Table of Content

25 December 2018, Volume 46 Issue 12

Previous Issue    Next Issue

Computer Science & Technology

Normal Estimation from Single Monocular Images based on Multi-Scale Convolution Network

XIAN Chuhua LIU Xin LI Guiqing JIN Shuo

2018, 46(12):  1-9.  doi:10.3969/j.issn.1000-565X.2018.12.001

Asbtract ( 291 )   PDF (2133KB) ( 320 )  

Related Articles | Metrics

Normal estimation from monocular images is one of the most important issues in computer graphics and computer vision research． Short of three-dimensional information，the corresponding normal is predicted from the monocular images，which is of great significance for 3D scene reconstruction，3D model recognition，3D semantic segmentation，etc． In order to find the solution to the problem，this paper adopts a multi-scale convolutional net- work structure to predict an end-to-end output of the image． The network consists of two scales，the first layer uses the DenseNet classification network with the best performance in ImageNet to process the input globally． The second level uses a fully convolutional network to further fine-tune the output obtained from the first level． The experimen- tal results show that the network proposed in this paper can achieve better results in normal prediction of monocular image even without using other pre-processing or post-processing steps．

An Improved Camera Pose Estimation Method for SLAM System

KONG Dehui LI Wenchao YU Yilan LI Jinghua YIN Baocai

2018, 46(12):  10-19.  doi:10.3969/j.issn.1000-565X.2018.12.002

Asbtract ( 308 )   PDF (2065KB) ( 115 )  

Related Articles | Metrics

Camera pose estimation is a key step in SLAM system，which affects the accuracy and efficiency of the whole SLAM system． At present，there are two main methods to estimate the pose of camera，namely，the feature point method and the direct method． The accuracy of feature point method depends on the number of feature points and the correctness of feature matching． When enough feature points cannot be extracted in the scene，the position and posture of the camera cannot be estimated accurately． The direct method estimates the position of the camera by the pixel’s photometric error，and does not need to extract the feature points． Therefore，the direct method can still estimate the position and pose of the camera more accurately when the feature point method is unable to work． But the direct method assumes the luminosity invariance，so the accuracy of the result is not as good as that of the characteristic point method． Aiming at the problems of camera pose estimation in SLAM，an improved camera pose estimation method is proposed in this paper． The main idea of this method is to combine the feature point method with the direct method to overcome the estimation of the position and posture of the camera when the feature point is lacking，and to improve the accuracy and robustness of the position and posture of the camera． In particular，first， a feature matching algorithm which combines camera motion model with image division is proposed． The algorithm improves the accuracy and quantity of feature matching while guaranteeing the matching speed． Secondly，on the basis of the feature points，by introducing the photometric information，an apparent shape weighted fusion method is proposed to estimate the position and posture of the camera． This method can still work steadily when the feature points are lacking． Finally，on the basis of the preferred key frame，the local and global fusion of camera pose opti- mization is realized，in which the local optimization is realized by constructing the common view relationship of the local key frame，and the global optimization is realized by the pose graph based on the closed loop detection． In or- der to verify the performance of the pose optimization method，a SLAM system based on this method is constructed， and the reconstruction experiments are carried out on the current popular scene image data set． The reconstruction results verify the effectiveness of this method．

Interactive Patterns Collage along a Given Boundary

Lü Lin CHEN Minghai XU Fan

2018, 46(12):  20-25,33.  doi:10.3969/j.issn.1000-565X.2018.12.003

Asbtract ( 280 )   PDF (1692KB) ( 215 )  

Related Articles | Metrics

Recent years have witnessed rapid development of digital fabrication techniques，causing manufacturing individual models to be accessible to ordinary users． Thus designing tools for customized objects along a user- friendly way is getting highly demanded． Inspired by texture synthesis and image mosaics，this paper proposes a method that generates an editable patterns collage design along a given boundary: users’inputting a close curve as the boundary and several pattern units; the method firstly generates an initial patterns collage design along the giv- en curve boundary automatically; then users can directly edit each pattern unit; adjustment of patterns collage lay- out would take place on the basis of a spring mass system meanwhile maintaining the boundary constraint automati- cally． After the whole interaction step，our method optimizes the collage result for local connectivity in order to meet the fabrication constraints，so that we can manufacture this design by digital fabrications such as 3D printing and digital cutting． To demonstrate the effectiveness，we show the fabricated results via 3D printing and digital pa- per cutting．

