Applied Mechanics and Materials Vols. 263-266

Abstract: Smooth support vector machine (SSVM) is a strong convex, smooth, and unconstrained optimization model transformed from traditional support vector machine (SVM). In this study, a new smooth function which is a rational function diffrent from those polynomial functions , is used to smoothen the model of support vector machine. A new SSVM based on this rational function (Rational-SSVM) is obtained. Furthermore, Rational-SSVM is better than those SSVMs based on polynomial functions by the precision analysis.

Abstract: In this paper, linking with the basic principle of FCM (Fuzzy c-means clustering) algorithm, on the basis of theory research, a method of the cluster analysis of FCM based on sober extraction algorithm is proposed. To insure the quality of image reconstruction and the edge information extraction, the characters of sober operator is analyzed. Firstly, the approximate optimal solution obtained by the improved FCM algorithm is taken as the original value, then combined with intensity-texture-position feature space in order to produce connected regions shown in the image. The final segmentation result is achieved at last. The experiment results prove that in the view of the image segmentation, this segmentation algorithm based on sober extraction algorithm provides fast segmentation with high perceptual segmentation quality.

Abstract: In this paper, linking with the basic principle of FCM (Fuzzy c-means clustering) algorithm, on the basis of theory research, a method of the cluster analysis of FCM based on artificial life is proposed. The artificial life approach is promising in image processing because it is inherently parallel and coincides with the self-governing biological process. Firstly, the approximate optimal solution obtained by the FCM algorithm is taken as the original value, then combined with intensity-texture-position feature space in order to produce connected regions shown in the image, the final segmentation result is achieved at last. The experiment results prove that in the view of the sport image segmentation, this algorithm provides fast segmentation with high perceptual segmentation quality.

Abstract: Moving target detection and tracking algorithm as the core issue of computer vision and human-computer interaction is the first step of intelligent video surveillance system. Through comparing temporal difference method and background subtraction, a moving object detection and tracking algorithm based on background subtraction under static background is proposed, in order to quickly and accurately detect and identify the moving object in the intelligent monitoring system. In this algorithm, firstly, we use background acquisition method to receive the background image, then use the current frame image and the received background image to perform background subtraction in order to extract foreground object information and receive the difference image; secondly, we use threshold segmentation and morphology image processing to process the difference image in order to eliminate noises and receive the clear binary moving object image; finally, we use the centroid tracking method to track and mark the moving object. Experimental results show that the algorithm can effectively and quickly detect and track moving object from video sequence under static background. This algorithm is easily realized and has good real-time and robust, which is automated and self triggered for background updating. The algorithm can be used in driver assistance systems, motion capture, virtual reality and other fields.

Abstract: In this work a multi-objective Simulated Annealing algorithm (SA) is applied to maximize the small signal gain (SSG) of Er3+-3µm: YLF crystal, which plays important role in the laser systems development (and further operation). For this purpose the rate equations Er:YLF crystal were solved numerically and 3 microns SSG for different Er3+ concentrations was maximized by Simulated annealing (SA) technique taken into account the pulse time (time-ON), time-OFF and the pumping rate (Rp) parameters. Results show that is possible to obtain valid laser gain for erbium concentration smaller than 9 mol % for the Er:YLF system after appropriate set of these parameters.

Abstract: Discretization of attributes with real values is an important problem in data mining. From the viewpoint of discernibility, consistency needs to be satisfied and cuts set for discretization is expected to be as small as possible. Consistent and minimal discretization problem is NP-complete. A particle swarm algorithm for discretization of decision tables is proposed.

Abstract: In order to obtain the optimal technological conditions of preparing biodiesel, artificial neural network was used to study the biodiesel processing model on transesterification method based on the single factor experiment and orthogonal experiment. The results of experiment indicated that we used the back propagation BP algorithm of artificial neural network to set the network prediction model based on the orthogonal test data can forecast the biodiesel conversion rate under different reaction conditions more accurately.The optimal conditions were obtained from this network model as follows: Molar ratio of methanol to oil was 6:1, the catalyst was 1.0% (w/w, based on oil), reaction temperature and reaction time was 65°Cand 2.5h respectively. Under the optimal conditions, the conversion rate of prediction was 94.93%, the conversion rate of experiment was 95.42% and the relative error was 0.51% compared with the predicted value. Therefore, the network k model could reflect inherent law of sample.

Abstract: As the power consumption becomes a hot topic in the field of NOC, many researchers are studying the design technology that minimizes the power consumption of NOC. Some articles realize the NoC mapping by ant colony optimization algorithm. However, by these methods, the system consumes much more CPU time when the scale of the problem or the amount of ants increases. In this article, by GPU acceleration, the scale of the problem to be solved can be increased linearly with the increasing of ants.

Abstract: Clustering is one of the prominent classes in the mining data streams. Among various clustering algorithms that have been developed, density-based method has the ability to discover arbitrary shape clusters, and to detect the outliers. Recently, various algorithms adopted density-based methods for clustering data streams. In this paper, we look into three remarkable algorithms in two groups of micro-clustering and grid-based including DenStream, D-Stream, and MR-Stream. We compare the algorithms based on evaluating algorithm performance and clustering quality metrics.

Abstract: A common form of variable step size LMS algorithm is presented，which is derived by the extensive analysis about several variable step size LMS algorithm. Using genetic algorithm parameter optimization, the algorithm get the optimal valueα、β、m and h quickly and efficiently，and not rely on experience or method of trial and error. MATLAB simulation results confirmed the theoretical analysis, the algorithm took on good convergence and tracking properties，and could be widely used in modern digital communication systems.