Abstract:

Abstract:

Abstract: TLD was a new long-term object tracking algorithm in recent years. It is different from the traditional tracking algorithm in that it combined the traditional detection algorithm and the tracking algorithm , and then the learning module was used to study the characteristics of the object. In this paper,an improved TLD object tracking algorithm based on FAST feature points was proposed to ensure that the selected feature points could be tracked correctly and reliably to improve the accuracy of the tracker. At the same time, for the tracking process, the learning module template cumulative effect is obvious, reduced the real-time performance. Using a dynamic template management mechanism, when the number of templates reached the threshold, by comparing the similarity between the template and the current target, the specific template is deleted and the number of templates was kept constant. Experiments results showed that the improved algorithm had higher tracking precision and real - time performance.

Abstract: This study focused on the fitness-based cooperative strategy and its improvement method. It was found that the problem of the most matching task could not be decided when the robot had the same fitness tasks. By adding the distance fitness function related to the robot’s starting and ending pisotion, the robot could choose the best matching task. At the same time, more realistic Gaussian distribution model is used to calculate the fitness of the external ability. The simulation results showed that the improved algorithm could not only achieve the optimal matching , but also be more efficient and energy saving than before.

Abstract: Bcause of the scale and the various shapes of down in the image, it was difficult for traditional image recognition method to correctly recognize the type of down image and got the required recognition accuracy, even for the Traditional Convolutional Neural Networks (TCNN). To solve the above problems, a Deep Convolutional Neural Networks (DCNN) for down image recognition was constructed, and a new weight initialization method was proposed. Firstly, these salient regions of images were cut from the images  using the visual saliency model.Then, these salient regions were used to train a sparse autoencoder and get a collection of convolutional filters, which accord with the statistical characteristics of dataset. At last, a DCNN with Inception module and its variants was constructed. To enhance the recognition accuracy, the depth of the network was deepened. The experiment results indicated that the constructed DCNN increased  the recognition acuracy by 2.7% compared to TCNN, when recognizing the down in  the images. The convergence rate of the proposed CNN with the new weight initialization method was improved by 25.5% compared to TCNN.

Abstract:

Research on vision from the perspective of network was a new direction. There was a contradiction between transmission efficiency and wiring cost in visual cortex network. Based on the principle of economy, the evolutionary game laws were proposed between the cost and efficiency under the limitation of energy by simulating the visual cortex network. The visual network model was built upon the sulci of 15mm in V1. The heterogeneous evolution of the visual cortex network confirmed the existence of contradiction between information transmission efficiency and wiring cost. The collaboration cliques generated by mixing the dynamic game strategies of efficiency and wiring cost validate the evolutionary game laws. The evolutionary game laws showed that there existed a dynamic evolutionary mechanism of class complex network in visual cortex network, and provided a new way for ordinary vision information management.

undefined

Abstract: In order to improve the payload-weight ratio, a novel quadruped walking robot was presented. And aiming at the impingement in the process of movement of the quadruped walking robot, a low contact compact trajectory planning method using high order polynomial curve was used. Each leg of the robot was composed of a 3-RRR parallel mechanism which was connected in series with a rotating mechanism. The coordinate system was established based on D-H method. Based on the low contact compact algorithm and the uniform linear motion, the trajectory planning was carried out respectively for the swing phase and the support phase of the foot end of the serial-parallel leg. The trajectory of the foot end in the global coordinate system of the hip joint was calculated according to the inverse solution model. The four driving function of the leg was calculated using Matlab based on the inverse kinematics solution of the hybrid leg’’s kinematics model. Finally, the leg model was exported into the ADAMS, the motion simulation of the hybrid leg was carried out by using ADAMS after adding constraints and the drive. The result showed that the foot end of the hybrid leg achieved the goal of low impact, and the curve of foot end of the speed and acceleration was smooth. The simulation results were consistent with the theoretical results and the rationality and validity of the algorithm were verified. This research provided the foundation for the further research of robot’s gait planning and motion control.

Abstract: The single hidden layer feedforward neural network was efficient with simple structure. Back Propagation Error (BP) algorithm was one of its typical learning algorithm which had one main shortcoming of the slow learning speed because of the use of the steepest descent method. Extreme Learning Machine (ELM) which could greatly accelerate the learning speed of networks was put forward. However, it demanded much more hidden neurons than BP algorithm to get the match accuracy, which led to redundant structure of networks and more testing time. Motived by the USA (Upper-layer-Solution-Aware）) which was a combination of the steepest descent method and ELM, in this paper, we proposed an algorithm based on the conjugate gradient method and train the network on different data sets. The Simulation results showed our algorithm had a better performance than USA and ELM with the same structure of the network.

