Abstract:A method to animate water in Chinese painting is presented in this paper. This approach first detects positions of water forms in the hand drawn Chinese painting and synthesizes those forms using autoregressive model. Next, water forms are rendered using an image based rendering technique to produce water forms with the look of Chinese painting. Finally, water forms are animated in hand drawn environments thus the authors bring water forms in Chinese painting to life. The given examples demonstrate the capability of our method to capture the statistic properties associated with the sample data and to produce a variety of water forms for non-photorealistic animation.

Abstract:In general, most 3D parametric design systems use plane and sphere as basic tools to draw a three-dimensional diagram. In this paper, we introduce a class of new drawing tools: conicoid. The scope of three-dimensional diagram that can be drawn with conicoid is strictly larger than that with plane and sphere only. We prove that a diagram can be drawn with conicoid sequentially if and only if it can be described by a set of triangular equations of degree less than nine. Spatial Appolonius drawing problems can be completely solved with conicoid.

Abstract:Two different algorithmic approaches, the integral method and the piecewise degree reduction method, to the degree reduction of interval B-spline curves are presented. Examples are provided to demonstrate the algorithms and to compare the two different approaches. The experimental results show that the piecewise degree reduction method is generally more efficient and produces tighter bound than the integral method.

Abstract:Illumination is an important aspect in realistic graphics rendering and most image related applications. The paper proposes a pure image based rendering method to synthesis the variation of illumination impacted on an object. Rather than estimating the BRDF function or any parameters in a reflectance model, Singular Value Decomposition (SVD) is used to estimate a low dimensional approximation of the image set of a Lambertian object under both changing of illumination and geometrical pose. In the method, light directions are analytically derived from the samples and the base images as well as images with known of the class of the objects. Images with any novel lighting directions can be rendered very efficiently by properly linear combination of the base images. Smooth morphing sequence of lighting and pose variation can be generated by linear interpolation of the SVD coefficients. A feature of this method is that it supplies a simple and compact way to represent dataset and synthesize novel images, which lends it to be suitable for many image-based applications.

Abstract:The convexity of curves and surfaces is an important property in the field of Computer Aided Geometric Design (CAGD). This paper tries to tackle the positive and convex problem of polynomials. Convexity can be solved by positivity. An algorithm for the positivity of polynomials is developed by extending the classic Sturm theorem. Hence, a necessary and sufficient condition for the positivity of polynomials of arbitrary degree is presented in this paper. A practical algorithm to express this condition in terms of the coefficients of the polynomials is also given.

Abstract:Filament winding pattern plays a key role in computer-aided filament winding, unfortunately there is no reliable method for non-axisymmetric filament winding pattern design. In addition, a substantial amount of data storage is an obstacle to non-axisymmetric filament winding. A filament winding pattern for elbows and a new data processing method are presented in this paper.

Abstract:A simple method in the approximation of ellipsoid by bicubic polynomials is given in this paper, the error is approximately 273×10^-6 for ellipse and 545×10^-6 for ellipsoid.

Abstract:MARS and Rijndael are analyzed by using power attack. It is shown that the complexities on MARS are 2²⁰⁸, 2¹⁶⁸ and 2¹¹⁶ for 256, 192 and 128 bits key respectively, the complexities on Rijndael are 2¹³¹, 2⁹⁹ and 267 for 256, 192 and 128 bits key respectively. Although the complexities are too big to implement, the approach presented in this paper reduces greatly the key size of MARS and Rijndeal.

Abstract:In this paper, model and design framework of a new database natural language (Chinese) interface technique based on the restrictive Chinese are presented. A middle language with the multi-track structure and anew policies in the transformation from middle language to SQL which implements the search approach passing relational way are described. The experimental results show that its expression is more natural and the understanding and processing capability of Chinese statements are improved evidently, because the system uses no-procedural Chinese inquiry statements.

