Table of Contents

30 April 2021, Volume 42 Issue 2 Previous Issue    Next Issue
---
Image Processing and Computer Vision
Contents
Image Processing and Computer Vision
Computer Graphics and Virtual Reality
Digital Design and Manufacture
BIM/CIM
Industrial Design
Total to Discuss


For Selected:

Download Citations EndNote Ris BibTeX

Toggle Thumbnails

Image Processing and Computer Vision

Select

Character region positioning of chip image based on improved convex hull detection CHAO Yuan, ZHOU Wei, SHAN Wen-tao, FENG Jun-ping 2021, 42(2): 165-173.  DOI: 10.11996/JG.j.2095-302X.2021020165 Abstract ( 156 )   PDF (959KB) ( 190 )   During laser printing, the semiconductor chip may suffer many defects, such as character tilt and incorrect character position. To address these defects, a character-region positioning method for multiple semiconductor chip-images was proposed based on improved convex hull detection. Firstly, multiple semiconductor-chip images were obtained using a three-axis image acquisition platform, and several single-chip images were segmented and extracted. Secondly, the Harris corner detection was employed to produce the image corner distribution map, and the convex hull detection algorithm was improved. Then, the corners in non-character regions were eliminated, and the outermost corner convex hull line was acquired. Finally, the minimum enclosing rectangle of the convex hull was fitted, and the character region was positioned. The experimental results reveal that, compared with the morphological filtering positioning method, the positioning method based on edge features, that based on character texture features, and that based on convex hull and the minimum external moment, the proposed method can more accurately position the chip’s character region, and the average running time of the single-chip image is less than half of the above-mentioned methods. Therefore, this proposed method can effectively reduce the computing load and improve the computing efficiency.  Related Articles | Metrics

Contents

Select

Contents of the Second Issue of 2021 2021, 42(2): 164-164.  Abstract ( 277 )   PDF (220KB) ( 52 )   CONTENTS Image Processing and Computer Vision Character region positioning of chip image based on improved convex hull detection ·························································· CHAO Yuan,  ZHOU Wei,  SHAN Wen-tao,  FENG Jun-ping (165) Human action recognition based on LSTM neural network ········································· YANG Shi-qiang,  YANG Jiang-tao,  LI Zhuo,  WANG Jin-hua,  LI De-xin (174) Research on the structured 3D animation and the compression ··· LUO Guo-liang,  WANG He,  ZHAO Xin,  CAO Yi-qin,  HUANG Xiao-sheng,  WU Chang-xing,  XIAN Chu-hua (182) Style transfer algorithm for salient region preservation ··············································· LIN Xiao,  QU Shi-cao,  HUANG Wei,  ZHENG Xiao-mei,   MA Li-zhuang (190) Real-time gun detection method based on compressed YOLOV3 ······················································· YE Ze-cong,  GAO Zhi-qiang,  CUI Xiao-long,  JIANG Rong-qi (198) Image super-resolution reconstruction based on dual attention mechanism ····························································································· LI Bin,  WANG Ping,  ZHAO Si-yi (206) Computer Graphics and Virtual Reality Mixed reality based remote collaborative assembly guidance ····························· LIU You-quan,  WANG Yuan-chao,  XU Kun,  LIU Zheng-xiong,  HUANG Pan-feng (216) Public opinion evolution analysis of “COVID-19 epidemic” based on sentiment feature ····················· GAN Yu-xiang,  WANG Ya-bo,  XUE Jun-xiao,  ZHANG Ruo-qi,  XU Shu-ning,  GUO Yi-bo (222) The construction of spherical interpolation splines with local properties ····························································· WANG Qian,  PAN Le,  ZHANG Jie-lin,  PENG Xing-xuan (230) Compression of directed surface chain code in 3D model ···························································· LIU Shang-wu,  WEI Wei,  DUAN Xiao-dong,  LIU Yong-kui (237) Curve modeling using rational fractal interpolation with variable parameters ···································· ZHANG Xin-yue,  LEI Yi-fan,  LIU Pei-pei,  BAO Fang-xun,  ZHANG Yun-feng (245) Multi-view color point cloud registration based on correntropy ························································· LIU Ying,  LI Yao-chen,  LIU Yue-hu,  FU Jing-wen,  MA Fei (256) Digital Design and Manufacture A light weight method of 3D printing model based on periodic cell ····················································· XIAO Wen-lei,  WANG Zhi-ming,  WANG Shi-ping,  ZHAO Gang (263) APDL modeling method of hydraulic radial gate structure ······································································ ZHANG Xue-cai,  CHEN Li-ye,  WANG Zheng-zhong (271) Simulation analysis of stress in power fitting for strain insulator-string following ice shedding from cables ···································· LI Chun,  LIU Xiao-hui,  HUANG Rui,  YE Zhong-fei,  WU Chuan,  ZHANG Bo (279) DIM/CIM Research on information collaboration and model construction for top-down design of prefabricated buildings ····································································································· ZHU Hui-xian,  XU Zhao (289) BIM evacuation design automation based on artificial intelligence ················ LIANG Yu-qing,  JI Jiu-mao,  YANG Jia-lei,  ZHANG Dong-sheng,  WANG Ke,  WANG Ling-yu (299) Research on named entity recognition of construction safety accident text based on pre-trained language model ··········································································· SONG Jian-wei,  DENG Yi-chuan ,  SU Cheng (307) A model adaptive method for Chinese word segmentation using IFC-based building information model ················································································ ZHANG Xin,  ZHOU Xiao-ping,  WANG Jia (316) Industrial Design Usability evaluation and application of multi-joint active and passive trainer based on user experience ···································································································· REN Ying-li,  WU Shi-jin (325) Research on evolutionary design of interactive interface for knowledge reuse and intelligent innovation ································································································· JING Chun-hui,  ZHI Jin-yi (332) Related Articles | Metrics

