Applied Mechanics and Materials Vols. 713-715

Abstract: Current trends in micro-architecture are towards larger number of processing elements on a single chip. It is challenging to tap the peak performance of those processors. In order to address this issue, the most promising solution is automatic parallelization. This approach does not require programmer too much effort in the process of parallelizing programs. Polyhedral model is a mathematical framework based on the powerful linear integer programming, which provides an abstraction concept to represent the nested loops computing and dependence of data access using integer points in the polyhedron. We propose an automatic transformation framework based on polyhedral model to optimize nested loop with affine dependences for parallelism and locality.

Abstract: Mechanical optimization design is a new design method in the development foundation of the modern mechanical design theory, the application of optimization design in mechanical design can make the scheme achieve some optimization results in the design requirements specified, without consuming too much computational effort. The corresponding mathematical models of ant algorithm and Cellular ant algorithm are established, according to the actual mechanical design problems, and used to solve the established mathematical model by computer, so as to obtains the optimal design scheme.

Abstract: Aiming at the problem that the traditional tracking method cannot track high speed high maneuvering target effectively, one modified fixed structure multiple model algorithm (M-FSMM) and one modified variable structure multiple model (M-VSMM) algorithm were proposed. The Constant Velocity (CV) model, Current Statistical (CS) and Modified Coordinate Turn (MCT) model were adopted in the M-FSMM algorithm, by means of Connected Graph (CG) thinking, the model connected graph was made up by models that can describe possible motion, the connected relation was set up and model self-adapting was designed to carry out the variable structure tracking that can quickly jump between models. Monte Carlo simulation results show that the two methods can track high speed high maneuvering target effectively, the computational quantum of M-FVMM algorithm is larger but the tracking accuracy and stability are better than the M-FSMM algorithm. They can be used to track near space hypersonic targets.

Abstract: This paper presents a model fused approach to Chinese dialect identification by combining multi-information including acoustic, phonotactic and prosodic feature. At first we analyze the way to translate language information into these features, and then propose a model fused framework for back-end classification. The experimental results show that the proposed method improves identification performance greatly and the prosodic features are more effective for shorter speech.

Abstract: At present, many popular methods for object recognition are based on regional visual feature vectors which ignore the global structure of the object, leading to the “semantic gap” between image content and its tag. Graph-based structural representation of an object can record regional visual features and global relationships between regions that compose the object. However, the computation complexity of graph comparing limits the application of graph model in object recognition. One method for improving the recognizing speed is to reduce the prototypes in concept space of specific object category. We presented a prototype selection method for sythetic object recognition based on structural graph model using greedy clustering and representative sample selection. Experiments are conducted on a 2-D CAD synthetic object database. And the results show that the prototype selection method reduces the time cost of object recognition greatly with an acceptable loss in accuracy.

Abstract: Parallel testing is the key to achieving parallel test task scheduling, and its core is allocating resources fairly and reasonably to the test tasks, then rearrange the execution order of the test tasks in meeting the priority relationship between the resource constraints and the conditions of the test tasks, making the whole test mission of this system can be completed in the shortest possible time and improving the test efficiency. The basic ant colony algorithm has been improved in this paper to fit the parallel test task scheduling and to obtain the task scheduling sequence that complete all testing tasks in shortest test time.

Abstract: In order to solve the problem such as measuring errors and faults in the color detection because of fatigue and individual differences, and to solve the problems of poor relevant of equipment and weak data-processing ability on the color detection. The solution of images color difference recognition system based on Labview is proposed. The solution is establishing the system simulation platform on Labview software.

Abstract: Aiming at detecting sea targets efficiently, an approach using optical remote sensing data based on co-training model is proposed. Firstly, using size, texture, shape, moment invariants features and ratio codes, feature extraction is realized. Secondly, based on rough set theory, the common discernibility degree is used to select valid recognition features automatically. Finally, a co-training model for classification is introduced. Firstly, two diverse ruducts are generated, and then the model employs them to train two base classifiers on labeled dada, and makes two base classifiers teach each other on unlabeled data to boot their performance iteratively. Experimental results show the proposed approach can get better performance than K-Nearest Neighbor (KNN), Support Vector Machines (SVM), traditional hierarchical discriminant regression (HDR).

Abstract: Aiming at some deficiencies of existing network intrusion detection system, the paper proposes a network intrusion detection system model based on data mining, applying data mining technology to network intrusion detection, and constructed the final test results of the system on the basis of Snort design. Experimental results demonstrate that this data mining based on cluster algorithm can effectively establish models of network normal activity and significantly accelerate intrusion detection, whilst its association analyzer can effectively unearth some new intrusion patterns from abnormal logs, and automatically construct intrusion detection rules.

Abstract: The evolution of elliptical hollow Gaussian beams propagating in strongly nonlocal nonlinear media is investigated. As examples, this paper mainly focuses on the evolutions of the transverse intensity and the beam width of elliptical hollow Gaussian beams with the beam order being 1 and 3. The results show that the evolutions of the transverse intensity and the beam width are both periodical and the size of the evolution period is determined by the input power. There exists the transverse reverse transform for the beam width when selecting a proper input power, which is quiet different from the circularly symmetric hollow Gaussian beams.