Table of Content

25 April 2022, Volume 50 Issue 4

Previous Issue    Next Issue

Contents



2022, 50(4):  0. 

Asbtract ( 38 )   PDF (311KB) ( 116 )  

Related Articles | Metrics

Computer Science & Technology

Semantic Textual Similarity Justification based on Multi-Model Ensemble



SU Jindian, HONG Xiaobin, YU Shanshan

2022, 50(4):  1-9.  doi:10.12141/j.issn.1000-565X.210427

Asbtract ( 1924 )   PDF (1147KB) ( 178 )  

Related Articles | Metrics

As the mainstream and typical methods in current natural language processing and artificial intelligence, various pre-trained language models perform differently on the downstream tasks, due to their different language modeling, feature representation, model structure, training tasks and pre-training corpus, et al. In order to better ensemble the knowledge in different pre-trained language models and utilize their learning abilities on the downstream tasks, we propose a multi-model ensemble method MME-STS (Multi-Model Ensemble for Semantic Textual Similarity) for semantic textual similarity justification tasks. The model structure and the corresponding feature representations are presented, and three different ensemble strategies based on average values, full-connected layer training and Adaboost algorithm with respect to model ensemble are also proposed. Experimental results show that MME-STS outperforms significantly over single pre-trained language model-based approaches on the two benchmark datasets of SemEval 2014 task 4 SICK and SemEval 2017 STS-B corpus in terms of Pearson correlation coefficient and Spearman coefficient metrics.

A Survey on Document-Level Relation Extraction



ZHOU Youhua, HUANG Han, LIU Haolong, et al

2022, 50(4):  10-25.  doi:10.12141/j.issn.1000-565X.210152

Asbtract ( 2258 )   PDF (2465KB) ( 443 )  

Related Articles | Metrics

Relation extraction (RE) is one of the most important tasks in information extraction of NLP, the result of RE can be used to downstream missions such as construction of knowledge graphs, knowledge base question answering, semantic search et al. which means RE has wide-ranging application scenarios and important research value. Recent years, RE achieves frutiful results, but most of them are limited in sentence-level RE, which focus on extract relation between two mentions within a single sentence. Reserches shows that a large number of relations can’t extract from a single sentence, in rencent years, document-level RE faces new opportunities and challenges with the development of deep learning and NLP. This study reviews the recent advances in document-level RE research, summarize a general technology roadmap of this task, and then analyzes the encoding and aggregation methods used in the researches, We also introduce the common datasets and evaluation metrics of this task. This paper ends up with forecasting the future development trend of this task.

An Imbalanced Classification Method based on Adaptive Sampling



CHEN Qiong XIE Jialiang

2022, 50(4):  26-34,45.  doi:10.12141/j.issn.1000-565X.210267

Asbtract ( 1661 )   PDF (2050KB) ( 62 )  

Related Articles | Metrics

In view of the problem that traditional resampling methods mostly use fixed sampling strategies and cannot change the sampling strategy according to the optimization requirements of the model, this paper proposes an adaptive sampling-based imbalanced classification method (Adaptive Sampling Imbalanced Classification, ASIC). This method dynamically adjusts the sampling probabilities of samples of different classes on the training set according to the performance of the classification model on the validation set, so that the sampling probabilities of different classes are dynamically determined by the requirements of the current classification model. At the same time, this method pays extra attention to the minority classes, and gives the minority classes a higher sampling probability under the same other conditions, so as to compensate for the negative impact of the insufficient example number of the minority class itself on the classification model, thereby improving the classification model's ability to recognize minority classes. The experimental results show that the classification model trained with the ASIC method is better than the comparison methods in terms of balanced accuracy and geometric mean, and the more imbalanced the data distribution, the more obvious the superiority of the ASIC method.

