Search

Enlarging the spatial resolution of features generated by fully convolutional networks (FCNs) can improve the performance of semantic segmentation. To achieve this goal, deeper network with deconvolutional structure can be applied. However, when the network architecture becomes more complex, the training efficiency may degrade. To address the joint optimization problem of improving spatial resolution through deeper networks and training deeper networks more effectively, we propose a Fully Convolutional Residual Continuous CRF Network (FC-RCCN) for semantic segmentation. FC-RCCN is composed of three subnetworks: a unary network, a pairwise network, and a superpixel based continuous conditional random filed (C-CRF) network. In order to generate full spatial resolution predictions with high-quality, a residual block based unary network with multi-scale features fusion is proposed. Even though the unary network is a deeper network, the whole framework can be trained effectively in an end-to-end way using the joint pixel-level and superpixel-level supervised learning strategy which is optimized by a pixel-level softmax cross entropy loss and a superpixel-level log-likelihood loss. Besides, C-CRF inference is fused with pixel-level prediction during the test procedure, which guarantees the method’s robustness to the superpxiel errors. In the experiments, we evaluatee the power of the three subnetworks and the learning strategy comprehensively. Experiments on three benchmark datasets demonstrate that the proposed FC-RCCN outperforms previous segmentation methods and obtains the state-of-the-art performance.

In recent years, the objectification of tongue image in Traditional Chinese Medicine (TCM) has become a topic of extensive discussion. As an important reference index in TCM diagnosis, tongue features play an important role in disease prevention and diagnosis. Among them, how to achieve the standardization and accurate segmentation of the tongue image and extracting quantitative feature information has become a key issue that restricts objective development of TCM tongue diagnosis. This article mainly reviews two aspects of tongue color standardization processing and tongue image segmentation methods. Review the currently widely used methods, including threshold segmentation, region segmentation, edge segmentation, and deep learning methods. The development and current situation of tongue image segmentation technology are explained, and the current problems are summarized. Put forward reasonable prospects and prospects for the development of this research field.

This paper presents an improved genetic algorithm using novel crossover and mutation (IGA-NCM) to solve the combined heat and power economic dispatch (CHPED) problems. The basic genetic algorithm (GA) has been augmented in three aspects. First, the selection operation is excluded from GA in order to avoid excessive losses of population diversity. Second, two kinds of adaptive crossover operations are used to sufficiently excavate the information of parents and yield potential offsprings. Third, a novel mutation operation is used to replace a few genes of each crossed offspring by those of the other crossed offsprings’ parents, which can further improve their quality. Furthermore, a new constraint handling method is proposed to repair the mutated offsprings and enable them to enter feasible regions easily. Experimental results show that our proposed IGA-NCM algorithm outperforms the other ones according to computation accuracy and runtime. Therefore, it is a potential alternative for the CHPED problems with or without prohibited operating zones.

An R2 indicator-based multi-objective particle swarm optimiser (R2-MOPSO) can obtain well-convergence and well-distributed solutions while solving two and three objectives optimisation problems. Ho...

In this paper, we propose a memory-based global differential evolution (MGDE) algorithm and a repair technique of constraint handling for the dynamic economic dispatch problems. On the one hand, MGDE modifies the mutation of DE/best/1, and uses a memory pool to provide more candidate solutions for this operation. Moreover, it adopts a randomly generated scale factor in the modified mutation to enhance its exploration capacity. In the crossover, a dynamical crossover rate is introduced to balance MGDE's global and local search capacities. On the other hand, a repair technique is designed for handling three kinds of constraints associated with generator capacity, power balance and generating unit ramp-rate. Moreover, a commonly used penalty function method is subsequently employed to handle the possible constraint violations associated with power balance and prohibited operation zones (POZs). To judge the performance of MGDE and the efficiency of the repair technique, we have solved six well-known DED problems taken from different sources. According to the experimental results, MGDE shows a superior performance in comparison with other improved DEs which also solve these problems. In the mean time, the repair technique of constraint handling has a high efficiency in eliminating or reducing the constraint violations.