Abstract:Reliable multicast is a useful network service but challenged by the Internet heterogeneity and the nonsupport of IP Multicast in wide-area Internet. In this paper, an application-level semantic reliable multicast (ASRM) is presented, which realizes application-level semantic reliable multicast data delivery. ASRM uses a hybrid way that combines unicast delivery with multicast delivery instead of using global IP multicast as the basis. ASRM provides a simpler and more natural way to name and locate the sessions.During the data forwarding,ASRM can use application-defined semantics to transcode content according to the transcoding models,rules and user'preferences.Thus,it can cope with the heterogeneity problem very well.ASRM is suitable for wide-area, limited scale,reliable multicast applications.

Abstract:Lacking of feature validity maintaining function has become a serious problem with existing feature modeling systems. In this paper, a representation for feature validity condition based on the extended attributed adjacency graph is proposed, and a novel approach is particularly proposed to feature validity maintaining using local feature recognition technique. The approach can not only automatically detect if the feature validity is destroyed, but also accurately determine the reason why the feature becomes invalid and the state of the destroyed feature.Furthermore,the approach can automatically renovate feature model based on the user's intention.

Abstract:It is difficult for conventional single population-based evolutionary algorithms (conventionalevolutionary algorithms-CEAs) to solve epistatic interaction problems. Based on computational model ofcooperative coevolution inspired by the coevolutionary phenomena of natural species, a coevolutionary algorithm (CoEA) for solving NK-landscape. problem is proposed. Some problems related to the interactions among species and individual s fitness computation are discussed. Mathematical analysis shows that the exponential increase. index of CoEA is higher than that of CEA for a schema which fitness is higher than the average fitness of population.Simulation results verify the theoretical result,and show that the coevolutionary algorithm is more efficient and effective than CEA in solving epistatic interactions problems.

Abstract:In this paper, a MBLP (modified BLP) security model is presented which is PAC (privileged access control) oriented. Some concepts such as role management, domain compartment, limitation of covert channels and defence of viruses, are introduced. MBLP is applied to SecLinux, which is a self-developed secure operating system the authors.

Abstract:Traditional work of quantity survey is time consuming and low quantity. To do the work automatically by computer is a good solution. A rule-based system for interpretation of vectorized construction drawings automatically (AISD) is presented in this paper. Based on summarizing the features and rules, it recognizes graphical primitives and symbols in construction drawings, analyzes relationships between them, interprets information of construction element and calculates the quantity of reinforcement steel. Based on the study of features of construction drawings,a group of rules are developed for interpretation of different kinds of construction drawings.Experimental results show that it provides a sound solution for understanding and interpretation of construction drawings.

Abstract:In order to support distributed multimedia application description and development, a multimediacommunication mechanism-QoS (quality-of-service)-controlled pipe is presented as a QoS-based platformpoint-to-point multimedia communication abstraction. The formal definitions of QoS-controlled pipe are introduced and its enforcement procedures are discussed in detail. It has been applied to distributed multimedia information-on-demand systems and distributed multimedia systems successfully. Experimental results show that it not only can act as a comprehensiv and integrated of point-to-point multimedia communication mechanism,but also can act as the basis of multimedia group communication mechanism.

Abstract:The shot based classification and retrieval is very important for video database organization and access. In this paper, a new approach NFL (nearest feature line) used in shot retrieval is presented. Key-Frames in shot are looked as feature points to represent the shot in feature space. Lines connecting the feature points are further used to approximate the variations in the whole shot. The similarity between the query image and the shots in video database are measured by calculating the distance between the query image and the feature lines in feature space.To make it more suitable to video data,the original NFL method by adding constrains on the feature lines is improved.Experimental results show that the improved NFL method is better than the traditional classification methods such as the nearest neighbor(NN)and the nearest center(NC).

Abstract:In order to satisfy the quality of service (QoS) of the application on Internet, Internet engineering task force (IETF) has presented a set of network service mode such as IntServ/RSVP (Integrated Service/Resource Reservation Protocol), DiffServ (differentiated service) and MPLS (multi protocol label switching). All of these service modes need a correlative QoS routing mechanism and algorithm to assign an adapted route. Based on the research of existing routing algorithms, a new pre-computed QoS routing algorithm CCPF(classified minimum cost path first)is put forward,which uses bandwidth as a basic metric and gives a weighted cost function as a path optimal criterion.By simulation,this algorithm is compared with the others.The results show that it can get differet routes among different bandwidth-ranges to support multipl service modes.

