Table of Content

25 June 2022, Volume 50 Issue 6

Previous Issue    Next Issue

Contents



2022, 50(6):  0-0. 

Asbtract ( 151 )   PDF (386KB) ( 166 )  

Related Articles | Metrics

Computer Science & Technology

A cross-modal face retrieval method based on metric learning



WO Yan, LIANG Jiyun, HAN Guoqiang

2022, 50(6):  1-9.  doi:10.12141/j.issn.1000-565X.210709

Asbtract ( 2768 )   PDF (2022KB) ( 649 )  

Related Articles | Metrics

Metric learning is an important technique to reduce modal differences. Existing cross-modal retrieval methods based on metric learning for cross-modal face retrieval tasks lack attention to pose differences and domain differences, and there are two problems in the process of metric learning: lack of learning of global information and the existence of a large number of redundant triplets. In this paper, a cross-modal common representation generation algorithm based on metric learning is proposed. Our study uses the yaw angle equivariant module to compensate for yaw angle differences so that we can obtain the image features with robustness, uses the multi-layer attention mechanism to obtain video features with differentiability; combines global triplets and local triplets to jointly train the cross-modal common representation generation network, then accelerates the convergence of the loss function through the screening of semi-hard triplets; combines domain calibration and transfer learning to improve the generalization of common representations. Finally, the results of comparison experiments on three face video datasets: PB, YTC and UMD Faces, demonstrate that our algorithm can improve the accuracy of cross-modal face retrieval, and the results of fine-tuning the cross-modal common representation generation network using different numbers of samples demonstrate that our algorithm can improve the accuracy of cross-modal retrieval of target domain images.

Image tampering localization based on mutil-scale transformer



LU Lu, ZHONG Wenyu, WU Xiaokun

2022, 50(6):  10-18.  doi:10.12141/j.issn.1000-565X.210603

Asbtract ( 2098 )   PDF (2425KB) ( 233 )  

Related Articles | Metrics

With the continuous development of digital image processing technology, image tampering is no longer limited to a single method such as image splicing. The traces of malicious tampering are hidden through the post-processing of the image editing software, which leads to poor results of traditional image forgery detection algorithms and the tampering localization methods based on deep learning. Aiming at the problem of low accuracy of existing image tampering algorithms, an end-to-end image tampering location network based on Multi-Scale Visual Transformer is proposed. The network combines a transformer and a convolutional encoder to extract the feature difference between the tampered area and the non-tampered area. Multi-Scale Transformer models the spatial information of image block sequences of different sizes, so that the network can adapt to tampered areas of various shapes and sizes. Experimental results show that the F1 and AUC scores of the proposed algorithm in the CASIA and NIST2016 test sets are 0.431、0.877、0.728 and 0.971, respectively, which are significantly improved co- mpared to the existing mainstream algorithms. Moreover, the algorithm proposed in this paper is robust against JPEG compression attacks.

Injection molding part size prediction method based on Stacking integration learning



SONG Jian, WANG Wenlong, LI Dong, et al

2022, 50(6):  19-26.  doi:10.12141/j.issn.1000-565X.210664

Asbtract ( 1649 )   PDF (4236KB) ( 109 )  

Related Articles | Metrics

Machine learning algorithms can handle high-dimensional and multivariate data and extract hidden relationships in data in complex and dynamic environments, which has a good application prospect in injection molding part size prediction. The performance of injection molding part size prediction system depends on the choice of machine learning algorithm, however, the traditional machine learning algorithm can not achieve good prediction effect in practical application. In this paper, a fusion model based on Stacking integrated learning method is proposed, and the optimal performance model is obtained by comparing different Stacking learner combinations and combining multiple types of learners. The performance of the model in injection molding parts size prediction is greatly improved compared with the traditional model, and the model prediction results can be explained back to the actual production according to the characteristics, providing decision guidance for the optimization of manufacturing processes and processes.