Abstract:

In the base station cooperation system, the centralized cooperative pre-coding based on regularized zero-forcing beam-forming (RZFBF) could obtain a similar capacity performance to that of dirty-paper coding. However, with the number of cooperative base station increasing, the centralized cooperative pre-coding had a quite high requirement for the backhaul capacity. In order to solve this problem, a distributed transmitted signal design scheme based on generalized approximate message passing (GAMP) algorithm was proposed. The scheme decomposed the overall compucentralized RZFBF pre-coding problem into a problem of signal estimation by constructing a virtual linear communication system, and decomposes the overall computational cost into many smaller computation tasks by exchanging information between adjacent base stations, which greatly reduced the requirement of backhaul capacity. Simulation results showed that the proposed algorithm could achieve approximate system tjroughput with the centralized cooperative RZFBF pre-coding at a lower computational complexity.  

Abstract: Followed the development law of power grid, the related index of the power grid development based on its own conditions and external factors to establish the diagnosis model by using ANP. By considering the energy, environment put forward, land and other resources on the development of power grid constraints, we innovative index was from the perspective of long-term development of the power grid. Considering the mutual connection and restriction between those indexes, use the model to diagnose the present situation and development space of the power grid development. Based on the example of diagnosis, the diagnosis model could guide the development direction of regional power supply, optimize energy structure, find the weak link in the development of power grid, evaluate the development space of grid, and grasp the development direction. It was a reasonable and feasible new type of power grid development diagnostic method.

Abstract: In this paper, Lead Magnetroheological damper，LMRD, was taken as a kind of new damper designed to reduce effectively the seismic response of structures. The position optimization of LMRD in structure was to improve the damping effect of controlled structure in LMRD, and it was taken the optimal structural quadratic performance index loss produced from removing damper as optimization goal at the same time. The result showed that the optimal position of LMRD in structure was depended on the structural properties and parameters, which had little relevance to the type of seismic wave, when considering the loss of structural optimal quadratic performance index without LMRD as the optimization objective. Also, LMRD could control the horizontal displacement response of structure, under the seismic load, effectively, rather than the acceleration.

Abstract: Constant strain rate loading could effectively solve the stress uniformity and the dispersion effect of the SHPB device in the application of rock-like materials based on a comprehensive analysis of the testing technology on the SHPB data processing principle and the conditions which SHPB device applied in the rock material. It was determined that conical-cylindric bullet was an effective way to realize constant strain rate loading. An analysis of different shape parameters was made by means of numerical calculation for the total length of 400mm conical-cylindric bullet, it was seemed that the optimal area ratio between small end-face and large end-face was 0.7, and the specimen length-diameter ratio of granite was between 0.75 and 0.82. The optimal cone length is 300mm. By comparing the waveform with traditional cylindrical bullets, it was proved that the designed conical-cylindric bullet could achieve the constant strain-rate loading of granite was verified from two aspects of numerical simulation and experiments.

Abstract: Traxial test were carried out to research the stress-strain relationship and strength property of flax fiber-reinforced sand. The variation of shear strength indexes of flax fiber-reinforced sand in different flax fiber content and different confining pressure was investigated. And the frictional mechanism between the flax fiber and sand in the shearing process was studied. The laboratory test results showed that, the strength and the capacity for resisting the deformation of flax fiber-reinforced sand were better than those of pure sand. Compared with the pure sand, the shear strength indexes of flax fiber-reinforced sand were changed, the cohesive and friction angle were increased, while the friction angle increased relatively less, the cohesive force increased greatly; in the certain number of rate of water and fiber length, the value of deviator stress, shear strength and cohesion firstly increased and then decreased with the increase of flax fiber content, and there existed an optimum fiber content.

Abstract:

The waste tire ground rubber was used as a modified material, to enhance the contaminants adsorption capacity of soil-bentonite backfill. Based on the consolidations tests of soil-bentonite modified with ground rubber, the effects of ground rubber and simulated leachate on consolidation and compression properties were studied. The soil-bentonite modified with ground rubber was medium to high compressibility soil. As ground rubber particles had different compressibility and hydrophobicity from clay particles, the compressibility and consolidation velocity of the samples increased slightly with the increase of rubber mixing content. For the sample modified with bigger ground rubber, the compressibility and consolidation velocity were lager. When the rubber mixing content is constant, simulated leachate would decrease the liquid limit, the combined water and the thickness of absorbed water, which would reduce the compressibility and increase the consolidation velocity of the sample.