Abstract:The traditional approaches are suffering from the problems of expensive computation, low efficiency, and lack of vitality. A new approach called tight-clothing and camera calibration based human animation is proposed in this paper. After videotaping the model dressed in tight-clothing, people can track human joint in image sequence automatically, then construct corresponding three-dimension human motion skeleton sequence under the perspective projection, finally the mapping of human motion onto the motion of other roles is completed by making BVH(biovision hierarchy)file.This approach has the characteristics of the rich source materal,efficient production,low cost and realistic animation result.It is possible to release the animators from the redious work.Also such technique can step into families for its low cost.

Abstract:How to describe the distributed array is a key point in parallel computing based on networks. Based on the general requirement of the parallel implementation, this paper first discusses the structure and necessary components of DAD (distributed array descriptor). Then a certain DAD structure used by p-HPF parallel compiling system is introduced in detail. Further, for the sparse and the dense storage model, the methods are given out to convert global arrays to local ones in three distribution Blocks, the Cyclic and the Block_cuclic(k)respectively.These methods have been implemented in p-HPF compiler and proved to be effective.Finally,the importance of the distributed array description standard for portability and reuse of the parallel computing systems is discussed.

Abstract:Name resolution mechanism is a very important field in the research of mobile agent technology. At present, for complexity of the algorithm, it is difficult for most traditional DNS-based name resolution mechanisms to implement position independent access of mobile object at application level. But it is convenient to construct communication model and navigation model for mobile agent system using global and position independent name resolution. In order to build a global/universal naming method and resolve position transparent problem at application level in mobile agent system,a name resolution mechanism--UON(universal object naming)is presented in this paper.Finally the detailed inolenebtation and the related algorithm are desctibed in three aspects(object registration,object move and mobile object access).

Abstract:Internet is facing two problems simultaneously: a faster switching/routing infrastructure and guaranteed quality-of-service (QoS). Each problem can be solved independently. High performance routers can be made faster by using input-queued crossbars instead of shared memory systems. QoS can be provided by usingpacket fair queuing (PFQ) algorithm. Until now, however, the two solutions have been mutually exclusive-all ofthe work on PFQ algorithm has required that routers use output-queuing or centralize shared menory.In this paper, on the basis of CIOQ(combined input output queuing)architecture,a packt scheduling algorithm DF²Q(distributed feedback fair queeuing)is designed and implemented.The most important feature of this algorithm is the introducing of feedback mechanism.the perfomance of DF²Qis analyzed and discussed.Experimental resuts show that it can avoid internal congestion effectively and improve the efficiency of resource utilizing.

Abstract:Coordination is one of the basic problems in distributed component systems. But up to now this problem has not been solved yet properly. According to the real application, the Concerto model is proposed, which is based on Petri net. Expanding the semantics of Petri net, Concerto model introduces control buffer and data buffer, which reflect the control dependency and data dependency respectively, to unify the present control-driven and data-driven coordination models. For the execution of Concerto model, driving mode,action rule and Concerto engine are forward.There are four kind of driving modes, depended on operation time,minimal time,maximal time and average time respectively.They are useful in application domains such as real-time systems,flow controlling,and task scheduling.As the core of model execution,Concerto engine controls the coordination of distrbuted component system by following certain arbitration rule, which resolves the deadlock and starvation.

Abstract:Line drawing is a fundamental task for raster graphics. Any acceleration of the traditional line drawing process is of great significance. In this paper, a new line drawing algorithm is presented. The algorithm differs from the traditional line drawing algorithm in that a line is directly represented as a series of displacement codes consisting of 0 s and 1 s, which can be easily determined from a formula. Based upon periodic characteristics of the codes, a new adaptive multi-pixel line drawing algorithm is put forward.Both exerimental results and analysis show this algorithm greatly reduces the computation of line drawing and accelrates the line drawing process.