Quadrotor Trajectory Planning Method Based on Local Soft-Constrained Optimization



CHEN Peng, JIANG Yongqi, YU Tianwei, et al

2022, 50(6):  27-36.  doi:10.12141/j.issn.1000-565X.210452

Asbtract ( 1623 )   PDF (6172KB) ( 71 )  

Related Articles | Metrics

In order to improve the efficiency of the quadrotor long-distance trajectory planning in 3D complex scene, this paper proposes a real-time quadrotor trajectory planning method based on local soft-constrained optimization. The method can be divided into the following two steps: Firstly, the safety distance constraint is added to theta* algorithm, and the heuristic function is improved by using the turning cost to reduce the time consumption caused by quadrotor turning, and finally the initial path composed of a small number of key points is generated; Secondly, the local optimization strategy is used to optimize the segments with potential safety hazards in the initial path based on soft constraints, and the Hodograph property of Bézier curve are used for time allocation to ensure the continuity, smoothness and dynamic feasibility of the trajectory and improve the flight efficiency of quadrotor. Experimental results show that the proposed method has shorter flight distance and flight time and higher planning efficiency while ensuring the safety of quadrotor. This method can also be successfully applied to the actual quadrotor flight.

Product Feature Extraction Method Based on Seed Constraint-LDA



CHEN Kejia, ZHENG Jingjing

2022, 50(6):  37-48,70.  doi:10.12141/j.issn.1000-565X.210124

Asbtract ( 1525 )   PDF (1892KB) ( 60 )  

Related Articles | Metrics

In order to classify and extract product features from reviews, make reviews displayed separately according to different product features, and improve the efficiency of making purchasing decisions for consumers, this paper proposes a product feature extraction method based on SC-LDA（Seed Constraint-Latent Dirichlet Allocation). Firstly, the TF-IDF (Term Frequency–Inverse Document Frequency) algorithm is used to automatically extract the keywords as a feature seed set. Secondly, document reorganization is adopted to solve the problem of multi-feature co-occurrence of the long text as well as sparsity of the short one and improve the rate of document reorganization. Then, must-link and cannot-link seed constraints are applied to define the probability expansion and contraction value, which affects the topic allocation of the LDA model and makes the training results more reasonable. Finally, the topics generated by SC-LDA are mapped to the prior feature categories. The advantages of the proposed method are verified by carrying out qualitative analysis in terms of feature categories as well as feature words and quantitative analysis in terms of accuracy, entropy as well as purity.

Electronics, Communication & Automation Technology

Particle filter method based on hybrid multi-strategy optimization



WEN Shangsheng, XU Hanming, CHEN Xiandong, et al

2022, 50(6):  49-59.  doi:10.12141/j.issn.1000-565X.210540

Asbtract ( 517 )   PDF (9574KB) ( 71 )  

Related Articles | Metrics

The standard particle filter has long problems referred as sample degeneracy and impoverishment. Requiring large number of samples to achieve suitable estimation accuracy, which reduces the comprehensive performance of the algorithm. This paper proposes a hybrid multi-strategy optimization particle filter method based on Levy flight strategy, differential evolution and success-history strategy. The Levy flight strategy enricifies the basic framework of the sample set, ineffective particles with low-weight are optimized through the differential evolution algorithm, and successful history strategy is used to adjust the parameters to achieve a balance between the global search and the local search, so as to prevent particles from falling into the local optimum when the motion scale is too large. Experiments show that the proposed algorithm can effectively improve the particle diversity, accuracy and sample degeneracy under low measurement noise, reducing the number of particles needed to estimate nonlinear systems.

Real Time Statistics Method of Escalator Passenger Flow for Embedded Devices



DU Qiliang, XIANG Zhaoyi, TIAN Lianfang

2022, 50(6):  60-70.  doi:10.12141/j.issn.1000-565X.210389

Asbtract ( 650 )   PDF (3571KB) ( 81 )  

Related Articles | Metrics

In order to address the problem that the accuracy and speed of the traditional statistical calculation method are difficult to balance, we design a real time method of escalator passenger flow statistics for embedded devices. Firstly, a distortion-free scaling method is proposed to maintain the consistency of information between the test and the training sample to avoid affecting the performance of the detection model; furthermore, the YOLOv4-tiny detection model is optimized by a dimensionality reduction module and group convolution, and then a YOLOv4-tiny-fast network is proposed, which significantly reduces the number of parameters and improves the inference speed while ensuring no loss of passenger detection accuracy; finally, a matching algorithm combining custom optimization matrix and occlusion processing is proposed to solve the passenger tracking problem with less computational effort. To demonstrate the effectiveness of the method, experiments are conducted with video of escalator entrances and exits in a real environment. The results show that the proposed algorithm achieves an average accuracy of 96.66% in passenger flow statistics on the embedded device platform, and the average detection speed reaches 25 frames/s, which is better than existing algorithms.