undefined

Abstract: To study the optimum preparation process of HDPE/SBS/waste tire powder mixed with three kinds of modified asphalt, SK-90 # asphalt was used to select SBS content, powder content, HDPE content and Shear rate, shear time and shear temperature to prepare compound modified asphalt. The results were analyzed by variance analysis method. The influence of each test variable on the performance of composite modified asphalt was obtained, and the optimal preparation scheme was obtained. The results showed that the SBS content had the greatest effect on the softening point and the elastic recovery rate. The content of HDPE had a great effect on the penetration. The SBS and HDPE content had an influence on the ductility. The content of the modifier and the shear temperature the effect of rotational viscosity was essentially the same. The optimum preparation scheme was as follows: the content of HDPE was 1.5%, the content of SBS was 5%, the content of waste tire was 20%, the shear temperature was 165 ℃, the shear rate was 4500r / min, the shear time was 45min. Under the optimal preparation scheme, the high temperature stability and low temperature crack resistance of the modified asphalt were greatly improved, and the viscosity of the modified asphalt was increased and the adhesion to the aggregate was enhanced, which improved the anti-rutting ability and the ability of resisting the water resistance. The elastic recovery rate could ensure the crack resistance of modified asphalt.

Abstract: The experimental results obtained from the conventional low Reynolds number wind tunnel test may produce errors due to the Reynolds number effect. And the improper simulation of turbulent wind parameters may lead to further increase of the error in the wind tunnel test. By using grille turbulence in the wind tunnel, a series of local turbulent wind fields were formed with the same turbulence intensity but different integral scale and with the same turbulence integral scale but different intensity.  This paper examined the Reynolds number effect for the surface wind pressure of rectangular structure and the effect of turbulence intensity and integral scale on the Reynolds number effect for the surface wind pressure. The results showed that: the Reynolds number effect of the mean wind pressure and the fluctuating wind pressure was the lowest in the windward side, the maximum in the leeward side and the second in the crosswind. The increase of the turbulence integral scale and intensity would lead to the further increase of the mean wind pressure’s rangeability in the sensitive region of the Reynolds number effect. The influence of the integral scale on the Reynolds number effect of the fluctuating wind pressure was smaller than that of the mean wind pressure, especially on the leeward side. And the turbulence intensity had little effect on the Reynolds number effect. The fluctuating wind pressure on the structure surface was mainly dependent on the turbulence intensity of the inflow, which was independent of the scale of the model.

Abstract: Solid acid catalyst SO2- 4/ZrO2/USY was prepared by precipitation- dipping method. And the effects of different parameters on the conversion of cellulose to ethyl levulinate were also investigated. The results showed that 10% SO2- 4/ZrO2/USY presented the highest catalyst activity when calcined at 400 ℃ for 3 h, and the high yield (38.1 mol%) of EL was obtained at 200 ℃ for 3 h, with a catalyst dosage of 10 g/L.  EL yields of 50.77mol%、39.55mol%、44.87mol% and 43.95mol% could be obtained from frutose, glucose, surcose and inulin, repectively. The catalysts were characterized by XRD、FT-IR、BET、NH3-TPD and TG-DTA. The prepared solid acid catalyst could be recycled and reused four times with good catalytic activity.

Abstract: Nb-doped TiO2 transparent conductive thin films were prepared on Glass substrates by magnetron sputtering method. The structure and the photoelectric properties of the films were characterized by using X-ray diffraction (XRD), atomic force microscope (AFM),UC-Vis spectroscopy and four probe resitance tester. The results showed that the anatase phase Nb-doped TiO2 thin films were obtained when the annealing temperature above 250℃, also the structure and the photoelectricl properties were improved with the temperature increased. The best visible light transmissivity reached to 80% and resistivity droped to 2.5×10 -3 Ωcm when temperature at 300℃. As the annealing temperature were risen to 350℃, the films began to appear rutile phase, then the photoelectric properties were decreased. Moreover Nb doping was beneficial to reduce the crystal phase transformation temperature of the TiO2 thin film. The Nb-doped TiO2 thin film absorption edge also produced blue shift, and the blue shift degree of the thin films absorption edge was different with the change of annealing temperatures.

Abstract: Glazed tiles of glass-ceramics were prepared by direct sintering using blast furnace slag as main raw material. The results showed that the composition of blast furnace slag was glass phase, which chemical composition was CaO, SiO2, Al2O3, a little of TiO2, and Fe2O3. When particle size was 3.3μm; the content of blast furnace slag was 80wt%, potassium feldspar was 10wt%, talc was 10wt%; crystallizing at 800℃ for 30min and sintering at 1160℃for 1.5h, the sample’s water absorption was 0.12%, micro-hardness was 6.5GPa; flexural strength was 128MPa, which met the GB performance requirement of glazed tile. The main crystal phase of glass-ceramics was wollastonite(contain a little of anorthite ). Crystal size was about 1μm and the morphology of crystal is globular and needle-like.