Video Object Detection Based on Correlation Feature and Convolutional Neural Network

LIU Yujie CAO Xianzhi LI Zongmin LI Hua

2018, 46(12):  26-33.  doi:10.3969/j.issn.1000-565X.2018.12.004

Asbtract ( 244 )   PDF (1406KB) ( 77 )  

Related Articles | Metrics

The problem of mutual restriction between speed and precision caused by using image detection algorithm in the field of video object detection，a video detection method based on correlation features and convolutional neu- ral network is proposed in order to make full use of the target’s motion between frames． Our methods are demon- strated as follows: firstly，an image detection algorithm is used to extract features from the current video frame; sec- ondly，the correlation features between the frames is employed to predict the feature maps of the current frame and finally，the target motion information from the associated features is used to predict the final result． The method proposed in this paper finally experimented on the ImageNet dataset，which is proved better than the current method since the precision is enhanced and a faster speed is maintained．

Single Color Image Dehazing Based on the Basis of H-NL-MTV Model

WEI Weibo LI Shuai PAN Zhenkuan HOU Guojia ZHAO Shengnan

2018, 46(12):  34-41,50.  doi:10.3969/j.issn.1000-565X.2018.12.005

Asbtract ( 230 )   PDF (1852KB) ( 128 )  

Related Articles | Metrics

The algorithm based on dark channel prior can effectively dehaze the hazy images in different scenes， but the dehazed images usually contain more noise． Non-local Multi-Channel Total Variation (NL-MTV) can be used to denoise color images while maintaining edge-preserving effects and the original texture can be preserved af- ter denoising． In this paper，a new image dehazing algorithm is proposed by combining these two methods． First- ly，we establish the energy functional of the atmospheric light with atmospheric transport function (H-NL-MTV model) and use Alternating Direction Method of Multipliers by introducing auxiliary variables to solve the energy functional，and finally use MATLAB software to simulate experiment． The data results show that the image ob- tained by the proposed algorithm is clear and natural，and that the edges of the image are kept well，and that the texture features are preserved．

A Multi-Feature Incremental Learning Neural Network for the Quality Enhancement of Video Reconstructed Pictures in H. 265/HEVC

DING Dandan CHEN Jingsen FEI Jialuo TONG Junchao PAN Zhigeng YAO Zhengwei

2018, 46(12):  42-50.  doi:10.3969/j.issn.1000-565X.2018.12.006

Asbtract ( 309 )   PDF (1476KB) ( 65 )  

Related Articles | Metrics

The new generation video coding standard H. 265/HEVC employs in-loop filter，which includes de-bloc- king (DBF) and sample adaptive offset filter (SAO)，to remove the blocking artifacts and reduce the distortions of reconstructed video frames． Both of DBF and SAO originated from signal processing theory，and the corresponding algorithms and parameters are designed and set manually． Although the computational complexity is relatively low， such filters may not deal with different kinds of contents well enough as the natural videos are much more complex． This paper formulates the loop-filter problem in video coding as an end-to-end regression problem，which can be solved by deep neural network． The relationship between reconstructed frames and original frames are mapped au- tomatically and as a result，the differences between them are minimized． The proposed Multi-Feature based Incre- mental Learning Network (MFILNet) includes 35 layers． The integrated network adopts global residual learning strategy and cascades several Feature Incremental Learning Blocks (FIBs) to extract features of different levels． Consequently，useful features are finally extracted，selected and enhanced to improve the perceptual ability of the network． Within each FIB，variable convolutional kernels are adopted． Inspirited by DenseNet，features from dif- ferent layers are fused，thus to facilitate information flow among layers． Experimental results show that with the scheme of combining density and sparsity，learning capability and generalization capability of the proposed network are boosted tremendously． Both objective and subjective quality of the video compressed frames is improved signifi- cantly． Consequently，the proposed network model is used to substitute the DBF and SAO in H. 265/HEVC． Up to 11. 2% and averaged 6. 32% BD-rate reduction is obtained． The model is also used after the DBF and SAO， 5. 24% BD-rate saving can be obtained in average．