Image Processing and Computer Vision

Select

Human action recognition based on LSTM neural network YANG Shi-qiang, YANG Jiang-tao, LI Zhuo, WANG Jin-hua, LI De-xin 2021, 42(2): 174-181.  DOI: 10.11996/JG.j.2095-302X.2021020174 Abstract ( 561 )   PDF (878KB) ( 437 )   Human action recognition provides the basic support for human-computer cooperation. Robots can enhance the flexibility and production efficiency of manufacturing system by recognizing and understanding the operator’s action. To resolve the problem of human motion recognition, the original 3D skeleton data was smoothed and denoised to conform to the smooth rule of human joint-point motion based on 3D skeleton data. The fusion feature composed of static and dynamic features was constructed to represent human action. The key frame extraction model was introduced to extract the key frames in human action sequences to reduce the computing load. A Bi-LSTM neural network model based on LSTM neural network was established to classify human actions, and the attention mechanism and Dropout were utilized to classify and recognize human actions, with the main parameters of the neural network optimized by the orthogonal test method. Finally, the open data set was employed for the action recognition experiment. The results show that the proposed model algorithm has a high recognition rate for human actions.  Related Articles | Metrics

Select

Research on the structured 3D animation and the compression LUO Guo-liang , WANG He , ZHAO Xin , CAO Yi-qin , HUANG Xiao-sheng , WU Chang-xing , XIAN Chu-hua 2021, 42(2): 182-189.  DOI: 10.11996/JG.j.2095-302X.2021020182 Abstract ( 143 )   PDF (1110KB) ( 95 )   With the increasingly wide applications of 3D animation, the compression methods of 3D animation data have become significant for the technologies in data storage, transmission, and rendering. There are multiple efficient technologies for 2D image compression in the domain of computer vision. Based on the structural differences between t3D animation data and 2D image data, the improvement of the image compression algorithm was explored for a new method of 3D animation compression. Specifically, the optimal vertex sequence of the 3D model towards compression was first computed. Then, based on the classical image compression algorithm flow, the sub-block size optimization and the quantization matrix optimization was formulated to improve compression. Finally, with different types of animation, experimental results show that the optimized image compression algorithm can guarantee both effectiveness and efficiency for the compression.  Related Articles | Metrics

Select

Style transfer algorithm for salient region preservation  LIN Xiao, QU Shi-cao, HUANG Wei, ZHENG Xiao-mei, MA Li-zhuang 2021, 42(2): 190-197.  DOI: 10.11996/JG.j.2095-302X.2021020190 Abstract ( 172 )   PDF (15311KB) ( 311 )   Style transfer based on neural networks has become one of the hot research issues in academia and industry in recent years. Existing methods can apply different styles to a given content image to generate a stylized image and greatly enhance visual effects and conversion efficiency. However, these methods focus on learning the underlying features of the image, easily leading to the loss of content image semantic information of stylized images. Based on this, an improved scheme was proposed to match the salient area of the stylized image with that of the content image. By adding a saliency detection network to generate a saliency map of the composite image and the content image, the loss of the saliency map was calculated during the training process, so that the composite image could almost maintain a saliency area consistent with that of the content image, which is conducive to improving the stylized image. The experiment shows that the stylized image generated by the style transfer model can not only produce better visual effects, but also retains the semantic information of the content image. Ensuring the undistorted ness of salient areas is a significant prerequisite for generating visually friendly images, especially for the content image with prominent salient areas.  Related Articles | Metrics