Abstract:Developing on opponent model, recognizing intentions and plans are necessary for effective communication in MAS. BDI is an effective way to describe mental state, which can be used to develop a generic opponent mental state model. As current research on the opponent mode focus on implications from unilateral action, they neither express simultaneous intentions and other relations, nor provide support to making decisions making at real time. A generic opponent agent model based on BDI developed in this paper improves the research done by AnandS.Rao and Milind Tamde.

Abstract:In this paper a unified framework for representing, reasoning and learning of uncertain information is put forward. The 4-valued recognition structure is used to measure the uncertain degree of uncertain information and thus more powerful expressive capacity is attained. Moreover, the framework supports the efficient acquisition of the uncertain information. These features make the framework more practical than existing theories of reasoning about uncertainty.

Abstract:In this paper, an edge contract based multiresolution modeling algorithm is presented, which generates the continuous simplified models by iteratively removing edges. A weighted control function is used to control the order of edge contract operation. The algorithm adjusts the weights of the control function in different simplification phase to realize different simplification criteria, which lead to the edge that is less important for appearance preservation is removed earlier among all the edge candidates.Furthermore,the independent setis used toavoid local excess simplification.Interpolation stencils are used to compute the generated point in each simplification operation.The experimenral results shoe the algorithm can anhieve satisfactory result.

Abstract:After analyzing the feature of open surfaces, simple close surfaces and multiple connected close surfaces, a 3D triangulation algorithm which doesn t need partition the surface is proposed in this paper. In the 3D triangulation process, the triangular net spreads from the boundary of triangulated field to untreated field, and finally covers the whole surface. This algorithm can apply to the scattered data points on open surfaces, simple close surfaces and multiple connected close surfaces. The experimental results show that this algorithm can satisfy foundation for the surface reconstruction in the next step.

Abstract:The traditional layered scalabilities can not satisfy the requirements of various multimedia applications commendably. The codecs need to provide rate scalability. In this paper, a high performance embedded bit stream is produced by adopting an improved version of zerotree wavelet coding algorithm and a rate scalable video codec is implemented. However, in the rate scalable coding scheme, a single bit stream needs to keep coding effectiveness in a wider range of bit rates. The performance of the existing solutions can not improve the qualities of the reconstructed frames effectively when the target bit rate is increased.It is pointed out that besides error propagation, the rate scalable algorithm needs to give attention to the frame dependency,then a new algorithm is proposed.The experimental results sgow thow that the proposed algorithm can make the single bit srteam produce better qualities at various data rates.This algorithm can also be used in other rate scalable coding schemes that adopt motion compensation to reduce the temporal redundancy.

Abstract:In this paper, a new E-Chunk based multi-engine machine translation model is proposed. The model is composed of a head-driven lexicalized parser, a word-similarity based E-Chunk match engine and a bilingual E-Chunk based transfer engine. The optimal E-Chunk tiling is constructed in a bottom-up style efficiently. Preliminary experimental results show that it is effective in domain oriented machine translation.

Abstract:Progressive meshes algorithm presented recently is used to solve key problems in Virtual Endoscopy in this paper. The advantages of the data structure, data simplification and navigation of virtual endoscopy are introduced. The speciality of the application of the progressive meshes algorithm to virtual endoscopy is analyzed,and some limits are pointed out also. An improved progressive meshes algorithm is presented and realized, and some satisfactory results are obtained. This improved algorithm can solve the data simplifcation and LoD based navion better than before.

Abstract:A new plan formalism method and a plan knowledge graph concept. Compared with Kautz s formalism used widely in plan recognition, this method is simpler and more direct. Because a concept of supporting degree is introduced into plan knowledge graph, the recognized plans are changed reasonably as new evidences collected. A plan recognition algorithm based on knowledge graph presented in this paper can change the plan recognition problem into the graph searching one. This method not only prompts efficiency but also gives the same result as Kautz's.