Mechanical Engineering

Optimal Temperature Control Strategy for Electric Transport Vehicle Drive Motor Cooling System#br#

XU Jun SUN Yumin LI Gangyan FENG Lan

2018, 46(12):  51-57.  doi:10.3969/j.issn.1000-565X.2018.12.007

Asbtract ( 326 )   PDF (2491KB) ( 137 )  

References | Related Articles | Metrics

In order to guarantee electric transport vehicle drive motor working in safety temperature range, output power of electric vehicle driving motor cooling system needs to be precisely controlled. Without considering the contrary relationship between drive motor’s heat loss and output power from cooling system, traditional control methods such as threshold triggering will cause higher total power consumption. An optimal temperature control strategy was proposed to minimize total power consumption of the cooling system. Taking the calculated motor’s temperature corresponding to the lowest total power consumption, this strategy can fulfill the aim of energy saving through controlling electronic fan’s rotating speed, which is deduced from analysis of relationship between total power and motor’s temperature. Taking a certain type electric transport vehicle as the application object, a simulation model was setup based on AMESim-MATLAB co-simulation platform to testify feasibility of the control strategy. Meanwhile, test results from a specially built platform have shown that the total power consumption can reduced about 3.8% compared with threshold triggering control mode.

Effects of Decoupling Membrane and Disturbing Plate on the Dynamic Characteristics of Hydraulic Damped Ｒubber Mount#br#

WANG Yanan HE Xin LIAN Chen LI Qingfeng

2018, 46(12):  58-65.  doi:10.3969/j.issn.1000-565X.2018.12.008

Asbtract ( 238 )   PDF (2194KB) ( 138 )  

References | Related Articles | Metrics

The structure of a Hydraulic Damped rubber Mount (HDM) with a column channel and decoupling membranes is analyzed. Using direct Fluid-structure Interaction (FSI) Finite Element Analysis (FEA) method, the integral finite element model is developed, and the calculation of the dynamic characteristics and the analysis of the working process under different excitation frequencies are carried out. The results show that the decoupling membranes deform in the low frequency range, and the notch peak value and the peak value of the dynamic stiffness and the peak value of the loss angle are changed with the variation of the number and the stiffness of the decoupling membranes. On the basis of the original design, disturbing plates with different structural parameters are added respectively and the dynamic characteristics of these HDMs are recalculated. The influence of the disturbing plate on the peak values and their corresponding frequencies of the dynamic stiffness and the loss angles is analyzed, and the laws of the dynamic characteristic curves changing with the diameter and rod length of the disturbing plates in the low frequency range are also studied. The effectiveness of the finite element analysis and calculation is verified by the experiment. The FEA modeling method based on the direct FSI technology is well in reflecting the real working process of the interior structures in the HDM and the change rule of the dynamic characteristics, which can be used in the performance analysis and design process of HDMs with complex structures.

Research on dynamic characteristics of a magneto-rheological damper featuring piston bypass holes

LI Guojie LI Liping Subhash Ｒakheja SHANGGUAN Wenbin

2018, 46(12):  66-73,102.  doi:10.3969/j.issn.1000-565X.2018.12.009

Asbtract ( 274 )   PDF (1941KB) ( 124 )  

References | Related Articles | Metrics

The damping force of a novel magneto-rheological (MR) damper was tested on MTS system in different excitation frequencies and different currents. The MR damper has three bypass holes in the piston. Experiment results indicate that the damping force has a quasi-linear relation with excitation velocity in a much broader range. Damping force also increases with applied currents. No damping force spurt at each excitation frequency. Damping force is sensitive to excitation frequencies. These characteristics make vehicles better riding comfort. At last, damping force was calculated adopting the local head loss with Bingham model or Eyring model. Damping force calculated by Eyring model matches better with experimental data. Maximum deviation of damping force in each operating condition is less than 2.3%. And if no local head loss was adopted, the damping force by calculation deviates far from experiment data. These results indicate that the calculation is reasonable and will contribute to design and optimization of MR damper.