Deep Multi-scale Attention Hashing for Large-scale Image Retrieval



FENG Hao WANG Nian TANG Jun

2022, 50(4):  35-45.  doi:10.12141/j.issn.1000-565X.210268

Asbtract ( 622 )   PDF (4771KB) ( 106 )  

Related Articles | Metrics

Aiming at the limited feature extraction capability and inefficient quantization constraint mechanism of existing hashing algorithms, a deep multi-scale attention hashing network was proposed for large-scale image retrieval. The whole network was composed of a main branch and a object branch. In the main branch, two modules of multi-scale attention localization and saliency region extraction were added to effectively localize and extract saliency regions of images, and the results were fed into the object branch to learn more detailed features. Subsequently, the multi-granularity features learned by two branches were fused to perform binary hash coding. In addition, a triplet quantization constraint was introduced to reduce quantization error while maintaining the similarity relationship between sample pairs. In order to verify the effectiveness of the proposed method, extensive experiments were carried out on two benchmark datasets. Experimental results showed that our method outperforms most existing hashing retrieval approaches.

Approach Of Width Measurement Of Irregular Ultra-Thin Heat Pipe Based On Machine Vision



LI Yong, SHE Peijian, OU Qiguan, et al

2022, 50(4):  46-55.  doi:10.12141/j.issn.1000-565X.210372

Asbtract ( 680 )   PDF (5223KB) ( 76 )  

Related Articles | Metrics

In order to realize the automatic measurement of width of irregular ultra-thin heat pipe, a pixel-by-pixel width search algorithm based on the central axis gradient of machine vision is proposed. First, the Canny algorithm is used to obtain pixel edges, and the improved quadratic interpolation method is further used to extract sub-pixel edges. Then, the thinning algorithm is used to extract the central axis of the enclosed area within the edge and crop it appropriately. Next, to construct the edge of the central axis area by morphological expansion and Gaussian filtering algorithm, to calculate the gradient angle, to approximate the smooth change of the true gradient angle with the help of two-sided mean filtering. Finally, searching for sub-pixel edge points along both sides of the gradient angle, and the edge point pair distance is the target width. It solves the problem that the conventional measurement algorithm cannot determine the width measurement point corresponding to the irregular deformation contour and the high-order transition curve area cannot be measured, and realizes the dense measurement of the width in the axial direction, which approximates the true edge width distribution to the greatest extent. The experimental results show that when the width of measured object is within 10 mm, the measurement uncertainty of the algorithm in the article is ±0.026 mm, which can well adapt to the irregular shape of the ultra-thin heat pipe, and realize stable, reliable, and high-precision automated measurement.

Electronics, Communication & Automation Technology

Energy Efficient Resource Allocation for Multicell Full-Duplex Distributed Antenna Systems



LIU Zhan FENG Suili

2022, 50(4):  56-64.  doi:10.12141/j.issn.1000-565X.210355

Asbtract ( 787 )   PDF (1440KB) ( 134 )  

Related Articles | Metrics

In this paper, we address the problem of energy efficient resource allocation in multi-cell full-duplex distributed antenna systems (DAS). Considering the user's quality-of-service (QoS) requirements and per-RAU and per-user power constraints, we jointly optimized the uplink and downlink beamforming factors, user transmit power, and remote antenna unit (RAU) selection to maximize system energy efficiency. Firstly, the objective function is converted from fractional order form to subtraction form through fractional programming, and an iterative resource allocation algorithm is proposed, which needs to solve internal problems in each iteration. Then, using hierarchical decomposition and successive convex approximation, the internal non-convex problem is transformed into a series of iterative convex optimization problems. Simulation results show that the proposed algorithm can significantly improve energy efficiency.

Research on Survivability Technology for Software Defined Network Based on Q-Learning Algorithm



WANG Weifa, ZHANG Daming, DAI Yi, et al

2022, 50(4):  65-72.  doi:10.12141/j.issn.1000-565X.210373

Asbtract ( 632 )   PDF (2040KB) ( 63 )  

Related Articles | Metrics

Aiming at the link invulnerability problem of software-defined networks, in order to make data transmission more robust, a destructive strategy based on Q learning algorithm is designed. This strategy chooses to use the interruption probability of each link in the network as a measure The indicator, through the Q learning algorithm, finds a path with low interruption probability as the backup path according to the network situation, so that when the network transmission fails, the backup path can be automatically switched to improve the anti-destructive performance. The Q learning algorithm is compared with the existing algorithm, and the advantages and disadvantages of each are analyzed. The experimental simulation results show that compared with the ant colony algorithm, the average throughput of the Q learning algorithm can be increased by 15%, and the average network transmission interruption probability can be reduced by 38%; compared with the shortest path algorithm (with backup), the average throughput should be improved 16.5%, the interruption probability of network transmission is reduced by 43% on average. It can be seen that the survivability technology based on Q-learning proposed in this article can greatly improve the survivability of SDN networks.