Abstract:Task scheduling is one of the crucial factors influencing the efficiency of a network of workstations. Fork-Join task graphs can represent many parallel structures. All of the existing algorithms which schedule Fork-Join task graphs have ignored the problem that communications cannot be executed in parallel in non-fully-connected NOW, and some of algorithms with high efficiency even did not take the problem of how to save the processors into account. In this paper, a new static task scheduling algorithm called TSA_FJ is proposed,which is based on task duplication and trying to synthetically take the problems of schedule length shortening,unparallel communications and processor saving into account.By randomly generating the task execution time and communication time,several fork-Join task graphs are got and the scheduling results of TSA_FJ are compared with that of other algorithms for the generated task graphs.It shows that TSA_FJ algorithm has the shortest scheduling length and uses much less processors.It is much suitable to non-fully-connected NOW.

Abstract:UML is a powerful language for object-oriented modeling, but it has some explicit drawbacks. For example, it lacks a model check mechanism so it is difficult to keep the consistence of Meta Model. Moreover, the code building of UML is so scant in spite of the strong power on the OOP analysis and design steps. All of these bring lots of inconvenience in the system building. In this paper, the UML modeling environment is extended by using object-process diagram (OPD), through which people can set a uniform standard for multi-modeling and make the extended UML has the ability of both OOP analyusis and design and visible program design.

Abstract:The automatic optimization of enterprise process model parameters is an optimization problem of the objectives and parameters system. This problem can be changed into the issue solving only one objective and parameters the system optimization with by linearity weighting method. Based on it, an automatic optimization technique in enterprise process modeling based on computer simulation is discussed in this paper, which combines FR (fletcher-reeves) method and TS (tabu-search) algorithm. The technique uses FR method to obtain the local optimization solutions,and uses TS algorithm to discover the global pptimum in the sense of probability.The concept of tabu-area table is firstly proposed,and the rabu-list is improved from the the two dimensions array to one dimension array,which uses to record the trace of optimize of optimize of the process model.FR method is improved by introducing search direction determining while near to the local optimization solutions so as to solve the peak and aigzag curve problem.The new technique can be generalized for model parameters optimization in the arbitrary curves,and be instructive for enterprise to implement BPR(business process reengineering)and ERP(enterprise resource planning)management.

Abstract:A new kind of support vector machine kernel is presented in this paper, which is called coordinate-transform kernel. The character of this kind of kernel is the mapped space has the same dimension as input space. Theoretic analysis and simulated experiment show that this kind of kernel is of accuracy and validity.

Abstract:FISH system is a new generation distributed object database system to support advanced applications like GIS, EC, CMS. In this paper, the design idea and the architecture of FISH system that adopts many novel techniques such as DSVM (distributed shared virtual memory), Persistent heap, Paged-Object, Transparent locking, Compact commit, and focuses on the implementation issues on Windows NT platform, including memory mapping management, shared memory management, RFC, multi-threads scheduling, page-fault handling are presented.The benchmaking testing based on 007 is made and the performance analysis is presented.The results show that the FISH on Windows NT cluster is as effective as that on Unix.

Abstract:In order to solve two difficult problems of premature convergence and slow searching speed of genetic algorithms in evolution neural network, a heuristic mutation operator is presented in this paper. Adaptive probability of mutation and heuristic mutation points selected is applied in it. When no evolution appears after many generations, the range of search will be extended by increasing probability of mutation, and a fine search will be started. The experiments of XOR problem demonstrate that the operator has fine adility of speedy convergence and maintains the diversity of the population automatically.

Abstract:Anaphoric resolution is a hot field in computational linguistics. According to the research, it is found that a kind of anaphora, meta-anaphora, which represents the form of sentence, paragraph or discourse (not the subject or object of the sentence), plays an important role in discourse. So meta-anaphoric resolution affects the computer s ability of discourse understanding. This paper focuses on the meta-anaphoric resolution. A new concept called Elementary Event focus (EE_focus) is introduced. And some meta-anaphoric resolition algorithms based on the concept are desinged.The algorithms examined by a large Chinese natural language corpus.The experimental results show that the EE_focus set can do good job in meat-anaphoric resolution.The workenriches the theory of discourse's structural representation.It is important in comfirming the cohesion and coherence of the Chinese discourses.

