For the linear edges,the method using the traditional Sobel operator for edge detection and direction discrimination can not accurately detect the direction.An algorithm for linear edge detection using Hessian matrix with half pixel precision is proposed.The pixels in the lost block of the image are classified,for different edges, we use different edge detection operator to detect the directions of the edges.The corrupted image is concealed by directional extrapolation according to the detected edge direction.The experimental results show that Hessian matrix with half pixel precision for linear edges detection can get more precise direction,the average PSNR(Peak Signul Noise Rate) of the reconstructed image is 0.2~0.4 dB higher than that of other existing techniques.

Extended research and practical application about DSA(Digital Signature Algorithm) are very wide, they are all based on the security of DSA itself. Aiming at the main known attacks of evaluating secret key with the help of public data, exhaustive groping attack, birthday attack and fabrication attack based on known message, security of DSA is analyzed. The computing expressions are given about the attack methods. It is indicated that these attacks are equal to or more difficult than solving the discrete logarithm problem. Upon that, security intensity of DSA is shown a certain content. The potential weakness of DSA is given, lying in random number k, message independent signature r, sharing modulus p  and  q, Hash function, etc. And the corresponding solutions are presented.A favorable random numbers generator is designed for selecting right random numbers, so the attacks on low exponential and same random number could be avoided. Length of modulus  p  is selected reasonably, so the attack on the common signature r   could been counteracted. DSA prime is used as modulus in the case of high security intensity and low speed, so the common modulus attack could be resisted. SHA(Secure Hash Algorithm) with enough length message digest is utilized, so security of inline hash function could be ensured.

To obtain a shorter response time, all-optical switch, in a photonic crystal high refractive index layers all mixed Kerr medium, based on the Kerr nonlinear effect caused by the whole band moving principle, design of the two kinds of one-dimensional photonic crystal all-optical switch. Application of FD-TD(Finite Difference-Time Domain) method, the preparation of Matlab computer program, for all-optical switching characteristics of numerical analysis. Discuss the frequency of mixed effects on the all-optical switching effects. Observation of photon localization effect of photonic crystals to enhance nonlinear phenomena, verify photon localization effect and photonic crystal structure of a complete cycle of the relevant number of layers, number of storeys photon localization effect was not obvious enough.

In order to achieve synchronization between the media transmission,we analyzed the characteristics of multimedia data, made the composition of multimedia data in the time domain, studied the relationship between multimedia synchronization, and compared the current method of transmission of multimedia synchronization. And a relative time-stamp synchronization scheme based on feedback control without synchronous clock of whole power was brought forward.In order to assure media synchronization we set up the buffer in the receiving end to offset delay. The method is simple, because it does not require accurate time, but only marking audio and video time-stamp simultaneously.

In order to solve the public transportation route choice problem, the feature of public transportation network,compares algorithms of shortest route,  and selection algorithm of public transportation network routes is analyzed based on two-way search. In order to select optimum route, thinking about the factors which are time of transfers, cost of running and times of transfers, a compound evaluation index model is obtained.An example is given to valldate possibility of the algorithm and the model.Based on the database theory, the algorithm expresses the public transportation network with the database, realizes the most superior riding pass choice with the database inquiry technology, it is easy to realize, and to be efficient.

Flexible flow shop scheduling problem is a complex job shop scheduling problem. And the problem has polynomial difficulty. Most of the other cases the problem has NP difficulty. In order to solve the problem, a greedy algorithm to solve flexible flow shop scheduling is given, and the capability of the algorithm is evaluated. Even the approximate result solved by the greedy algorithm has a certain error with the optimal result; however, it has small time complexity, so it still makes sense to the job shop scheduling problem solving.

In order to improve the security of ELGamal digital signature scheme and speed up signature verification, based on its detailed analysis, the ELGamal digital signature scheme was deformed, in view of some problems in the application of this signature algorithm. The deformed ELGamal digital signature scheme was analyzed about the two aspects of security and time complexity. The results showed that as the deformed signature scheme reduced the number of modular inversion, the time of digital signature was shortened, and the deformed signature scheme had higher security and a better time complexity.

The average fitness is considered to be a constant in traditional schema theory. But in practice, the average fitness will grow while the best schema grows, so the best schema will not be exponential growth exactly. For this problem, an equivalent form of schema theory is deduced from the individual schema with a more precise expression of the average fitness of colony. The result is verified by experimental simulation. It is showed that the fitting deviation of the traditional schema theory is 2 times and 3.4 times more than the equivalent form in the paper.

Because spatial data have the characteristic of spatial autocorrelation, it makes the MLS(Multivariate Linear Regression) model unfit to spatial prediction. Due to account for spatial information, the SAR(Spatial AutoRegression) model can be used for spatial prediction, but it is computationally very expensive. We add spatial information into input variables by replacing each input variables with the weighted average of its neighbors and feed the new input variables to a MLS model to estimate model parameters, and then make spatial prediction, where MLS stands for this model. Experimental results show that the MLS model and the SAR model have almost identical effects on spatial prediction, while the MLS model is computationally more efficient than the SAR model.