Non-rigid Point Set Registration Using Neighborhood Structure and Driving Force Criterion



HE Kai, LIU Zhiguo, LI Dashuang, et al

2022, 50(4):  73-80.  doi:10.12141/j.issn.1000-565X.210313

Asbtract ( 687 )   PDF (12082KB) ( 94 )  

Related Articles | Metrics

Finding the correct correspondence is the key point to non-rigid point sets registration. Traditional point set registration methods usually produce mismatches due to only selecting the global spatial distance of features as the criterion but ignoring the neighborhood structure information. To solve this problem, a non-rigid point sets registration algorithm is proposed using neighborhood structure and driving force criterion. A method on the basis of the consensus point drift (CPD) algorithm is proposed to calculate the local mixing distance. Combining it with the original space distance is helpful to improve the matching precision. Besides, a new driving force criterion is constructed based on the improved shape context, which is helpful to improve the searching speed in the original matching process and decrease the matching error in the later process. Finally, we solve the correspondence of each point using the Expectation Maximization (EM) algorithm. Experimental results on commonly-used international point set datasets demonstrate that our method surpass the state-of-the-art ones in terms of robustness and accuracy when deformations, noises, outliers or occlusions exist. Moreover, the proposed algorithm can also achieve ideal registration results on the real images.

A Deep Neural Network Based on Layer-by-Layer Fusion of Multi-Scale Features for No-Reference Image Quality Assessment



YANG Chunling YANG Yajing

2022, 50(4):  81-89,141.  doi:10.12141/j.issn.1000-565X.210231

Asbtract ( 679 )   PDF (2937KB) ( 108 )  

Related Articles | Metrics

The existing deep network-based no-reference image quality assessment algorithms for authentic distortions have poor representation ability for the quality of natural scene images, limiting their evaluation accuracy and generalization ability. To solve this problem, this paper proposes a deep neural network for no-reference image quality assessment to fuse multi-scale features layer-by-layer (MsFF-net). The pre-trained ResNet50 is first used to extract multi-scale features of the image. Then, a multi-scale features fusion module is proposed, which gradually fuses adjacent-scale features layer by layer to obtain multi-scale fused features that can accurately represent the image quality. The low-dimensional features are further extracted from the multi-scale fused features to obtain multi-granularity image quality perception features. Finally, regression is performed on the low-dimensional features by using a fully connected network which is adaptively generated by the highest-level features. The simulation results show that MSFF-net outperforms most of the current methods on authentic distortions databases, and it achieves competing performances on synthetic distortions databases.

An Improved Tomographic Reconstruction of Phase Objects Based on High-order Transport of Intensity Equation



CHENG Hong, ZHANG Xiaolong, LIU Yong, et al

2022, 50(4):  90-99.  doi:10.12141/j.issn.1000-565X.210348

Asbtract ( 829 )   PDF (10528KB) ( 157 )  

Related Articles | Metrics

Classical tomography method usually only considers the intensity reconstruction of the object, and often loses the phase, but the phase contains the information of the depth, shape, refractive index and so on of the surface of the object, which is more important than the intensity. As a classical phase retrieval algorithm,the Transport of Intensity Equation can directly calculate the phase from the known intensity. In this paper, a new algorithm can reconstruct the phase information of objects by tomography.The proposed algorithm is improved based on the high-order Transport of Intensity Equation to obtain higher precision phase results, and then reconstruct the three - dimensional phase information of the object through filtered back projection. Adding rotation center correction steps to the back projection reconstruction process makes the results more accurate. The experimental results show that the algorithm can obtain high precision three-dimensional phase reconstruction results while solving the low phase accuracy caused by intensity differential approximation constraints.