Abstract:JAPS-II (Java automatic parallelizing system version 2) is a Java parallelizing compiler that exploits and implements intra and interobject parallelism of serial Java programs. Its target architecture is NOW based on distributed memory computer system. In this paper, the infrastructure of JAPS-II and the key techniques to implement JAPS-II which include dataflow analysis for analyzing object parallelism, optimization for improving object parallelism and reducing executing cost, and techniques of class restructure and code generation are introduced.Related experiment results are also included.

Abstract:Traditional conceptual design methods cannot express designers intent directly because of the lacking of 3D interface mechanisms. Conceptual design based on virtual prototype can not only provide a 3D virtual design environment, but also meet the needs of low costs, short cycle and flexibility of modern design. Considering the characteristics of conceptual design and the feature classification of virtual prototype, a novel virtual prototype-based product model V-desModel in conceptual design is presented in this paper.It introduces virtual feature concept in product's view models,and employs expandable 3D entity-constraint graph to describe the constraints between design objects.V-desModel can effectively support the process of cooperative design and concurrent design.By employing the concept of virtual feature and method of virtual feature generation,V-desModel can solve the insufficient information problems in constructing virtual prototype during the conceptual design of product.It can not only expangd the means and ranges of conceptual design,but also be beneficial for the exertion of creativtity during conceptual design.

Abstract:In this paper, a method to construct an integer version of the linear transform is presented so that the integers of input data can be mapped to integers. Based on the introduction of three kinds of basic integer reversible transform, by means of matrix decomposition, the sufficient and essential condition that a linear transform can be rebuilt into integer version and its construction method are proposed. The transform is reversible, so it is suitable to the lossless data processing such as the lossless digital audio or image compressions.

Abstract:While facing the challenge of e-commerce, the enterprises pay more attention to the business process re-engineering (BPR). A reasonable model to describe the business process is the key point to BPR sucess. Such a model should integrate all related business information and should be enacted by a system. Referencing on the WfMC (workflow management coalition) meta-model, a workflow model based on ECA rules and business activity decomposition is presented. The ECA rules imply the execution dependencies among the zctivities.Using rewriting techniques,the ECA(event-condition-action)model is transformed to Trigger-Action model to make the interpreation more efficient.Rewriting the event to the occurrence time of the event makes the event expression have more expressive semantics.Activity decomposition model can support the hierarchical project management.

Abstract:Multi-Dimentional vectors indexing is one of the key technologies in multimedia database systems. This paper focuses on the dynamic indexing structure based on the vector space. An ER-Tree indexing structure is proposed which is efficient for medium and high dimensional vectors. In the ER-Tree, the efficiency of retrieval could be improved by the ways of partition of the vector space with hyper-sphere and R*-Tree s re-insert technology introduced to enhance ER-Tree s capability to represent the data-sets'features.By introducing safe-inserted-node and safe-deleted-node concept into the structure,the performance of tree creation is reenforced.The algorithms are also given for the creation of ER-Tree.The empirical test test results show that(1)the tree creation algorithm proposed is about 10 times faster thanR*-Tree.(2)the effectiveness of retrieval is highly improved.

Abstract:A Java chip operating system (JCOS) is presented in this paper, which is an object-oriented operating system for embedded applications based on Java chip. JCOS is composed of three kinds of components including severs, kernel and transition layers. The model of JCOSis constituted by the abstract characteristics of these components. The specific components inherit above characteristics and specialize their features. By this technology, JCOS is developed and maintained easily, and has better scalability, portability,portability and code reuse ability.Architecture,model and implementation of JCOS are introduced in this paper.At last,JCOS iscompared with som related work.

Abstract:An automatic text-classification algorithm based on sequence is presented in this paper. It utilizes the semantic relevance on two levels: relevance between sentences (subpattern) and between keywords which represent specific meaning (concept node) in one sentence. In this way, each keyword can be combined with dynamic weight. For subpatterns which contain no keywords, Markov model is used to estimate the amplitude of their signals, thereby the feature sequence for the text which needs to be classified is created.In the experiment of classifying Chinese documents,it is BEP value is about 83％.Furthermore,it is easy to implement in actual system.