Incremental learning based on neuron regularization and resource releasing



MO Jianwen, ZHU Yanqiao, YUAN Hua, et al

2022, 50(6):  71-79,90.  doi:10.12141/j.issn.1000-565X.210404

Asbtract ( 619 )   PDF (1989KB) ( 90 )  

Related Articles | Metrics

Aimimg at the catastrophic forgetting problem caused by the image classification of deep learning systems in an incremental scene. A incremental learning method which based on neuron regularization and resource releasing mechanism was proposed. This method is based on the framework of Bayesian neural network. Firstly, grouping the input weights by neurons and restrict the standard deviation of weights to the same value according to the groups. Then in the training process, the corresponding strength regularization was performed for the weights of each group according to the standard deviation after unification. Finally, in order to improve model's continuous learning ability, a resource releasing mechanism was proposed. This mechanism maintains model's learning ability by guiding the model to selectively dilute the regularization strengths of some weights. Experiments on several common datasets show that the proposed method can explore the continuous learning capability of the model more effectively, and a better model can be learned even in a fixed capacity environment.

Feature-domain Multi-Hypothesis Prediction Reconstruction Neural Network for Compressed Video Sensing



YANG Chunling, LING Xi, LÜ Zeyu

2022, 50(6):  80-90.  doi:10.12141/j.issn.1000-565X.210507

Asbtract ( 578 )   PDF (2614KB) ( 65 )  

Related Articles | Metrics

In the prediction-residual reconstruction framework, multi-hypothesis prediction based on temporal correlation is the key step of compressed video sensing reconstruction. This paper studies the accuracy prediction method by utilizing rich features based on deep learning, and a novel feature-domain multi-hypothesis reconstruction network for compressed video sensing (FMH_CVSNet) is proposed. In FMH_CVSNet, the feature domain multi-hypothesis prediction module (FMH_Module) is firstly proposed, which improves the prediction ability by reasonably constructing the motion estimation module and the hypothesis weight calculation module based on the characteristics of video signal. Secondly, the two-stage multi-reference motion compensation mode is proposed, which makes the constructed hypothesis sets much better for sequences with different motion and the further improves the prediction accuracy. Simulation results show that FMH_CVSNet achieves better reconstruction performance under various experimental conditions, improves by 4.76dB compared with the traditional multi-hypothesis algorithm 2sMHR and improves by 3.87dB compared with CNN based compressed video sensing reconstruction algorithm VCSNet-2.

Full Life-cycle Intelligent Detection, Diagnosis and Analysis for Sludge Bulking



LIU Yiqi, HUANG Zhipeng, YU Guangping, et al

2022, 50(6):  91-99,110.  doi:10.12141/j.issn.1000-565X.210561

Asbtract ( 819 )   PDF (1704KB) ( 99 )  

Related Articles | Metrics

As a typical multivariate statistical analysis method，the partial least squares (PLS) has been widelyapplied in the anomaly monitoring of key performance indicators． However，the data in complex industrial processes may show dynamic or nonlinear characteristics，or both，so the linear model based on PLS is not suitable for thisprocess and may increase the false alarm rate． Therefore，a multi-feature extraction algorithm based on PLS was proposed． Firstly，the dynamic features were extracted based on the dynamic internal model of PLS to obtain thequality-related dynamic subspace and dynamic residual subspace． The PLS regression modeling was carried out forthe dynamic residual subspace to further separate the quality related features of static information． Then，the residualsubspace was projected to a high dimension to construct a nonlinear PLS model to extract the nonlinear characteristics of variables． Finally，the statistics was constructed in each feature space，and a complete multi-feature hybrisystem process monitoring strategy was designed． The example analysis results of Tennessee-Eastman process show
that the proposed method has faster fault detection speed and can achieve better fault detection effect．