In order to solve several problems of typical VPN(Virtual Private Network) network, including the deficiency of extension mechanism, deficiency of management, singleness of access terminal and access methods, a new network building mode called P2P（Peer-to-Peer）-VPN is discussed .The advanced search mechanism of P2P technology will be adopted to VPN network to improve the flexibility, expandability, access terminal and access method diversity of traditional VPN network. Meanwhile, the weak security of the network caused by P2P technology can be avoided by the effectiveness verification from the thirdparty certification and identity authentication.

In order to mine knowledge hiden in the structures that does not often changed in the XML(Extensible Markup Language) document changing history, this paper proposes a method to fiund the frozen structures, then uses a documentvector model composition by a group of frozen structures to represent an XML document, and uses the weighted Jaccard coefficient as similarity, then cluster XML documents based on the relative stable frozen structures which found in the XML document historical change process. Through experiments show that XML documents can be effective clustering base on frozen structures, after cluster, XML documents in each cluster have similar not often changed structures.

 In order to solve the problem of using INIC(Intelligent Network Interface Controller) to structure the MOST²⁵ （Media Oriented Systems Transport）ring network,the theory and the method of MOST²⁵ network are introduced,and  the design of the MOST²⁵ audio amplifier node and its implementation are provided. In the design of this node, the network controller used the intelligent network interface controller OS81050, the encoder used CS4341,the amplifier used NE5532. The application example of the design which has achieved a good sound effect is feasible.The results demonstrated that the design is effective and feasible, it has the advantages such as high fidelity, easy to install/uninstall.

The development condition of collaborative product design is never-ending varying, the process of product development need to be managed and various resources is dynamically organized according to development process to realize optimization of design process. This paper analyzes the characteristic of modern product collaborative development process and concludes the demand realizing process management model and puts forward multi-view process management modeling method. Model describes the restriction relation between related element and information flow. The model sustains function view and behavior view,organization view,information view of product development making by process information.

The robust stability analysis of three degrees of freedom helicopter singular systems is complex and difficult to design. Based on state-space theory, the linear dynamics model of three degrees of freedom helicopter was constructed. The conclusions that three degrees of freedom helicopter system is no feasible solution and nonrobust stability are obtained. The conditions for robust asymptotic stability of singular system was derived and proved using LMI(Linear Matrix Inequalities). The ruboust stability of state feedback control rate is found.The three degrees of freedom helicopter system state response curves, indicate that designed controller has better stability and robustness than the LQR(Linear Quadratic Regulator) controller.

 For the uncertainties of the multivariable, nonlinear and strong coupling flexible inverted pendulum system, considering the uncertainty of the system, a robust LQR（Linear Quadratic Regulator）controller was designed for the flexible connecting linear inverted pendulum. The simulation and the experimental results showed that the robustness of the proposed LQR controller is belter than the original LQR controller,and has a better antiinterference ability and stability.

In order to improve the robustness of human eye positioning, accuracy and detection speed,and improve the applicability of the algorithm,a methord was proposed that combining color information based on YC_bC_r color space with gray level projection and human iris saturation information based on HIS color space. Color was extracted with Gaussian distribution model of improved algorithm base on C_b, C_r Component of Image,and detected face region through analysis of color regions.The graylevel projection methods to locate eyebrow regions was adopted,the human iris saturation information was used to locate eyes accurately. Experimental results show that the right accuracy of eye positioning is 85%, positioning time is 0.072 6 s.

The traditional traffic monitoring methods only analyze protocols below the transport layer, which can not identify the  P2P（Peer-to-Peer） applications. In order to address the issue, we discuss characteristics and application scenario of traffic monitoring methods such as  RMON(Remote Network Monitoring), xFlow and probe. A test has been made in the metro access layer, convergence layer, and MAN（Metropolitan Area Network） export by using probe and MAN traffic distribution models are achieved and analyzed. The results show that it is validity to monitor traffic by using probes.

In order to improve the effect of the Moving human detection in computer vision, this paper used the OGHMs method. In the implementation process, we have simplified OGHMs（Orthogonal Gaussian-Hermite Moments）,used the Gaussian smoothing to the movement image, and added an erode algorithm to get more accurate results. Simultaneously several methods used in moving human detection are analyzed, such as temporal difference, background subtraction and optical flow. The experimental result indicated that the OGHMs which is added erode process has some advantage of common algorithms, and has strong anti-jamming capability. This algorithm has better solved the problem of large quantity of calculaion, it is not necessary to carry on the process of background rebuilding and updating. The detection result of this algorithm contains more movement information, it can finish the moving human detection more integrallty and accurately.