2021 volumne 42卷 Issue 05
LI Xuexiang, CAO Qi, LIU Chengming
Abstract: Image super-resolution reconstruction based on generative adversarial networks (GAN) is subject to the dataset training with an unstable result. To solve this problem, a new NM-SRGAN model is established. The cycle-gan is firstly used as the preprocess module to make the model free from the dataset for training with better input of the image, and the model cancels BN layer to solve the unstable results. Besides, covariance matrix is adopted to capture the second-order information of the image, and second-order loss function is added with a focus on the changes of the image details. The new VGG loss function is used to improve the marginal texture of the image. The proposed NM-SRGAN model is verified by four standard datasets, and the resulting images are assessed by the objective evaluation indices. Compared with the existing models, NM-SRGAN model has an improved evaluation value of 0.19, 0.03, 0.13, and 0.02 dB, respectively, reaching up to the maximum among the four datasets. Results show that the proposed method, compared with traditional algorithms, has achieved better improvements in stability and image quality with better details.
Ye Jihua; Guo Qiyi; Jiang Aiwen; Li Xin;
Abstract: To solve cross-age face recognition tasks, this paper introduces the direct sum module on the basis of the multi-task convolutional neural network that simultaneously performs two tasks of face recognition and age classification, and proposes the feature subspace with direct sum multi-task convolutional neural network (FSDS-CNN). The network uses two parallel subnets to extract the identity-related feature and age-related feature from the deep feature, then the direct sum constraint is applied to the feature subspaces corresponding to these two related features, so that the correlation between identity-related feature and age-related feature is decreased as much as possible. Through the joint supervised learning of multiple loss functions, the network can obtain age-invariant face identity feature that is robust with age. Cross-age face recognition and verification experiments is conducted on three datasets (Morph Album 2, CACD-VS and Cross-Age LFW). In the CACD-VS dataset, the proposed method achieves the optimal result of 99.7% on the evaluation metric of AUC; in the Cross-Age LFW dataset, the method respectively achieves the optimal results of 10.1% and 10.2% on the evaluation metric of EER and FNMR when FMR is 0.1. At the same time, the ablation comparison experiments are conducted on the three datasets to verify the effectiveness of the direct sum module. The results show that the correlation between identity features and age features is effectively reduced by the direct sum module in FSDS-CNN, and then effectively improves the performance of cross-age face recognition.
Zhang Jianxin; Guo Si Jing; Zhang Guolan; Tan Lin;
Abstract: This paper aims to modify the two-scene detection model Faster R-CNN. Specifically, this model uses Resnet101 to extract features which are processed by pyramid structure FPN to extract the shallow and high-level features of Resnet101. The shallow feature map of Resnet101 is input into Inception Module structure to obtain the convolutional features of multiple sizes, and finally the proposed model uses the pixel attention mechanism and channel attention mechanism to emphasize the target position and weaken the rest, which makes the detection target more accurate. This network avoids the problem of insufficient feature extraction of trunk network, and integrates features of various scales to distinguish fire area and non-fire area, thus effectively improves the detection accuracy of fire image data sets, and mean average precision MAP is 0.851.
Wang Xipeng; Li Yong; Li Zhi; Zhang Yan;
Abstract: Due to the limitation of video information, target tracking in the case of occlusion is still a difficult problem to solve. Aiming at the problem of occlusion in the target tracking process, it is proposed to introduce image depth into single target tracking algorithm. Firstly, the monocular image depth estimation algorithm is used to estimate the depth of the image to obtain the depth information of the image. Secondly, the target tracking algorithm based on the siamese region proposal network is combined with the image depth to construct an occlusion discriminating module, which uses the change of the target depth information to determine the occlusion. Finally, the occlusion discrimination score and the anchor response score are weighted integrated. According to the final response score, the anchor of the target tracker is reordered to avoid interference by obstructions. Experimental results on the OTB-2015 dataset show that the algorithm can effectively deal with the influence of occlusion on tracking performance, with an average success rate of 0.623 and an average tracking accuracy of 0.853, which is 1.7% and 0.9% higher than the benchmark algorithm, repectively.