Select

Real-time gun detection method based on compressed YOLOV3  YE Ze-cong, GAO Zhi-qiang, CUI Xiao-long , JIANG Rong-qi, 2021, 42(2): 198-205.  DOI: 10.11996/JG.j.2095-302X.2021020198 Abstract ( 190 )   PDF (3653KB) ( 139 )   The detection of such dangerous targets as guns has always been one of the important research subjects in the field of security. Manual inspection of guns and other dangerous objects through surveillance video is inefficient, the accuracy of which can be easily affected by unfavorable conditions of inspectors due to continuous working. Therefore, a real-time gun detection method using the pruning method to compress the YOLOV3 model was proposed. In order to improve model accuracy, the K-means ++ algorithm was adopted to cluster Anchor size of data set. Then the pruning method of “channel + layer” was employed to compress the trained model. Finally, by means of retraining, the accuracy was restored to that before compression. The experimental results show that while still maintaining high accuracy, this method can reduces not only the model’s occupation of memory resources, but also the computing load, greatly improving the inference speed of the model. Compared with the YOLOV3 method, this method reduces the model parameters by up to 1/52 on the jetson nano platform and increases the reasoning speed by 6 times, with the accuracy almost unchanged, thus meeting the requirements of real-time and high-precision detection of such dangerous objects as guns.  Related Articles | Metrics

Select

Image super-resolution reconstruction based on dual attention mechanism LI Bin , WANG Ping , ZHAO Si-yi 2021, 42(2): 206-215.  DOI: 10.11996/JG.j.2095-302X.2021020206 Abstract ( 207 )   PDF (3047KB) ( 221 )    In recent years, the convolutional neural network (CNN) has achieved desired results in the field of single image super-resolution (SISR). Deep networks can establish complex mapping between low-resolution and high-resolution images, considerably enhancing the quality of reconstructed images, compared with the traditional methods. Since the existing SISR methods mainly increase the receptive field of convolution kernels by deepening and widening the network structure, and employ equal processing methods in spatial domains and channel domains of varying importance, a large number of computing resources are wasted on unimportant features. In order to address the realistic problems of the existing models, the algorithm proposed in this paper captured implicit weight information in channel and space domains through dual attention modules, so as to allocate computing resources more effectively and speed up the network convergence. The fusion of global features through residual connections in this network not only focused on learning the high-frequency information of images that had been lost, but also accelerated the network convergence through effective feature supervision. In order to alleviate the defects of the MAE loss function, a special Huber loss function was introduced in the algorithm. The experimental results on benchmark show that the proposed algorithm can significantly improve the image reconstruction accuracy compared with the existent SISR methods.  Related Articles | Metrics

Computer Graphics and Virtual Reality

Select

Mixed reality based remote collaborative assembly guidance  LIU You-quan , WANG Yuan-chao , XU Kun , LIU Zheng-xiong , HUANG Pan-feng 2021, 42(2): 216-221.  DOI: 10.11996/JG.j.2095-302X.2021020216 Abstract ( 150 )   PDF (10179KB) ( 123 )   A remote collaborative assembly and maintenance system was proposed based on mixed reality technology. HoloLens’s spatial perception capability of the depth camera and RGB camera was employed to map the interaction from the 2D gestures on the expert side to the 3D animation markers on the operator side, where the gestures were recognized using an improved method. These simple and vivid 3D markers were located with spatial anchors to provide the operator with clear directions about the next steps for expert-operator-mode collaborative tasks, freeing the operator’s hands from interacting with the computer by gestures. The experiments verify the feasibility of such user-friendly guide interfaces for remote collaboration of assembly and maintenance tasks.  Related Articles | Metrics

