Journal of Computer and Robotics
Volume:9 Issue: 2, Summer and Autumn 2016

Image segmentation is an important task in image processing and computer vision which attract many researchers attention. There are a couple of information sets pixels in an image: statistical and structural information which refer to the feature value of pixel data and local correlation of pixel data, respectively. Markov random field (MRF) is a tool for modeling statistical and structural information at the same time. Fuzzy Markov random field (FMRF) is a MRF in fuzzy space which handles fuzziness and randomness of data simultaneously. This paper propose a new method called FMRF-C which is model clustering using FMRF and applying it in application of image segmentation. Due to the similarity of FMRF model structure and image neighbourhood structure, exploiting FMRF in image segmentation makes results in acceptable levels. One of the important tools is Cellular learning automata (CLA) for suitable initial labelling of FMRF. The reason for choosing this tool is the similarity of CLA to FMRF and image structure. We compared the proposed method with several approaches such as Kmeans, FCM, and MRF and results demonstratably show the good performance of our method in terms of tanimoto, mean square error and energy minimization metrics.

Different approaches have been proposed for feature selection to obtain suitable features subset among all features. These methods search feature space for feature subsets which satisfies some criteria or optimizes several objective functions. The objective functions are divided into two main groups: filter and wrapper methods.  In filter methods, features subsets are selected due to some measures like inter-class distance, features statistical independence or information theoretic measures. Even though, wrapper methods use a classifier to evaluate features subsets by their predictive accuracy (on test data) by statistical resampling or cross-validation. Filter methods usually use only one measure for feature selection that does not necessarily produce the best result. In this paper, we proposed to use the classification error measures besides to filter measures where our classifier is support vector machine (SVM). To this end, we use multi objective genetic algorithm. In this way, one of our feature selection measure is SVM classification error. Another measure is selected between mutual information and Laplacian criteria which indicates informative content and structure preserving property of features, respectively. The evaluation results on the UCI datasets show the efficiency of this method.

Clustering is one of the known techniques in the field of data mining where data with similar properties is within the set of categories. K-means algorithm is one the simplest clustering algorithms which have disadvantages sensitive to initial values of the clusters and converging to the local optimum. In recent years, several algorithms are provided based on evolutionary algorithms for clustering, but unfortunately they have shown disappointing behavior. In this study, a shuffled frog leaping algorithm (LSFLA) is proposed for clustering, where the concept of mixing and chaos is used to raise the accuracy of the algorithm. Because the use of concept of entropy in the fitness functions, we are able to raise the efficiency of the algorithm for clustering. To perform the test, the four sets of real data are used which have been compared with the algorithms K-menas, GA, PSO, CPSO. The results show better performance of this method in the clustering.

Residue Number System (RNS) is a carry-free and non-weighed integer system. In this paper an improved three-moduli set  in reverse converter based on CRT algorithm is proposed. CRT algorithm can perform a better delay and hardware implementation in modules via other algorithms. This moduli is based on p that covers a wide range on modules and supports the whole range of its modules in dynamic range. With growth in moduli, many types of modules have been proposed. By using dynamic range we can solve many problems in Residue Number System (RNS) by just one three moduli. In proposed moduli set of this paper in Residue Number System (RNS), the internal circuit is improved and thus, complexity of circuit, energy consumption and power consumption in our proposed design is improved. These improvements are shown in evaluation in terms of CSA adders, CPA adders and delay.

Many appealing multicast services such as on-demand TV, teleconference, online games and etc. can benefit from high available bandwidth in multi-radio multi-channel wireless mesh networks. When multiple simultaneous transmissions use a similar channel to transmit data packets, network performance degrades to a large extant. Designing a good multicast tree to route data packets could enhance the performance of the multicast services in such networks. In this paper we want to address the problem of multicast routing in multi-radio multi-channel wireless mesh networks aiming at minimizing intermediate nodes. It is assumed that channel assignment is known at prior and channels are assigned to the links in advance. Aiming at constructing multicast tree with minimum number of intermediate nodes and minimum number of interfered nodes we propose a heuristic algorithm called Maximum Multicast Group Nodes (MMGN). Simulation results demonstrated that our proposed method outperforms LC-MRMC algorithm in terms of throughput and packet delivery ratio.

In this paper, we propose to optimize manufacturing methods of memory cells by produced silicon nanoparticles via electrical spark discharge of silicon electrodes in water to reduce the energy consumption for low power applications. The pulsed spark discharge with the peak current of 60 A and a duration of a single discharge pulse of 60 µs was used in our experiment. The structure, morphology, and average size of the resulting nanoparticles were characterized by means of X-Ray Diffraction (XRD), Raman spectroscopy and transmission electron microscopy (TEM). TEM images illustrated nearly spherical and isolated Si nanoparticles with diameters in the 3-8 nm range. The optical absorption spectrum of the nanoparticles was measured in the violet-visible (UV-V is) spectral region. By measuring of the band gap we could estimate the average size of the prepared particles. The silicon nanoparticles synthesized exhibited a photoluminescence (PL) band in the violet- blue region with a double peak at around 417 and 439 nm. It can be attributed to oxide-related defects on the surface of silicon nanoparticles, which can act as the radiative centers for the electron-holepairsre combination.