Display OmittedThe improvisation process of LHS algorithm. Opposition-based learning (OBL) technique is employed in improvisation process. The purpose is to increase the diversity of solution.The current best harmony and worst harmony are used to adjust the parameter BW. An adaptive global pitch adjustment is designed to enhance the exploitation ability of solution space.In the proposed algorithm, a new harmony and its opposite harmony are generated in iteration. Then a competition selection mechanism is established to improve solution precision.The effects that varying the parameter HMS and HMCR have on the performance of the LHS algorithm is also analyzed in detail. In this paper, we propose an improved harmony search algorithm named LHS with three key features: (i) adaptive global pitch adjustment is designed to enhance the exploitation ability of solution space; (ii) opposition-based learning technique is blended to increase the diversity of solution; (iii) competition selection mechanism is established to improve solution precision and enhance the ability of escaping local optima. The performance of the LHS algorithm with respect to harmony memory size (HMS) and harmony memory considering rate (HMCR) are also analyzed in detail. To further evaluate the performance of the proposed LHS algorithm, comparison with ten state-of-the-art harmony search variants over a large number of benchmark functions with different characteristics is carried out. The numerical results confirm the superiority of the proposed LHS algorithm in terms of accuracy, convergence speed and robustness.

Display Omitted A new population partitioning strategy is employed to PSO algorithm.In the current population, global neighborhood exploration strategy is presented to enhance the global exploration capability.A local learning mechanism is used to improve local exploitation ability in historical best population.Stochastic learning and opposition based learning operations are employed to accelerate convergence speed and improve optimization accuracy in global best population.IGPSO performs better for engineering design optimization problems. In this paper, an improved global-best-guided particle swarm optimization with learning operation (IGPSO) is proposed for solving global optimization problems. The particle population is divided into current population, historical best population and global best population, and each population is assigned a corresponding searching strategy. For the current population, the global neighborhood exploration strategy is employed to enhance the global exploration capability. A local learning mechanism is used to improve local exploitation ability in the historical best population. Furthermore, stochastic learning and opposition based learning operations are employed to the global best population for accelerating convergence speed and improving optimization accuracy. The effects of the relevant parameters on the performance of IGPSO are assessed. Numerical experiments on some well-known benchmark test functions reveal that IGPSO algorithm outperforms other state-of-the-art intelligent algorithms in terms of accuracy, convergence speed, and nonparametric statistical significance. Moreover, IGPSO performs better for engineering design optimization problems.

The nuclear support, which is used to support pipe or equipment with nuclear safety level, is of great importance for nuclear power plants. In this work, we studied that the door type nuclear support was transformed to disassemble easily by installing bolts. The 3D models of nuclear support before and after transformation were established in ANSYS Workbench software. With the reasonable setting of the contact conditions and boundary, the deformation, stress and natural frequency were calculated in ANSYS Workbench by using the finite element method. The results show that bolted connections will lead to the loss of stiffness of support and natural frequency decrease. The order natural frequency of the support decreases with addition of the thickness of plates and the bolt circle diameter has little effect on global stiffness of the support. The finite element analysis results provide reference for the transformation design and optimization of nuclear support and have practical engineering value.

In order to solve the problem of excessive vibration during the operation of the refrigeration unit in nuclear power plant, taking the double motor refrigeration unit in operation as the research object, we studied the effective measures to reduce the vibration and improve the stiffness of the double motors in the near resonance region. On the basis of theoretical calculation, the weaknesses of the equipment stiffness were reinforced and reconstructed, and the secondary analysis and calculation of the stiffness and seismic performance of the reformed refrigeration unit were carried out to determine the feasibility of the optimization scheme. Based on the analysis, we optimized the equipment. After that, there was a significant decrease in the vibration of the refrigeration unit. The vibration measurement results indicated that the vibration value of the unit has dropped to the qualified range, effectively avoiding the equipment damage caused by excessive vibration. We provided a complete set of analysis ideas for the research and solution of vibration problems of rotating equipment in nuclear power plant.

Tooth-marked tongue is one of the important tongue features in Traditional Chinese Medicine tongue diagnosis. It reflects the dysfunction of liver, spleen and kidney, and contains rich pathological information. However, the recognition of tooth-marked tongue is challenging. Most existing methods only focus on the classification of tooth-marked tongue, and the exact location and number of teeth marks were not involved, which had no great practical significance for the subsequent treatment based on syndrome differentiation. In this paper, we try to solve these problems by proposing a method based on Mask Scoring R-CNN framework and transfer learning which can extract, visualize the teeth marks and identify the number of teeth marks. First, the tongue image is fed into residual network of depth 101 layers (ResNet-101) to generate regions of interest (RoIs) via region proposal network (RPN) and RoI feature via RoIAlign. Next, MaskIOU is then predicted using the predicted mask and RoI characteristics as inputs. The proposed approach achieved F1 score of 0.95. According to these experimental results, the approach can robustly detect and segment teeth marks, which can provide a basis for the severity analysis of tooth-marked tongue.