Select

Public opinion evolution analysis of “COVID-19 epidemic” based on sentiment feature  GAN Yu-xiang , WANG Ya-bo , XUE Jun-xiao , ZHANG Ruo-qi , XU Shu-ning , GUO Yi-bo 2021, 42(2): 222-229.  DOI: 10.11996/JG.j.2095-302X.2021020222 Abstract ( 139 )   PDF (939KB) ( 96 )    In order to analyze the evolution of public opinion under emergencies and discover the law of the evolution of public opinion, a sentiment feature-based public opinion evolution analysis method was proposed, includdinga News Sentiment Analysis Module and a Public Opinion Evolution Analysis Module. The News Sentiment Analysis Module was based on the BERT pre-training model and the BiGRU model, where BERT was extracted as a word embedding, and BiGRU was employed to extract the contextual links of the textual feature vector to achieve a better understanding of the sentiment polarity of public opinion data. In the Public Opinion Evolution Analysis Module, this paper modeled the dynamic visualization of the sentiment features of public opinion in the time dimension, and then based on the visualization results, enabled the resolution of evolutionary patterns of public opinion data. Finally, a numerical experiment was conducted using one million pieces of the COVID-19 news data from January 1, 2020 to February 19, 2020. The experimental results show that the method proposed in this paper can effectively analyze the sentiment polarity of public opinion data.  Related Articles | Metrics

Select

The construction of spherical interpolation splines with local properties   WANG Qian , PAN Le , ZHANG Jie-lin , PENG Xing-xuan 2021, 42(2): 230-236.  DOI: 10.11996/JG.j.2095-302X.2021020230 Abstract ( 137 )   PDF (574KB) ( 121 )   The high dimensional spherical spline curves fitting technology has received wide attention in computer animation and inertial navigation. In practical applications, spline curves are usually required to interpolate the given data points with certain continuity and local properties. Thus, the previous methods are limited in certain regards. For this reason, a new method, based on spherical Bézier curves, of constructing spherical spline in arbitrary dimensional space was proposed. Firstly, the higher order derivative vectors at the interpolation points were fitted by a reflection. Then, necessary and sufficient conditions for C2 Hermite interpolation were given. Finally, the C2 spherical Bézier spline was constructed, using only interpolation points. The proposed method exhibitslocal properties. The disturbance of some points will not impact other parts of the spline. The splines possess explicit expressions not involving nonlinear equations. Numerical experiments show that the method can be widely applicable and efficient.  Related Articles | Metrics

Select

Compression of directed surface chain code in 3D model  LIU Shang-wu, WEI Wei, DUAN Xiao-dong, LIU Yong-kui 2021, 42(2): 237-244.  DOI: 10.11996/JG.j.2095-302X.2021020237 Abstract ( 61 )   PDF (3719KB) ( 49 )   Firstly, an improved MC normalization method was proposed that was applicable to 3D triangular face chain code algorithm. The per unit length in voxel was set as 2 in the improved MC algorithm, and the 27 points in a voxel was employed to rebuild a contoured surface, eventually obtaining the high-quality standardization triangular mesh model. Secondly, with the new normalization model, a new face traverse method was proposed. Based on the 3D triangular face chain code algorithm, the priority traversal right connection face principle was utilized to take control of the direction of face traverse. This method can reduce the number of traverses, and extend the average length of the face chain code. Experimental results show that the new normalization and the new traversal method, compared with the original 3D triangular face chain code, can significantly improve the compression effect.   Related Articles | Metrics

Select

Curve modeling using rational fractal interpolation with variable parameters ZHANG Xin-yue , LEI Yi-fan , LIU Pei-pei , BAO Fang-xun , ZHANG Yun-feng 2021, 42(2): 245-255.  DOI: 10.11996/JG.j.2095-302X.2021020245 Abstract ( 99 )   PDF (1096KB) ( 100 )   In order to deal with irregular data from complex real phenomena, a constructive approach to fractal curves was proposed using the rational fractal interpolation. First, based on the traditional rational spline with shape parameters, we constructed one class of rational Iterated Function Systems (IFS) with function vertical scaling factors. Rational IFSs were hyperbolic and rational fractal interpolation curves were defined. Then, some important properties of rational fractal curves were investigated, including smoothness, stability and convergence. Finally, lower and upper bounds in the box-counting dimension of rational fractal curves were estimated. The presented rational fractal interpolation with variable parameters generalized the traditional univariate rational spline, which is more suitable for fitting irregular data or approximating a function with continuous but irregular derivatives, and is more flexible and diversiform. Numerical examples and curve modeling show that this method can not only significantly outperform the Bézier interpolation, B-spline interpolation, and polynomial-based fractal interpolation methods in terms of visual effects, but also display prominent advantages in numerical comparison of root mean square errors.  Related Articles | Metrics