Abstract:The existing form frame line detection algorithms are either time consuming or with low robustness. Furthermore, all these approaches do not use the constraint information between form frame lines. In this paper, a novel bottom-up form frame line detection algorithm is proposed based on the directional single-connected chain (DSCC). Defined as an array of black pixel run-lengths, DSCC works very well as an image structure element or a vector in this vectorization algorithm. By merging multiple DSCCs under some constraints,people are able to extract the form frame lines automatically yut fast.With the help of the con straints between form frame lines,the robustness of the approach is increased drastically by getting rid of pseudo lines and completing broken lines.Byfiltering DSCCs created by noise and speeding up the merging of DSCCs,the speed of this algorithm is comparable with the well-known projection method.Experimental results show that this algorithm is fast,resistant to moderate serious line break and skew of any angle.

Abstract:Zerotree-Based wavelet coders, EZW and SPIHT, are two excellent coding algorithms currently, but the imperfect point is that they need more bits to express the parent-children relationship. The another coder based on the morphological representation of wavelet data (MRWD) is also successful, but the shortage is that it doesn t use the interband similarity of wavelet coefficients. In this paper, a new wavelet-based coder is developed by taking the advantages of the above coders and based on morphology. The main property of the new coder is that it uses following characteristics of waveler coefficients;(1)within-subband clustering of significant coefficients;(2)cross-subband similarity;(3)decay of magitude of waveletcoefficients acrooss subband.The advantage of the new coder is to utilize the property of intraband clustering and interband similarity to predict the interband clusters based on morphological operator and then to overcome the shortages of the above three.Wxperimental resuts results show that the performance of the new coder is to utilize the property of intraband clustering and interband similarity to predict the interband clusters based on morphological operator and then to overcome the shortages of the above three coders.Experimental results show that the performance of the new coder is superior ti the above coders.In specially,the new coder has the outstanding performance for the images including a large portion portion of texture.For xeample,for Barbara image 512×521，at 0.25b/p, the new coder outperforms EZW,SPIHT and MRWD and MRWD by 1.68dB,and0.59dB.

Abstract:Elicited by the original steoro matching method, considering the specialities of the obstacle detection task, and in order to cope with its difficulties, a new fast and robust obstacle detection method is presented in this paper. Being different from origin methods based on the depth of obstacl object, the presented method uses a newcharacterization-plan s direction. This charaction identifies the vertical surface of obstacle object from thehorizonal surface of ground.

Abstract:QUICKFIT, a quick algorithm is developed by analyzing the properties of the preparatory algorithm of Bound Fit and integrating Bound Fit with MULTIFIT for scheduling independent tasks on identical parallel processors to minimize the make span. For the same tasks and machines, QUICKFIT needs a fewer iterations than MULTIFIT and Bound Fit do to obtain the same make span. Experimental results show that the greater the task-to-machine ratio is, the better the performance of QUICKFIT is than MULTIFIT and Bound Fit do to obtain the same make span.Experimental results show that the greater the task-to-machine ratio is,the better the performance of QUICKFIT is than MULTIFIT and Bound Fit,and the make span of QUICKFIT is the least among those of LPT,MULTIFIT and Bound Fit in most of cases,and QUICKFIT is suitable for the large scalee identical scheduling.

Abstract:A new solution of the shadow and target detection for high-resolution SAR (synthetic aperture radar) images is presented in this paper. At first, shadow detection through the RCS (radar cross section) reconstruction algorithm based on the model of equivalent number of looks is described, and then by nonlinear iterated process, the target detection is achieved. Compared with the CFAR (constant false alarm rate) detector, this algorithm can detect the shadow and more targets of interest.

Abstract:The function of population isolation is stated by the analyzing the domain of attraction of local optimum. The separation among populations and the adaptive-gathering in a population are achieved by local evolution, so the multi-model function optimization is transformed to uni-modal function optimization. Combining with the (μ+1) evolutionary approach based on space ranking that has a swift convergence speed in uni-modal function optimization, the evolutionary algorithm is presented in which populations separate and genther aunction optimization,the evolutionary algorithm is presented in which populations separate and gether automatically according to domanin of attraction,The experimental result is given at the end of the paper.