Research on Multi-characteristic Complex Process Monitoring Method based on Partial Least Squares



KONG Xiangyu, CHEN Yalin, LUO Jiayu, et al

2022, 50(6):  100-110.  doi:10.12141/j.issn.1000-565X.210679

Asbtract ( 385 )   PDF (2235KB) ( 44 )  

Related Articles | Metrics

Partial least squares (PLS) as a typical multivariate statistical analysis method, has been widely used for the monitoring of key performance indicators. However, in the complex industrial processes, the data may exist dynamic, nonlinear features, or both. The linear PLS is not suitable for these cases, and may lead to high false alarm rate and false alarm rate. Therefore, this paper proposes a multi-feature extraction algorithm based on partial least squares for the forementioned complex process. The proposed algorithm extracts dynamic features based on dynamic internal PLS (DiPLS) to obtain quality-related dynamic subspace and dynamic residual subspace; In order to further separate the quality-related features of static information, PLS regression modeling is carried out on dynamic residual subspace; In addition, to extract the nonlinear features of variables, the residual subspace is mapped to a high dimension to construct a nonlinear PLS model; By constructing statistics in each feature space, a complete process monitoring strategy for multi-feature hybrid system is designed. Finally, through the example analysis of Tennessee-Eastman (TE) process, it is verified that the proposed method has faster fault detection speed and can achieve better fault detection effect.

Mechanical Engineering

Intelligent Pushing Method and Experiment of Feeding Assistant Robot



ZHANG Qin, HU Jiahui, REN Hailin

2022, 50(6):  111-120.  doi:10.12141/j.issn.1000-565X.210621

Asbtract ( 2173 )   PDF (11308KB) ( 879 )  

Related Articles | Metrics

Regular feed pushing is an important link in the feeding process of dairy cows, and the intelligent feeding assistance of robot instead of human is becoming the development direction in the future. The feeding robot can complete multiple feeding throughout the day, which is widely used in the feeding of dairy cows in large and medium-sized pastures. However, the existing feeding robot has a single function, which can only complete the uniform feeding function, and can not meet the personalized feeding needs of cattle. To solve this problem, this paper proposes an intelligent pushing method of feeding assistant robot. The QR code label is introduced as the positioning label of cow neck rail, and the detection frame area of QR code label and cow head is obtained based on YOLOv4 deep learning model. The detection frame area of QR code label is recognized and tracked in real time through preprocessing and prediction algorithm, and the QR code label and cow head are matched to determine the position of feeding neck rail; According to the position matching information of cow-code and the residual forage distribution information, the robot push plate was controlled to change the pushing angle to realize personalized pushing and meet the individual feeding needs of dairy cows. Research and test results show that the proposed intelligent push method has a recognition rate of 96% for the QR code; In the case of losing 60 consecutive frames, the tracking and prediction accuracy of the QR code is less than ± 2.85%; The processing time of each frame in GPU is 34.4 ms; The accuracy of intelligent feeding is 100%, which can meet the real-time requirements of intelligent pushing in complex environment.

Mixed-lubrication Analysis of Journal Bearing Considering Wear and Turbulence



SONG Xintao, WU Wei, YUAN Shihua

2022, 50(6):  121-128.  doi:10.12141/j.issn.1000-565X.210536

Asbtract ( 1629 )   PDF (2094KB) ( 282 )  

Related Articles | Metrics

Based on the average flow equation and the Ng-Pan turbulence model, using the boundary condition of mass conserving (JFO), a mixed lubrication model considering wear and turbulence was established. A finite difference method was used to solve the model. Under the condition of constant external load, the effects of wear and turbulence on oil film pressure, oil film thickness and friction coefficient were studied. The results show that wear significantly changes the oil film pressure and oil film thickness distribution, and has an important influence on the Stribeck curve of bearing. Under high-speed operating condition, turbulence reduces the maximum oil film pressure of the bearing, and increases the minimum oil film thickness, cavitation zone, and friction coefficient. Wear and turbulence have a critical impact on the lubrication performance of the bearing, and the two together determine the lubrication performance of the bearing.