The traditional pulse coupled neural network model only considers the spatial position information between neurons, but neglects sensitivity of human visual perception to luminance change, especially when the brightness is high or low, the visual sensitivity of the human eye is relatively low, and in a certain range of brightness, the visual sensitivity of the human eye is relatively high, so it is easy to have false segmentation. In order to solve this problem of the traditional pulse coupled neural network model, this paper redescribes the connection weights between neurons by using the human eye luminance weight factor, and obtains the new pulse coupled neural network model based on visual perception. Through the simulation experiment of life images, experimental results show that the proposed algorithm is better than the traditional pulse coupled neural network model.

In order to enhance the performance of harmony search algorithm, a competition harmony search algorithm is presented in this paper. The proposed algorithm has three modifications. Firstly, a parallel search mechanism is designed in the improvisation process, which is aim to accelerate searching speed. second, a novel dimension selection strategy is presented for improving searching efficiency. Moreover, estimation of distribution operation is used based on history information. Besides competition updating operation is used to replace the original updating strategy. Experiment results demonstrated that the proposed algorithm has better performance than the state-of-the-art HS algorithms.

A memory-based simulated annealing (MSA) algorithm is proposed for the fixed-outline floorplanning with soft blocks. MSA constructs a memory pool to store some historical best solutions. Moreover, it adopts a real-time monitoring strategy to check whether a solution has been trapped in a local optimum. In case a solution encounters this predicament, it will be replaced by the one from the memory pool, and the current temperature will be regenerated by continuously perturbing the new solution several times. To meet the fixed-outline requirements, a new auxiliary function is formulated based on the geometric structure of the current floorplan, and it is very helpful in driving MSA to search towards potential solution space. Concretely, the area information of all violated blocks is utilized to construct an auxiliary function. Moreover, the excessive area of a violated block can be weighted by three different coefficients, which depend on the relative position of the block and the fixed-outline. Additionally, due to its simple topology and strong applicability, B $$^{\star }$$ -tree representation is employed to perturb a solution in each generation. The efficiency of the proposed method is demonstrated on six GSRC floorplan benchmark examples with various white space and aspect ratios. Two groups of Matlab simulations show that our approach can achieve better floorplanning results and satisfy both the fixed-outline and non-overlapping constraints while optimizing circuit performance.

Inspired by the improvisation process of music players, a population-based meta-heuristic algorithm-harmony search (HS) has been proposed recently. HS is good at exploitation, but it can be poor at exploration, and its convergence performance can also be an issue in some cases. To address these disadvantages, the distance bandwidth (bw) adjusting methods proposed in recent literatures are summarized and the exploration ability of HS improvisation is investigated in this paper. Further, the relationship between improvisation exploration and each parameter under asymmetric interval is derived, and an iterative convergence sufficiency of the iteration equation which consists of variance expectation and mean expectation is proven theoretically. Based on these analyses, a modified harmony search (MHS) algorithm is proposed. Moreover, the effects of the key parameters including HMS, PAR and HMCR on the performance of the MHS algorithm are discussed in depth. Experimental results reveal that the proposed MHS algorithm performs better than HS as well as its state-of-the-art variants and other classic excellent meta-heuristic approaches.

In this paper, a novel modified particle swarm optimisation (NMPSO) approach is presented to handle the many-objective knapsack (MOK) problem. NMPSO relies on the global best particle to guide the search of all particles in each generation. Furthermore, a randomisation-based mutation is adopted to overcome the premature convergence. A normalised penalty method (NPM) is devised to reach a compromise between objective functions and inequality constraints, which enables particles to explore solution space more precisely. In summary, the contribution of our work can be summarised in two aspects: 1) a more powerful approach called NMPSO is proposed; 2) a reasonable NPM is devised. Five improved PSOs are used to handle the MOKs with different number of objective functions and dimensions. Experimental results show that NMPSO has higher efficiency than the other four approaches. It uses the lowest computational cost and achieves the smallest penalty function values for most MOKs.