Multicarrier Noise Reduction DCSK Scheme with No Intra-signal Interference



ZHANG Gang DONG Jiangtao HE Lifang

2022, 50(4):  100-109.  doi:10.12141/j.issn.1000-565X.210177

Asbtract ( 643 )   PDF (2690KB) ( 72 )  

Related Articles | Metrics

Aiming at the dominant drawbacks of traditional multicarrier chaotic keying system relating to poor bit error rate and low information transmission rate, a multicarrier noise reduction differential chaos shift keying scheme with no inter-signal interference is proposed in this paper. The designed orthogonal chaotic signal generator is used to obtain two chaotic carrier signals with orthogonal time domain, thus achieving the purpose of no inter-signal interference. Meanwhile, the introduction of multiple predefined subcarriers and orthogonal modulation techniques further increases the transmission rate of the system. The received signal is averaged for noise reduction using a moving average filter at the receiver, which improves the BER performance of the system. Theoretical BER formulas are derived and Monte Carlo simulation experiments are performed under AWGN channels and multipath Rayleigh fading channels. The research results show that the simulation results are basically consistent with the theoretical analysis, and the system has strong anti-noise performance and multipath interference resistance; compared with the traditional MC-DCSK system, the transmission rate can be increased by four times under the same carrier number, and the SNR can be improved by nearly 3 dB under the same BER performance.

Chemistry & Chemical Engineering

Study On the Effect Of Humic Acid Loading On The Adsorption Of Zn(Ⅱ) From Aqueous Solution By Palygorskite



TANG Yuting DING Sichun HAN Chenglin

2022, 50(4):  110-118.  doi:10.12141/j.issn.1000-565X.210455

Asbtract ( 705 )   PDF (2837KB) ( 66 )  

Related Articles | Metrics

There are strict regulations on the discharge of Zn(Ⅱ) wastewater in China. Among adsorptive material that can efficiently remove Zn(Ⅱ) from wastewater, palygorskite (PAL) is considered to be a promising adsorptive material. To verify the effect of humic acid (HA) on the adsorption of Zn(Ⅱ) by PAL, static adsorption experiments were carried out to compare the adsorption rates of Zn(Ⅱ) by humic acid/palygorskite (HA/PAL) and pure PAL, and the performance and mechanism of removing Zn(Ⅱ) from aqueous solution by PAL and HA/PAL were studied. The adsorption equilibrium time of PAL was 120 min, and temperature had little effect on the adsorption process, and the adsorption rate of Zn(Ⅱ) by PAL increased with the increase of the dosage and the decrease of the Zn(Ⅱ) initial concentration. The adsorption equilibrium of HA/PAL was reached at 120 min, and the adsorption process of Zn(Ⅱ) by HA/PAL was also not affected by temperature, and the increase of the dosage and the decrease of the Zn(Ⅱ) initial concentration improved the adsorption rate of Zn(Ⅱ) by HA/PAL. The adsorption process of Zn(Ⅱ) by PAL fitted the Langmuir model, and the pseudo-first-order kinetic model could better fit the experimental data of PAL. However, the adsorption process of Zn(Ⅱ) by HA/PAL was more consistent with Freundlich model, and the pseudo-first-order kinetic model also better fitted the experimental data of HA/PAL. According to thermodynamic parameters, the adsorption of Zn(Ⅱ) by PAL and HA/PAL is endothermic. Under the same conditions, the adsorption rate of Zn(Ⅱ) by PAL is higher than that of Zn(Ⅱ) by HA/PAL, and the influence of all experimental conditions on PAL is greater than that of HA/PAL. The characterization results of SEM, BET and FTIR showed that HA was successfully loaded on PAL surface and reduced PAL specific surface area.

Study on Purification of Kapok Flavonoids by Macroporous Resin and Its Anti-acne Efficacy



ZHENG Bisheng, WEI Ｒuijing, GUO Chaowan, et al

2022, 50(4):  119-128.  doi:10.12141/j.issn.1000-565X.210383

Asbtract ( 690 )   PDF (2620KB) ( 82 )  

