorithm has better classification generalization.

To improve forecast accuracy of water demand when using Projection Pursuit (PP) model which are high-dimensional, non-normality and nonlinear, an Ant Colony Algorithm (ACA) was used for the parameter optimization of the model. ACA was improved to self-adaptive control pheromone on the grids divided by definitional domains of the model parameters. A case for water demand prediction was emulated according to the improved ACA and PP model. Then prediction accuracy from the improved ACA was compared with the results from Artificial Immune Algorithm (AIA) and BP Artificial Neural Network (BPANN) model, respectively. It is shown that: 1) the absolute relative errors of fitting accuracy are less than 2% from ACA and less than 10% from AIA and BPANN; 2) the absolute relative errors of prediction accuracy are less than 6%, 11% and 12% from ACA, AIA and BPANN, respectively; 3) ACA can converge to global optimal solution with higher convergence rate. Therefore, the improved ACA for optimizing the parameters of PP water demand prediction model is significantly better than the AIA and BPANN. This method can be applied to other similar high-dimensional and nonlinear problems.

This paper proposes a classified strain limiting method which deals with unreal stretch deformation in clothing simulation with physical-based mass-spring model. The method primarily include two processing module. The first module is classification. it uses velocity computed by integrating system as input parameters first, and then judges some point whether needs strain limiting by the energy method, finally through the judged result divides point set into two types : needing strain limiting and not. The second module is strain limiting. It defines the threshold value of spring deformation and three variables for representing the restrictive proportions in principal strain direction, and then by the line strain theory computes the strain tensor of spring, finally obtains specific restrictive proportions and updates the position of corresponding point. The method could guarantee natural simulation results and eliminate unreal stretch deformation, and does not require all elements of the strain limiting for processing, reducing the computational cost to ensure real-time. Results indicate that the method have good effect and efficiency.

Focusing on the random packet loss, the high failure rate of failure detection for network storage system with random packet loss was studied. A Fault Detection (FD) for network storage with random packet dropout was presented. The residual generation and residual evaluation as well as False Alarm Rate （FAR） were used in the approach. First, residual generation was carried out in the periodic system framework. Then, residual evaluation was got by making use of the stochastic properties of the random packet loss. Finally, performance evaluation of the computation of FAR is fulfilled with the assistance of Chebyshev's inequality, and the algorithms of fault detection were given. The simulation results show that this approach can effectively detect the fault. Moreover, this approach is sensitive to fault.

Considering massive scale of cloud storage solutions, the traditional data distribution strategy confronts challenges to improve scalability and flexibility. This paper proposed an efficient data distribution strategy. Based on consistent hashing algorithm, the strategy introduced the virtualization technology, and employed virtual node to improve load balance. Moreover, the strategy used a new capacity-aware method to improve the performance of the cloud storage system. The evaluation experiments demonstrate that the proposed data distribution strategy improves system performance in both homogeneous and heterogeneous distributed storage architectures.

The existing TCP multicast congestion control mechanisms do not have speed smoothness, Round-Trip Time (RTT)-fairness, and high efficiency in highspeed networks. To solve these inherent disadvantages derived from TCP, an Adaptive High-Speed Multicast (AHSM) congestion control mechanism was proposed based on population ecology theory. The new mechanism estimated the bottleneck bandwidth and the rate of background traffic at each receiver. Then the estimated values were used in population ecology model to calculate requested service rate. The receiver with lowest requested service rate was selected as representative to control the sending rate via feedbacks. The simulation results show that the new mechanism has speed smoothness and RTT-fairness, it also has high transmission efficiency in high-speed networks. It can share bandwidth with unicast flow in both low and high-speed networks.

Poor generalization of feature parameters classification and large category computation reduce the classification performace of K-Nearest Neighbor (KNN). An improved KNN based on union entropy under the attribute reduction condition was proposed. Firstly, the size of classification impact of data feature was measured by calculating the union entropy of two feature parameters relative to any two condition attributes, and the intrinsic relation was established between classified features and the specific classification process. Then, the method which reduced condition attributes according feature union entropy set was given. The theoretical analysis and the simulation experiment show that compared with the classical KNN, the improved algorithm has better classification performance.

In order to resolve the share-space problem about parsing trees in operator-precedence based parsing of Binary Combinatorial Grammar (BCG), a chart algorithm based parsing approach was proposed to reduce the spent space and to improve the parsing efficiency. All edges produced during the parsing process were stored by a tabular structure. By tracing the table, parsing trees were constructed according to the span of edges, so that these edges were no longer restricted in a particular tree. Then, the pruning in parsing process was achieved according to the characteristics of BCG grammar. The experiment showed that all of the spent time, produced edges and result trees of this approach were less than those of traditional chart algorithm and operator-precedence based algorithm.

The schemes to solve the high bit error problem in wireless networks were discussed. Aiming to improve the performance of TCP protocal in mobile environments, two typical TCP models of the pure end-to-end approaches, TCPW and TCP-Real,  were researched. The pure end-to-end schemes have the advantage of preserving the end-to-end semantics of TCP protocol. Simulation results with different bit error conditions of these protocols reveal an evident improvement in terms of TCP performance in relative to others, especially when the bit error rate is high.

It is necessary to study those watermark algorithms appropriate to the special feature of fractal transform. By analyzing related parameters in the process of image fractal transform and taking some trials, a new watermark algorithm embedding and extracting the watermark information was proposed by modulating the parameter of contrast factor which mapped the domain blocks to their best matched blocks in fractal transformation. Higher watermark capacity could be embedded into image at low algorithm complexity without having perceptible influence on the quality of the image. The extraction algorithm avoided the loss of watermark information because of the missed synchronization between original image and watermarked image, which improved the accuracy in extraction. Experiment results show the effectiveness of the algorithm.