Select

Multi-view color point cloud registration based on correntropy  LIU Ying , LI Yao-chen , LIU Yue-hu , FU Jing-wen , MA Fei 2021, 42(2): 256-262.  DOI: 10.11996/JG.j.2095-302X.2021020256 Abstract ( 161 )   PDF (2796KB) ( 138 )   For the pair-wise registration of color point clouds, we proposed a color point cloud registration algorithm based on Correntropy to enhance the robustness and accuracy of traditional registration methods. On the basis of the iterative closest point algorithm, the proposed algorithm employed hue of HSV color space combined with the traditional three-dimensional coordinates to form a four-dimensional space to assist registration, and utilized Correntropy to reduce the impact of outliers and noise on registration, so as to achieve more accurate registration results. After the pair-wise registration was completed, the result calculated by this algorithm was taken as the initial value of multi-view registration, and then a more accurate multi-view registration result was achieved through the motion average algorithm that reduced the cumulative error. Experimental results of pair-wise registration show that the proposed approach is superior in accuracy and robustness compared with other approaches. In addition, the experimental results of simulated data and real data on public datasets show that the result computed by this algorithm can be well combined with the motion average algorithm as the initial value, and that the reliable results of multi-view point cloud registration can be obtained.  Related Articles | Metrics

Digital Design and Manufacture

Select

A light weight method of 3D printing model based on periodic cell  XIAO Wen-lei, WANG Zhi-ming, WANG Shi-ping, ZHAO Gang 2021, 42(2): 263-270.  DOI: 10.11996/JG.j.2095-302X.2021020263 Abstract ( 131 )   PDF (1737KB) ( 77 )   In order to obtain the material-saving effect and optimized physical and mechanical properties of 3D printing models, starting with the lattice shape variability, a lattice-based 3D printing lightweight structure generation method was proposed. The resulting lightweight structure was employed to replace the solid space of the given model. Firstly, a general unit cell description method was proposed to design the geometric and topological characteristics of unit cell. Secondly, the topology skeleton structure, on which the solid modeling was based, was constructed by periodically arranging lattice units in the bounding box of the model. Finally, a geometric modeling method based on mesh splicing was utilized to directly construct the STL model, and the lattice solid structure with controllable mesh quality can be obtained quickly without Boolean operation. The 3D printing molding process suitable for lattice structure was verified through object printing. Comparisons and analyses were conducted on the geometrical and mechanical properties of the five kinds of typical lattices serving as a preliminary basis for lattice design and selection. The examples show that this method can not only save material and improve the strength-weight ratio of the model, but also guarantee the printability, self-balancing, and the optimization of mechanical properties of the lightweight model. Compared with previous methods, it is advantageous in variability and efficiency and is applicable to various engineering applications.  Related Articles | Metrics

Select

APDL modeling method of hydraulic radial gate structure ZHANG Xue-cai, CHEN Li-ye, WANG Zheng-zhong 2021, 42(2): 271-278.  DOI: 10.11996/JG.j.2095-302X.2021020271 Abstract ( 156 )   PDF (4834KB) ( 109 )   The purpose of this paper is to address the problems of unclear process, inefficient method, and low quality in establishing geometric model and finite element model of hydraulic radial gate structure in the process of finite element analysis. Using the APDL parametric language of the general finite element calculation software ANSYS, the specific process and corresponding APDL command were given for the rapid establishment of high-quality geometric model and finite element model of radial gate. This method can ensure the authenticity of geometric model, correctness of finite element model, appropriateness of boundary conditions and rationality of analysis results. As a result, the key technical problems in the process of finite element analysis of radial gate were solved, and the design quality of the gate was enhanced. Based on the parametric modeling, the relevant parameters in gate design can be accurately and rapidly modified, and the design efficiency of gate can be boosted. The proposed method can shed light on high-quality design and analysis of the same type of gate structures, and lay a technical foundation for improving and enriching the current gate design specifications.  Related Articles | Metrics