Related Articles | Metrics

Study the extraction process of kapok flavonoids and evaluate its anti-acne efficacy. Using Guangdong Zhaoqing kapok as raw material, after ultrasonic-assisted alcohol extraction, the most suitable macroporous resin type was investigated and the purification process of the crude kapok flavonoids was determined. The in vitro inhibition zone experiment method was used to determine the purified and refined kapok flavonoids. The inhibitory effect of Staphylococcus epidermidis and Propionibacterium acnes; the in vitro whitening activity method was used to determine the inhibition rate of kapok flavonoids on tyrosinase, and the whitening activity of kapok flavonoids was further verified by the zebrafish embryo melanin inhibition test; Finally, the in vitro antioxidant activity determination method was used to investigate the scavenging ability of kapok flavonoids on free radicals such as ABTS?+ and DPPH. By comparing the adsorption rate and resolution rate of 8 kinds of macroporous resins to kapok flavonoids, D101 was screened out as the best resin type, and the best static adsorption desorption conditions were investigated as follows: adsorption time 6 h,resolution time 4 h, the ethanol concentration of the analytical solution is 60%; the best dynamic adsorption analysis conditions are: loading volume 2 BV, analytical solution elution volume 0.625 BV. After purification, the purity of kapok flavonoids increased from 32.5% to 98.2%. The IC50 values of the purified kapok flavonoids on ABTS?+ and DPPH free radical scavenging rates were 0.62±0.011 mg/mL and 11.8±0.624 μg/mL, respectively.The IC50 of the inhibition rate of Kapok flavonoids on tyrosinase is 1.69±0.002 mg/mL. When the concentration of kapok flavonoids was 1.0 mg/mL, the inhibition rate of melanin in zebrafish embryos was 12%, and it had significant whitening activity.Through the inhibition zone test, when the concentration of kapok flavonoids is 12.5 mg/mL, it is moderately sensitive to Propionibacterium acnes and highly sensitive to Staphylococcus epidermidis, and this antibacterial effect presents a dose-dependent effect.In summary, the kapok flavonoids purified and prepared in this study have multiple skin care activities such as whitening, antioxidation, antibacterial and anti-acne, and have good application and development value for anti-acne products.

Study of Vulnerability of Chemical Storage Tanks Subjected to Coupling Effect of Blast Wave and Fragment



ZHAO Jie LAI En CHEN Guohua

2022, 50(4):  129-141.  doi:10.12141/j.issn.1000-565X.210487

Asbtract ( 973 )   PDF (15019KB) ( 282 )  

Related Articles | Metrics

Studying vulnerability of storage tanks under the coupling effect of blast wave and fragment can contribute to improve safety level of chemical industrial parks. In the study, the failure limit state equation of chemical tank that subjected to the coupling effect is established based on the energy density rule and maximum plastic strain criterion, which considering the sequential action of blast wave and fragment and the strain-rate effect. Comparative analysis of the theoretical rupture velocity of the storage tank and the maximum speed that obtains by numerical simulation to verify the rationality of the equation. The analysis of the vulnerability of a 5000m3 vertical vault tank shows that the strain-rate effect can enable the tank less destruction of the combined loadings and reduce the vulnerability of the tank. The action sequence of blast wave and fragment significantly affects the failure probability of the tank, while fragment impacts the tank firstly is more dangerous than blast wave. Through a single univariate analysis shows that the vulnerability of the storage tank is positively correlated with explosive strength and impact loading of fragment, but it is negatively related to the impact angle of fragment. The study results are helpful for providing guidance for the prevention of domino effect accident subject to coupling effect of blast wave and fragment and formulating pre-control strategy.

The influence of pulse vibration molding on the structure and performance of UHMWPE hollow parts



FENG Yanhong WANG Ｒuisong YIN Xiaochun ZOU Xinliang LIN Cheng

2022, 50(4):  142-148.  doi:10.12141/j.issn.1000-565X.210110

Asbtract ( 639 )   PDF (3578KB) ( 82 )  

Related Articles | Metrics

In order to improve the mechanical properties and wear resistance of ultra-high molecular weight polyethylene (UHMWPE) hollow parts, a self-developed hollow part forming mold was used, and pulse vibration was introduced during the compression molding process, so that the material can be formed under pulse pressure. The effect of impulsive vibration on the microstructure and macroscopic properties of the part was studied. Through tensile test, mortar abrasion test, DSC test, density test and 2D-WAXD test, the results show that the application of pulse vibration can increase the density, orientation and crystallinity of the sample, and its tensile properties and wear resistance are better than those molded under static conditions.