Performance Analysis and Parameter Optimization of a Novel Hybrid Rotary Table

MA Jinkui TIAN Zaihao LU Changhou CHEN Shujiang NIE Yulong

2018, 46(12):  74-83.  doi:10.3969/j.issn.1000-565X.2018.12.010

Asbtract ( 328 )   PDF (2517KB) ( 131 )  

References | Related Articles | Metrics

Based on a novel hybrid rotary table, the load carrying property was analyzed theoretically and experimentally, and an optimization model of geometric parameters of the spiral groove with the objective function of hydrodynamic load carrying capacity was proposed. For starters, the geometric model of the friction pair was established and the governing equation was solved numerically to obtain the pressure distribution and lubrication performance. Experiments were also carried out to verify the validity of the simulation. The hydrodynamic load carrying capacity was set as the objective function and spiral angle, wedge depth and wedge number were determined as design variables after analyzing the influence of wedge geometric parameters on the objective function to establish the optimization model. In order to seek the optimum combination of these three variables, Cuckoo Search was applied to solve the optimization model and the law of the optimal solution under various working conditions was obtained. Through comparison with optimized performance from the original performance, dynamic performance of the optimized rotary table was found to be improved, which proved the reliability of the optimization method.

Integrated Design Cloud Platform for Automobile Manufacturing System Based on Internet of Things#br#

LIU Shihao LI Binyong

2018, 46(12):  84-92.  doi:10.3969/j.issn.1000-565X.2018.12.011

Asbtract ( 191 )   PDF (1890KB) ( 107 )  

References | Related Articles | Metrics

In order to improve the intelligent level of automobile manufacturing system, an integrated design method for automobile manufacturing system based on internet of things and cloud computing is proposed. In according to the proposed method, an internet of things-based integrated design cloud platform for automobile manufacturing system is constructed, in which the user's new design requirements can be real-time processed. In the process of on-line monitoring of the automobile manufacturing system, the data collection and transmission is carried out by using the Internet of Things. The state data of the system is analyzed under the cloud computing platform, then the fault and design defects of the automobile manufacturing system is found, which provides a targeted improvement strategy for the conceptual design and detailed design of the system. In the automobile manufacturing system’s conceptual design stage, the fuzzy integral method is used to solve the multi-index integrated decision-making problem in order to obtain the optimal concept design scheme. In the automobile manufacturing system’s detailed design stage, the optimal topology and its parameters of the system is solved by genetic algorithm and cloud computing technology, and the automobile manufacturing system is online improved on the basis of the obtained design results. The use of cloud computing and online improvement method is conducive to improving the intelligent level of automobile manufacturing system. Taking the proposed method into practical application, the performance index optimization problem of an automobile intelligent manufacturing system is solved under the new processing requirements of a batch of automobile crankshaft, which makes the processing speed and product yield of the automobile crankshaft improved largely. After the automobile intelligent manufacturing system is improved, the average machining speed of the crankshaft for automobile is 4 piece/h and the product qualification rate is 97%.The research shows that the proposed integrated design method for automobile manufacturing system based on internet of things and cloud computing has high engineering application value.