Abstract:In this paper, in order to solve the problems of poor description ability and flexibility existed in existing workflow management system, a new workflow modeling method based on coordination theory and feedback mechanism is put forward. The proposed method extends the traditional activity network model and introduces the new modeling mechanism. Several new modeling elements, such as request node, service node, coordination node, and multiple instance node are added to the modeling method. The detailed realization mechanism and application scenarios are given in this paper and two examples are presented too.Research results show that compared with traditional actibity network model,the proposed method has significant asvantages in enhancing description ability,reducing model complexity,and improving system flexibility and adaptability.The proposed method has important value in designing and developing workflow management systems with high reliable,flexibility,andadaptability.

Abstract:In this paper, a cost function relative to residual errors is introduced, and a linear algorithm by exploiting the strategy of weighted translation transformation is presented. Firstly, the original input data is weighted and the centroid coordinates are calculated, and the origins of coordinates are translated to their centroids. Then, the matching points are normalized. Finally, the eight parameters of fundamental matrix (F-matrix) can be solved and the procedure of estimating the fundamental matrix with high accuracy can be achieved.Experimental results show that this algorithm performs very well in terms of robustness to outliers and noises.The algorithm is superior to other algorithms in residual errors and average epipolar distance and improves the accuracy of F-matrix.

Abstract:With the increasing of Internet traffic, the RED (random early detection) algorithm depending on the average queue length has inherent flaws. RED cannot efficiently prevent the packet loss even if it combines with explicit congestion notification (ECN). Based on comparison between RED and BLUE, an enhanced active queue management mechanism, EBLUE (enhanced BLUE), is presented. The TCP congestion control with EBLUE is researched, and the token bucket is used to mark priority traffic at the network entry in order to provide the minimum rate guarantees.Finally,under the NS environment,the simulated experiment and performance evaluation are done in packet loss percent and link efficiency for EBLUE by comparing the mechanism with relative ones.The experimental results show that this mechanism can support bandwidth reserbation services effiiently.

Abstract:The KQML (knowledge query manage language) is a representative model in multi-agent interaction. In this paper, some conclusions are given about the necessary communication support to the agent-oriented program on the knowledge level based on the analysis of KQML model. Firstly, the agent statue model and the transform model of KQML communication are founded to analyze the necessary conditions in performing interaction with the synchronal and asynchronous KQML model respectively. Secondly, the deadlock and starvation problems in the communication process with the synchronal and saynchronous KQML model respectively.Secondly,the deadlock and starvation problems in the communication process with KQML model are analyzed feom the KQML stansform aspect,and the cardinal readons and the solutions are given.At last,the advantages and the disadvantages of the synchronal and asynchronous KQML model are listed respectively,and the choosing principle is given.

Abstract:In this paper, the shape variation law of Bezier surfaces, B-spline surfaces and NURBS surfaces (BBN surfaces) is investigated under the modification of a control point of the surfaces. The relationship between the geometric elements of the surface before and after deformation and the deforming displacement is built up by decomposition a BBN surface into the summation of a convex surface and a surface with special shape. The law of a convex BBN surface losing its convexity is also obtained. The achievements in this paper can guide the algorithm design and shape control in CAD modeling.

Abstract:The design and implementation of the remote file browser (RFB) based on Java/CORBA is introduced in this paper, which is developed for Dawning3000 parallel machine. On traditional parallel machines, users organize and manage the file system through command line interface, which makes users feel very inconvenient. To solve this problem, a GUI-based file browser is implemented to support the access to the file system of parallel machines. The RFB employs client/server mode, and separates GUI from the implementation of the access to file system.Here,GUI is implementes with language and Java GUI Swing package,which make it platform-independent,and communiction between client and server is implementes with currently prevalent CORBA middleware.The RFB supports remote access to the file system of parallel machines within both local area network and wide area network,and hides the differnce of client platform and location of paralle machines,which highly improves the usability of parallel machines.