JOURNAL OF SOFT COMPUTING AND INFORMATION TECHNOLOGY (JSCIT)
Year:2016 | Volume:6 | Issue:1 |Papers Count:7

Emerging 3D technology partitions a larger die into smaller parts and then stacks them in a 3D fashion. This technology can lead to a paradigm shift in on-chip communication design providing higher orders of bandwidth and lower latency. However, due to the aggressively scaled transistors in modern technology nodes, the reliability issue has become into a major concern. In this paper,we leverage these ultra-low-latency vertical links to design a fault-tolerant 3D NoC architecture. In this architecture, permanent and intermittent defects on links and crossbars are bypassed by borrowing the idle bandwidth from vertically adjacent links and crossbars. Evaluation results under synthetic and realistic workloads show that the proposed fault-tolerance mechanism offers higher reliability and lower performance loss, when compared with state-of-the-art fault-tolerant 3D NoC designs.

In the engineering systems, one of the most important measurable parameters is temperature. For measurement of temperature as for sensitivity, accuracy and term of physical required, different sensors as RTD, NTC, PTC, Thermo-couple are exist. One of the most important sensors is NTC that despite the many benefits, less than it used to be. Nonlinearity characteristic and self-heating phenomenon in the NTC are the biggest deterrents in practical applications. In this study, a new structure based on application of artificial neural network (ANN) has been suggested for modeling of the nonlinear and the self-heating phenomenons of NTC. Simulation results based on read generated laboratory data are presented to demonstrate the performance of the proposed structure. Mean square error of 0.0480 and 0.0370 are achieved based on MLP and RBF neural network, respectively.

Forest canopy density classification map is one of the main sources of information used in forest management. In conventional methods multispectral images are used to generate the map. In this study, aerial panchromatic images as a valuable data source are used to generate this map. Statistical image texture quantization methods including first statistical and second statistical based on GLCM matrix and also geostatistical method used to generate new features from high spatial resolution image. Generated features beside main image used as classification input feature space. Supervised classification was used and about 90% accuracy was obtained. This method is mainly usable in areas with low variety in forest cover type like Zagros and Iran-Turanian region.

In this comparative study, three global DEMs, namely, SRTM90, SRTM30 and ASTER have been compared to a national DEM, derived from 1:25000 maps by NCC of Iran. The goal of this study is to assess the accuracy of different DEMs, and lastly show which of them matches both NCC and reality more closely. To achieve these goals, a reference DEM has been derived from 1:2000 maps of the areas in question. Experimental results on two different areas showed that NCC DEM outperforms than the other DEMs. As is obvious, SRTM30 matches both NCC DEM and reality more closely. This relationship is true for slope products except for SRTM90, which is sensitive to the topography. This product has minimal correlation with reference DEM, derived from 1:2000 maps. Generally speaking, RMSE values of slope data are similar for all four DEMs in the two case studies, implying that slope data are more robust than the elevation data.

In this paper, the fuzzy proportional- integral controller (PI) optimized by particle swarm algorithm is applied to control the frequency of island Micro-grids. Micro-grids are distributed energy sources that usually use renewable energies in order to produce and transmit electrical power to distributed loads in both connected and islanded modes. Because of natural variations of power that produced by renewable energy sources and uncertainties of power systems, classic controllers do not have a good performance. So nominal values of PI parameters and interval of fuzzy membership functions are optimized using PSO algorithm. Fuzzy system updates PI parameters momently. Simulations show the better performance of proposed controller in terms of RMS, overshoot and undershoot, frequency of oscillations and settling time in facing different load disturbances in comparison to classic PI controller, fuzzy PI controller and another PSO fuzzy controller. Results indicate the robust performance of the proposed controller in dealing with variation of system parameters.

In this paper a method for online recognition of Farsi handwritten digit is presented. Four sets of Point Features and a set of global features are extracted from preprocessed patterns. In this study a suitable structure for feature vector, which contains only a set of point features and global features, to improve the performance of classifier, is presented. Therefore, numerous experiments with each of the point feature set and the global features using support vector machine (SVM) classifier, with one versus all (OVA) and one versus one (OVO) approaches is done. In this paper, for presenting a fast, accurate and reliable method, SVM classifier with OVO approach is proposed for online recognition of Farsi handwritten digits. This method is applied on online-TMU database. The best recognition rate with point feature set (Dx , Dy)s and global features is achieved. The average recognition rate is 98.08%.

Colorization is the process of color allocation into grayscale images or films. Since several colors are the same level of illumination, the colorization of a grayscale image into a colorized one has not a unique and automatic solution, and the human has a cardinal role in the colorization process. In other hand, in a grayscale image, the information of the neighbor pixels of a pixel are effective in its color. In this paper, a novel idea has been proposed for colorization of a grayscale image based on image segmentation, with minimum human interference. The proposed method use a collection of reference images consist of some classes of natural color images. Some example of classes are tree, mountain, jungle, sea, human flower and etc. As soon as the user import a test grayscale image into the algorithm, its type (class) is selected by the user. The user can select more than one class for a test image. The reference images of the selected class are used as the reference images in the colorization process. On the other hand, the test image is segmented, and for each segment, the most similar segment in the set of the segments of the selected reference images is specified. The segment of the grayscale test image is colorized using the fuzzy theory based on the specified segment of the reference image. This process is done for all the segments of the test image. Finally, a post process is applied to match the color of the neighbor pixels. The minimum human interference and the use of the information of the neighbor pixels are the most important advantages of the propose method.