Dynamic Modeling Method for Bracket Type Parallel Posture Adjustment Mechanism with Actuation Redundancy

WANG Zhihao CHEN Wenliang WANG Min LIU Wei

2018, 46(12):  93-102.  doi:10.3969/j.issn.1000-565X.2018.12.012

Asbtract ( 340 )   PDF (2008KB) ( 122 )  

References | Related Articles | Metrics

Bracket type posture adjustment mechanism is widely used in aircraft assembly field to adjust the position and posture of the work-piece. Due to the large mass and inertia of the bracket, the kinematics control is difficult and dynamic control needs to be introduced. Independent variables of the posture parameters of the bracket type parallel posture adjustment mechanism in the posture reverse solution and the independent variables of the driving parameters in the posture positive solution were determined. Based on the independent variables, the positive and inverse solutions of velocity and acceleration were analyzed. Newton-Euler method was used to model the dynamics of a new type of three-degree-of-freedom redundantly actuated automatic drilling and riveting bracket type parallel posture adjustment mechanism. Combined with the analysis of the deformation coordination of the telescopic barrel, the posture error and structural error analysis of the posture adjustment mechanism, the dynamic model also took into account the influence of the sliding pair friction of the positioner. Then, the supplementary equations related to the force of the spherical joint, the deformation of the top of the telescopic cylinder, the position and posture error of the posture adjustment mechanism and the position error of the center point of the spherical joint was established. It increased the number of constraint equations in the dynamic model. The dynamic model is mutual verified by Matlab and Adams. Compared with the existing dynamic model, the precision of driving force solution of the dynamic modeling method is more accurate. It proves the validity of the modeling method and can be applied to other types of dynamic control of parallel mechanisms based on positioners.

Traffic & Transportation Engineering

Study on the Vertical Vibration Characteristics of Steel Spring Floating Slab Track in Subway

LI Xiang ＲEN Zunsong

2018, 46(12):  103-110,120.  doi:10.3969/j.issn.1000-565X.2018.12.013

Asbtract ( 318 )   PDF (2440KB) ( 115 )  

References | Related Articles | Metrics

Abstract: In the light of the fact that the serious track vibration noise incidents and abnormal rail corrugation were observed in some subway lines, the research on the reason of rail corrugation was carried out. Using multi-body dynamics simulation software SIMPACK to build dynamic model of subway vehicle and track structure, the response of vehicle-track system dynamic performance and vibration characteristics of the flexible rail system was studied on rail corrugation formation. The results show that changing the track parameters such as track plate length, thickness, steel spring vertical stiffness and span, the natural frequencies of track plate all vary, but the vibration characteristics of it are basically unchanged. There is an obvious resonance phenomenon in the inner rail and track plate at 116 and 161Hz, which may cause inner rail corrugation at wavelength 132 and 95mm. The vertical bending vibration modes and frequency characteristics of the both rail sides and slab track are known to be the main reasons of the initial stages or development of rail corrugation made by the vertical vibration of track system. Which also provides a relevant thought and method about rail corrugation research problem of the steel spring floating slab track.

Numerical Analysis of Corrugated Steel Plate - Composite Beam Structure Based On Equivalent Stiffness Method

LI Baijian ZHU Liangsheng FU Xinsha

2018, 46(12):  111-120.  doi:10.3969/j.issn.1000-565X.2018.12.014

Asbtract ( 225 )   PDF (4828KB) ( 183 )  

References | Related Articles | Metrics

To explore the performance of Corrugated Steel Plate - Composite Beam Structure, based on an old bridge reinforcement engineering located in Hunan province, an equivalent rectangular beam numerical model was established through equivalent stiffness method, and the method of stress conversion was discussed. The properties of Composite Beam Structure under vehicle loads and construction period were analyzed by using equivalent numerical model. It is found that equivalent rectangular beam model can be used to analyze Composite Beam Structure, stress of equivalent rectangular beam can be converted to stress of Composite Beam Structure by the formula of this paper, stress of most points are the same except for the boundary of structure; the most unfavorable condition occur in the double rear axle symmetry acting on the vault, bending moment and axial force of structure is maximum in this condition, which need to be considered when calculated internal forces of structure under live loads; Composite Beam Structure has high construction speed, about 4~5 months construction period, the most unfavorable situation during the construction is the eccentric loading situation during backfilling, moment of structure is maximal and axial force is minimal at this time. At the end of construction, axial force of structure is maximal. These two conditions during the construction process cannot be ignored; CHBDC method is not suitable to calculate internal forces of Composite Beam Structure, which should be analyzed by other methods.