Lu Lulu; Chen Shuyue; Wang Liping; Xu Xia;
Abstract: Aiming at the problems of time-consuming and labor-consuming manual identification of water microfibers, and the weak robustness of traditional image processing algorithms for identifying water microfiber images, an improved MobileNetV2 network identification method for microfibers is constructed. In the feature extraction part, the feature reconstruction strategy is adopted. Firstly, the deep convolution features are compressed to obtain the global receptive field. Then, the fully connected layers are used to generate weights for each channel to establish the interdependence between the channels. Finally, the channel is weighted to the original in terms of features to complete the reconstruction of the original features. In addition, different sizes of downsamplers are used to capture and fuse feature information of different scales to enhance the detailed feature information of microfibers, and to improve the model′s learning ability and recognition effect of microfibers. The improved MobileNetV2 network′s microfiber recognition accuracy rate reaches 97.96%. Compared with the original MobileNetV2 network, the recognition accuracy rate is increased by 2.54%. At the same time, the false recognition rate and the missed recognition rate are also significantly reduced. In comparison to ResNet, DenseNet, VGG16 and NasNet networks, the model size is compressed several times, the accuracy of microfiber recognition is improved, and the false recognition rate and missed recognition rate are greatly reduced. Experimental results show that the network model can extract more complete feature information for microfiber. While strengthening the microfiber feature to identify the directivity, the model is reduced, and the difficulty of deployment in mobile devices is reduced as well. The improved model recognizes microfibers with higher accuracy and better stability.
ZHANG Sanchuan, MING Zhu
Abstract: Local path planning is the key to the active safety of intelligent driving vehicles. In order to solve the theoretical problems of local minimum and unreachable target existing in the traditional artificial potential field method, in this paper the distance between the experimental vehicle and the obstacle is introduced as the repulsive force regulator based on the measurement function of the azimuth(θ0)of the obstacle by millimeter-wave radar, which makes sure that the repulsive forces near the target point are not too large. At the same time, the additional force of target gravity is introduced with direction angle θ (>θ0) and controlled by target distance k·S(M,Mg), which makes experimental vehicle break away from the minimum point. The numerical simulation results of MATLAB show that: when the gain coefficient of the additional force (k) is between 5~7, a stable and safe local planning path can be obtained, and no minimum point appears. The variation of the peak value of repulsive force and resultant force in the improved artificial potential field decreases exponentially with the increase of the distance between the obstacle and the starting point of path planning, the repulsive force of the experimental vehicle near the target point attenuates to 0, and the target is reachable; Compared with the traditional artificial potential field, the single-step calculation time is slightly increased, but there is no oscillation interval for the planned path. The simulation time is 0.26 s, and the timelessness is basically the same. The safety index is increased from 0.018 8 of the traditional artificial potential field method to 0.305 0, which greatly improves the safety of local path planning.
Yu Fangxing; Ji Bo; Cheng Quanrun; Lu Hongxing; Liu Hongchuan;
Abstract: The dual-cavity optical feedback interference system is often used for high-sensitivity sensing of moving objects. Its dynamic behavior can be solved by the Lang-Kobayashi (L-K) equation, and the accuracy of the solution will have a decisive influence on the measurement accuracy. In order to improve the measurement accuracy of the dual-cavity OFI system for moving objects, a sixth-order Runge-Kutta algorithm to solve the L-K equation is proposed. By analyzing the principle of the numerical solution method of differential equations, more interval points are selected to calculate the average slope of the integral curve on the basis of the fourth-order Runge-Kutta algorithm, so as to make it closer to the real value and further improve the solution accuracy. At the same time, the simulation software of moving object motion detection is designed and implemented based on the optoelectronic signal dual-cavity OFI system for simulation experiments, and the simulation results of the sixth-order Runge-Kutta algorithm is compared with the Euler method and the fourth-order Runge-Kutta algorithm. Experimental results show that compared with Euler′s method, the solution accuracy is improved by about 22% on average; Compared with the fourth-order Runge-Kutta algorithm, the solution accuracy is improved by about 6% on average. The sixth-order Runge-Kutta algorithm can improve the solving accuracy of L-K equation, thus generating more accurate simulation results and improving the sensing sensitivity of the dual-cavity OFI system.