Finite element study on the influence of temperature on the scaling effect of specific cutting energy



YE Guigen, LI Xinjian, ZHANG Peng, et al

2022, 50(6):  129-136.  doi:10.12141/j.issn.1000-565X.210524

Asbtract ( 587 )   PDF (3508KB) ( 47 )  

Related Articles | Metrics

The scaling effect of specific cutting energy is one of the most challenging problems in metal cutting. There is no consensus reached on the mechanism of the scaling effect, and the contribution of the cutting temperature on the scaling effect was rarely studied. Here the orthogonal cutting experiments under different initial temperature were conducted. It shows significant dependence of the scaling effect phenomenon on the temperature effect. The scaling effect phenomenon becomes more obvious at lower initial temperature. The finite element model of orthogonal cutting was further developed to study the contribution of temperature on the scaling effect. The results show that, there exists two temperature gradient zones near the ends of the primary shear zone along the shear direction. The length of the gradient zones are independent on the uncut chip thickness. The existence of the temperature gradient zones causes the average temperature in the primary shear zone to drop with decreased uncut chip thickness, which gives rise to the increase of flow stress, leading to the nonlinear increase of the specific cutting energy.

Flow Characteristics Research of Children Upper Airway Based on 3D Printing Technology



CAI Huikun, XUE Haoyang, XU Chang, et al

2022, 50(6):  137-144.  doi:10.12141/j.issn.1000-565X.210496

Asbtract ( 461 )   PDF (7958KB) ( 83 )  

Related Articles | Metrics

Aiming at the problems of limited analysis and difficult validation on flow characteristics of children upper airway, this paper presents an experimental bench using upper airway models manufactured by 3D printing technology to study these problems. The maximum difference between experimental results and clinical nasal resistance test is 11.1%, which proves the experimental method based on 3D printing technology is reliable and accurate, and it can be a direct and easy method for the research of flow characteristics of children upper airway. After that, the effect of different levels of hypertrophy of tonsil on flow characteristics is studied. It is found that, when the narrow level of oropharynx reaches 56%, where the level of hypertrophy of tonsil reaches III degree, flow characteristics of upper airway will obviously change, in terms of increasing flow velocity and resistance, decreasing flow volume, and a large inverse pressure and pressure difference in oropharynx, which can an explanation of the symptoms for OSAHS child, and also will be a significant reference to clinicians on tonsillectomy operation and its postoperative recovery plan.

Experimental Study on Friction and Wear and Efficiency of Water Hydraulic Axial Piston Pump with Biomimetic Non-smooth Surface Slipper Pair



LIANG Yingna, GAO Jianxin, GAO Dianrong

2022, 50(6):  145-154.  doi:10.12141/j.issn.1000-565X.210484

Asbtract ( 2351 )   PDF (52434KB) ( 855 )  

Related Articles | Metrics

In this paper, a certain type water hydraulic axial piston pump was taken as a prototype, and the original smooth surface swash plate was replaced with a biomimetic non-smooth surface one. The flow-pressure, volumetric efficiency-pressure and mechanical efficiency-pressure of the test pump at three different pressures of 7 MPa, 10 MPa, and 12 MPa were tested, and the worn surface of the swash plate was observed using laser confocal microscope and scanning electron microscope. The results show that the slipper pair with non-smooth surface can produce hydrodynamic lubrication effect and have chip holding capacity due to pits. Self-lubrication can be realized during the friction process, achieving the effect of reducing drag and wear. With the increase of working pressure, friction marks in the half-circle high pressure area of the non-smooth surface swash plate are gradually obvious, grooves on the worn surface become wider and deeper, and adhesion wear and oxidation wear are aggravated. That is the friction and wear are aggravated. The volumetric efficiency, mechanical efficiency and total efficiency of the non-smooth surface slipper pair test pump are increased by 0.2%-0.6%, 0.1%-1.7% and 0.1%-2.3% respectively, compared with those of the smooth surface slipper pair test pump.