Segmentation and Quantitative Analysis of Corrosion Images Based on Deep Neural Networks

WANG Dalei PENG Bo PAN Yue CHEN Airong

2018, 46(12):  121-127,146.  doi:10.3969/j.issn.1000-565X.2018.12.015

Asbtract ( 496 )   PDF (1985KB) ( 320 )  

References | Related Articles | Metrics

A novel detection method based on deep neural networks is proposed in this article to handle the tough problem of standardizing and quantifying in corrosion detection, and detection as well as quantitative analysis is achieved through semantic segmentation on corrosion images. The “encoder-decoder” architecture is selected in designing deep neural networks, which divides the process of computation into encoding part (downsampling) and decoding part (upsampling), and a segmentation mask with the same resolution as the input image is obtained after computation which indicates whether a pixel belongs to corrosion. Sutong bridge corrosion dataset including 440 corrosion images with resolution of 709×1067 is used to train the segmentation network, and raw images are augmented to 6156 manual labeled images with resolution of 256×256. The training process took about 7 h containing 50 epochs with binary accuracy of 92.55% in training set and 90.56% in validation set. Besides, segmentation network is also applied on raw images and the detection results show that majority of corrosion can be recognized. Corrosion area, corrosion rate and total corrosion rate are defined to quantitatively analyze the corrosion regions, and these indices can be directly calculated with segmentation mask which provide data support for daily maintenance for steel structures.

Optimization Model of Recharging stations for electric vehicles with battery swapping

ZHANG Yong GU Tengfei

2018, 46(12):  128-138.  doi:10.3969/j.issn.1000-565X.2018.12.016

Asbtract ( 270 )   PDF (2869KB) ( 126 )  

References | Related Articles | Metrics

In order to realize rapid recharge for electric vehicles(EV), the locations and batteries holdings of recharging stations for electric vehicles on city road network are studied with battery swapping charging mode. Therefore, the path choice of electric vehicles traveling to recharging stations is firstly considered, the optimal model for recharging stations to minimize total social cost was established, which solves optimal locations of the recharging station and EV's demand to swapping batteries; and then the delay cost of user to swap battery was obtained based on live-death model of EVs and fully-charged batteries at recharging station, thereby establishing a battery holding's optimization model to minimize the total social cost at battery swapping station. The research shows that the optimization model of battery-swapping station can achieve the link flow of road network, the demand of battery swapping at stations, optimal totals and locations and construction sequences of stations’. The penetration of EVs has a more complicated impact on the number and locations of swapping stations and the traffic network equilibrium. The increase of swapping stations amount can reduce the travel cost, but the marginal effect is diminishing. As the penetration of electric vehicles increases, the total number of swapping batteries in the road network will increase correspondingly, but not every exchange station will increase the amount of swapping batteries. Increasing the swapping battery holding at swapping station will increase the probability of fully-charged batteries, but the marginal effect is also diminishing. The advancement of charging technology will increase the probability of fully-charged batteries. The research results can provide decision-making basis for the planning and operation of urban swapping stations.

The Shortest Path Algorithm for Large-scale Traffic Network based on Cloud Computing

ZHANG Dongbo LIN Yongjie LU Kai SHOU Yanfang XU Jianmin

2018, 46(12):  139-146.  doi:10.3969/j.issn.1000-565X.2018.12.017

Asbtract ( 244 )   PDF (1332KB) ( 120 )  

References | Related Articles | Metrics

Cloud computing usually provides dynamically scalable and virtualized computing resources through the Internet, and has widely been used in various fields, especially mass traffic processing. The MapReduce parallel programming model is a novel framework that supports the design of cloud computing algorithms, and can invoke clustered computers at different locations to conduct huge data processing. This study proposed a new computing logic on the base of MapReduce, and developed a parallel searching algorithm with subgraph partitions to search the shortest path in large-scale real traffic networks. In tests of the field networks, the proposed method can provide high-quality shortest path searching service within acceptable calculation time.