Huang Qian; Wang Shuqin; Deng Shaohong; Fan Linjun;
Abstract: Uncertain factors often affect the rescue operations and effects of troops. In the conditions of limited resources and urgent time, it is very important to select the locations of the troops, allocate the tasks of disaster relief, plan the rescue routes, organize efficient rescue, and achieve the overall optimal effect of disaster relief, overcoming the influences of uncertainty. Assuming that the time of troops′ movement and the time required for disaster relief are all in normal distribution, a multi-objective stochastic programming model of location routing problem (LRP) with the minimum total cost and time of disaster relief is established. The random constraints are transformed into the objective function by introducing the penalty factors. The normalized sum of each objective function value is taken as the fitness function value. Based on this, an improved genetic algorithm is proposed. The experimental results show that the total rescue time of the improved genetic algorithm is shorter than the one of basic genetic algorithm, and the improved ant colony algorithm has shorter total relief time and lower disaster relief cost, but the penalty value is very big, which verifies the superiority of the improved genetic algorithm
MA Ding1, 2, FEI Xuan1, MU Xiaowu2
Abstract: In network function virtualization environment, topology abstractions are crucial to create unified topology view and conceal underlying details. To map service function chain effectively, a VNF-aware virtua-lization layer constructing algorithm based on adjustable hop count is proposed considering network topology abstraction. Firstly, virtual nodes mapping is implemented via analyzing the virtual network functions required in service function chain request. Secondly, virtual links mapping is implemented via proposed non-redundant link mapping method with adjustable hop count. To evaluate the performance under different hop count, the layered graph algorithm is used as the service function chain mapping algorithm which is executed over the constructed virtualization layer. The experimental results show that: when the value of hop count equals 3, compared to the situation in which the value of hop count equals 2, the overall performance is improved significantly; when the value of hop count increased to 4 and 5, compared to the situation in which the value of hop count equals 3, the cost of constructing virtualization layer increases by 27% and 52%, respectively. However, the overall performance improves slightly. Finally, simulation experiments show that the proposed algorithm can effectively map service function chain requests and determine the optimal virtualization layer in terms of performance and cost for the specific physical network.
LU Peng1,2, WANG Hanzhang1,2, MAO Xiaobo1,2, ZHAO Yuping2,3, LIU Chao1,2, SHANG Lijia4, SUN Zhixia5
Abstract: Classification of pulse wave based on deep learning relies on a large number of labeled data, however, limited clinical data and expensive labeling costs hinder the pulse wave classification and recognition. A pulse wave classification model based on convolutional autoencoder networks (CAE-Net) is designed in this paper. Firstly, the convolutional autoencoder (CAE) is constructed, which combines the local feature extraction ability of convolutional neural network (CNN) and the compression reconstruction and dimension reduction characteristics of autoencoder (AE). And considering the characteristics of pulse wave, the time domain feature constraint of pulse wave is introduced into the mean absolute error loss function to improve the self-learning ability of CAE for low dimensional features. Secondly, the CAE-Net is constructed by reusing the coding layer network and weights of the pre-training CAE, then the network is fine tuned by using labeled pulse waves. Experiments on cardiovascular disease dataset show that the classification accuracy of CAE-Net is 98.00%, and the F1 score is 94.40%. Compared with other classification models, the designed network can extract features with high discrimination, reduce the dependence on the labeled pulse waves, and perform well in the classification of small sample pulse wave data.
MA Ge, JIA Suimin
Abstract: In order to meet the increasing spectrum demands of cognitive vehicular network, spectrum allocation problem is studied. Cognitive vehicle network is the formulation of cognitive radio technology into vehicle network. For the problem of cognitive vehicle network spectrum allocation,which is modeled to optimize the throughput of cognitive nodes, an immune optimization algorithm based on graph theory model is proposed. To ensure the optimization of algorithm, it concludes matrix-coding scheme, antibody correction mode, and proportion of cloning. The simulation results show that the proposed algorithm can obtain high cognitive node throughput and is suitable for the spectrum allocation of cognitive vehicle network.
Huang Jiayu; Liu Yuanzhen; Gao Yuxuan; Chao Chaoxu;
Abstract: Based on the creep test of recycled thermal insulation concrete under constant temperature and humidity, this research studies the different development of creep value, creep coefficient and elastic modulus of recycled thermal insulation concrete when the replacement rate of recycled coarse aggregate changes. Based on the ACI-FIP (1990) model, a new modified model is proposed to predict accurately the development state of creep coefficient of recycled thermal insulation concrete when the replacement rate of recycled coarse aggregate is different. Experiments showed that with the increase of the replacement rate of recycled coarse aggregate, the elastic modulus of recycled thermal insulation concrete decreased, and the creep value and creep coefficient increased significantly. Compared with thermal insulation concrete, the elastic modulus of recycled thermal insulation concrete with recycled coarse aggregate replacement ratios of 50% and 100% decreased by 10% and 12%, respectively. At 180 days, the creep of recycled thermal insulation concrete with recycled coarse aggregate replacement ratios of 50% and 100% increased by 6% and 17%, and the creep coefficient increased by 12% and 31%, respectively. Comparing the predicted results of the modified creep model with the test results, through linear regression analysis, it is found that the linear regression coefficients (R2) of recycled thermal insulation concrete models are all between 0.91 and 0.93, indicating the model has a good degree of fit and it can better predict the change of concrete creep coefficient with variation of recycled coarse aggregate replacement ratios.
Shi Xiaoyi; Liang Yan; Wan Dekun; Chen Huai; Fang Leilei;
Abstract: The large-span steel box girder has complex forces during the jacking construction process, and the temporary structure of the guide beam has a greater impact on the safety of the long-span steel box girder jacking construction process. In this paper, a large-span steel box girder pushing construction is taken as an engineering example. According to the instability of the guide beam at the variable cross-section that occurs during the actual pushing process, the finite element software ABAQUS is used to establish its finite element calculation model, combined with the on-site guide beam instability, the section deformation verifies the accuracy of the finite element model. By analyzing the local instability and buckling deformation characteristics of the guide beam, the stability of the guide beam under four different reinforcement schemes of one row of longitudinal stiffeners, two rows of longitudinal stiffeners, one row of longitudinal stiffeners and two row of vertical combined stiffeners, and two rows of longitudinal and two rows of vertical combined stiffeners are further studied. According to the analysis of the calculation results, the first order instability characteristic values of the guide beams in the four reinforcement schemes are increased by 1.9%, 63.8%, 26.0% and 93.5%, respectively. The effect of longitudinal stiffeners is better than that of vertical stiffeners. Combining the actual force situation of the site structure, the longitudinal stiffening ribs can be preferentially used for reinforcement. According to the results of numerical analysis and the time requirements of high-speed traffic, a guide beam reinforcement plan that conforms to the actual project and is convenient for construction is determined: two rows of longitudinal stiffeners are used to strengthen. The reinforcement effect shows that the reinforced guide beam did not lose stability during the follow-up jacking construction process, ensuring the safety of the follow-up jacking construction of the steel box girder. In view of the uncertainty of the jacking construction process, the jacking construction process should be monitored in real time, and the beam should be dropped in time when abnormal conditions occur. The research results of this article can provide references for similar projects.
MIAO Yanchun1, ZHANG Yu1,2, LEI Chuang1, LI Minghou1, LIU Yuanzhen1, LI Zhu1
Abstract: Based on the meso-scale heterogeneity of recycled aggregate thermal insulation concrete (RATIC), MATLAB software was used to generate a 2-D polygonal random aggregate model of RATIC by Monte Carlo method, and then the finite element analysis software ABAQUS was used to simulate the uniaxial compression mechanical properties of meso-scale RATIC based on coupled thermo-mechanical modeling. Firstly, the heat conduction behavior of RATIC at different temperatures was simulated. According to the simulation results, the effects of meso-scale constituents at different temperatures (100, 200, 300, 400, 500, 600, 700 and 800 ℃), such as the thermal parameters (conductivity, specific heat and thermal expansion coefficient) and the mechanical parameters (strength, elastic model and Poisson′s ratio), on the meso-scale RATIC mechanical properties were explored. Furthermore, a comparative analysis was conducted to study the uniaxial compression failure modes at different fire temperatures of RATIC under simulated and experimental conditions. The results show that the temperature stress weakens the strength of RATIC when the temperature exceeds 400 ℃. And at 800 ℃, there is a maximum temperature stress of 3.309 MPa generated inside the specimen. The high temperature damage of RATIC specimens under uniaxial compression first appears in the interfacial transition zone, and then develops to the mortar. It is mainly concentrated on the free end of the specimen, and with the increase of the fire temperature and loading time, the damage shows a gradual increase trend. The results indicate that the meso-scale model can be well used to simulate the uniaxial compression mechanical properties and failure patterns of RATIC at high temperature.
Wang Mingdong; Zhou Wei; Li Xiaolei; Li Zhongwen; Wang Zixi;
Abstract: As a new AC/AC conversion device, modular multilevel matrix converter (MMMC) has advantages in low frequency and large capacity applications such as fractional frequency transmission system, whereas its bridge arm current control is complicated. In this paper, the topological structure of MMMC is analyzed and the mathematical model is established to study the structure of bridge arm current. In order to solve the problem of poor control precision of bridge arm current by traditional direct control method, quasi-proportional re-sonant (QPR) control strategy is introduced, and a quasi-proportional resonant controller based on input and output frequencies is designed according to the characteristics of fractional frequency transmission system. According to the control requirements of the system active and reactive power, the outer loop control strategy of constant active power control and constant reactive power control is given. At the same time, the relationship between the capacitive voltage of the sub-module and active power is deduced, and the voltage balance strategy of the submodule based on active power of the system is proposed by applying double Clarke transform. Finally, the MMMC-FFTS system simulation model is built on the simulation platform, the steady-state and transient response of the system are simulated respectively, and the simulation results of direct control and QPR control methods are compared. The simulation results show that the improved control method has good steady-state and transient performance, and the accuracy of reference tracking is higher, which verifies the feasibility and superiority of the proposed method.
Meng Qinglong; Wang Wenqiang; Li Weilin; Xiong Chengyan; Li Yang; Responsibility;
Abstract: Aiming at the problems of air conditioning system participating in power grid demand response, the demand response (DR) of multi energy interaction in building power grid is comprehensively studied and analyzed from the perspective of HVAC system characteristics. The definition and classification of HVAC demand response are summarized, and the methods of using model predictive control (MPC) algorithm, genetic algorithm (GA) and other algorithms to predict the potential of HVAC demand response are discussed. The principles and applicability of DR strategies such as resetting regional temperature, increasing air supply temperature, resetting chilled water temperature and so on are summarized and analyzed. The analysis shows that for DR projects where the user′s thermal comfort is improved after the implementation of DR, this strategy can be considered to reduce energy consumption during daily system operation, and the combination of active energy storage strategy and conventional DR strategy can effectively solve the load rebound problem of DR events. Therefore auxiliary services should be considered for those users with large adjustable air-conditioning load.
ZHU Qiang1, ZHANG Dongsheng2, FAN Yuheng1, ZHANG Jing1, ZHAO Hongliang1
Abstract: The vacuum reaction infiltration method was used to prepare C / Cu composites with carbon fiber/resin carbon (C/C), carbon fiber (Cf), graphite (Graphite), and glassy carbon (GC) as matrix carbon. The OM, XRD, SEM, EDX and other detection and analysis methods were used to study the influence of different matrix carbon on the wetting behavior of C/Cu composites and the microstructure of the interface layer. The results showed that C/C and Cf had good wettability with Cu-Ti alloy, and the contact angles were 56.26° and 40.12°, respectively. However, Graphite and GC did not wet with Cu-Ti alloy, and the contact angles were both larger than 90°. The good wettability between Cu-Ti alloy and C/C、Cf is mainly because the TiC layer which is formed by reaction-diffusion during the vacuum reaction infiltration can be wetted well with C and Cu, improving the C/Cu interface wettability and interface bonding state. The TiC interface layer is dense and uniform, and had the largest thickness when the matrix carbon is Cf. Finally, by considering the wetting behavior and the interface layer microstructure, the order of precedence for different carbon matrix is as follows: Cf 、C/C、Graphite、GC.
Zheng Jin; Zhang Qi; Song Meng; Wang Dong Shuang; Yang Kai; Jiao Mingli;
Abstract: To study the changes of the microstructure of PET in D5 is helpful to understand the dyeing mechanism of PET with D5 as the medium and make a reasonable dyeing process. In this study, molecular models with different proportions of polyester, D5 and 1-methylnaphthalene coexisted, and the effects of solvents, accelerators and temperatures on the microscopic characteristic parameters of polyester were simulated by molecular dynamics. The glass transition temperature (Tg) of pure PET system constructed by molecular simulation was in good agreement with the Tg of PET measured by experiment. After adding D5, the Tg of PET decreased significantly. With the addition of 1-methylnaphthalene, Tg decreased further. The addition of D5 alone did not significantly improve the kinematic properties of PET chains, but when 1-methylnaphthalene coexisted with D5, the kinematic properties of PET chains were significantly improved. The presence of D5 and 1-methylnaphthalene increases the total free volume of PET. Molecular simulation of polyester microstructure showed that D5 and 1-methylnaphthalene were beneficial to increase dyeing rate, reduce dyeing temperature and improve dyeing effect.
Copyright © 2023 Editorial Board of Journal of Zhengzhou University (Engineering Science)