Select

Simulation analysis of stress in power fitting for strain insulator-string following ice shedding from cables LI Chun , LIU Xiao-hui, HUANG Rui , YE Zhong-fei , WU Chuan , ZHANG Bo 2021, 42(2): 279-288.  DOI: 10.11996/JG.j.2095-302X.2021020279 Abstract ( 60 )   PDF (3456KB) ( 64 )   In order to address the damaging impact on the electric power fittings of the tension change of conductors due to ice shedding from cables, the wire model of UHV transmission line and the integral model of strain insulator-string were established respectively using the ABAQUS software. Based on the method of changing density, the ice shedding from six-bundle conductors with uniform ice-coating was simulated. Under non-synchronous ice shedding conditions, research was undertaken into the dynamic tension of sub-conductors and the torsion angle of the six-bundle conductor. The results show that the maximum dynamic tension of sub-conductors without ice shedding is larger than that of the static tension of iced conductors, and that the dynamic tension of the conductor is negatively correlated with the torsion angle of the conductor bundle. Finally, the finite element analysis was conducted based on the effect of static tension of iced conductor, and the maximum dynamic tension, the minimum dynamic tension, and the steady tension of conductors caused by iced shedding on the stress of strain insulator-string. The result reveals that the changing relationship between the stress of each electric power fitting of strain insulator-string and the tension of sub-conductors is inconsistent. The failure of the electric power fittings with greater stress and the sudden shape change were analyzed separately. The research results can inform the design and optimization of electric power fittings.  Related Articles | Metrics

BIM/CIM

Select

Research on information collaboration and model construction for top-down design of prefabricated buildings  ZHU Hui-xian, XU Zhao 2021, 42(2): 289-298.  DOI: 10.11996/JG.j.2095-302X.2021020289 Abstract ( 98 )   PDF (1634KB) ( 126 )   In order to comply with the national policy of building industrialization and intelligent development and solve such problems as the disconnectedness to professional information, insufficient use of models, and low efficiency of data transmission, a top-down design method based on the building information modeling (BIM) technology was proposed. Starting with the design stage of prefabricated building, the BIM technology-based top-down design process of prefabricated building was elaborated in detail. Combined with the characteristics of standardized component design and assembly of prefabricated buildings, an assembly model based on top-down design was proposed, and the structure of the model was described from the perspectives of hierarchy and specialty. To tackle the problems of standardization and consistency of data information during the transmission of professional models, the entities and property sets of assembly model were extended based on the industry foundation classes (IFC) standard, and the information collaboration using the IFC standard-based top-down design was proposed for each specialty and each stage. Finally, the application of the example verified the feasibility of the IFC-based top-down design of the extended assembly model, which can provide a useful method for collaborative design of prefabricated buildings.  Related Articles | Metrics

Select

BIM evacuation design automation based on artificial intelligence  LIANG Yu-qing, JI Jiu-mao, YANG Jia-lei , ZHANG Dong-sheng, WANG Ke , WANG Ling-yu, 2021, 42(2): 299-306.  DOI: 10.11996/JG.j.2095-302X.2021020299 Abstract ( 163 )   PDF (967KB) ( 151 )   To reduce the time cost of manual design of the Building Information Modeling (BIM) fire evacuation routes and to enhance design efficiency, an improved algorithm was proposed based on Deep Q Learning (DQN) and A* algorithm, through which a BIM evacuation design automation tool developed. First, the evacuation path of the room was drawn by A* algorithm. Then, the path from the evacuation door to safety exit in floor evacuation was determined using the improved DQN algorithm, and was drawn by A* algorithm. On the basis of the DQN algorithm, we redesigned reward matrix assignment and added reward matrix verification to improve the correctness of floor evacuation. Finally, we applied the improved algorithm-based evacuation design automation tool to practical projects. The results show that this improved algorithm can not only draw the route correctly, but also increase the efficiency by 2 to 3 times compared with manual drawing. The efficiency of this algorithm can be further improved by server deployment and hardware upgrades. At present, the automatic design tool has been adopted in several actual projects of Shanghai Digital Architecture Fabrication Technology Center (SFAB), Tongji Architectural Design (Group) Co., Ltd.  Related Articles | Metrics

Select

Research on named entity recognition of construction safety accident text based on pre-trained language model  SONG Jian-wei , DENG Yi-chuan , SU Cheng, 2021, 42(2): 307-315.  DOI: 10.11996/JG.j.2095-302X.2021020307 Abstract ( 177 )   PDF (536KB) ( 118 )   The construction safety accident analysis plays an important role in construction safety management, but the construction safety knowledge scattered in accident reports cannot be reused, nor can it shed sufficient light on construction safety management. Knowledge graph serves as a tool for structured storage and knowledge reuse, such as retrieval of accident cases, analysis of accident-related paths, and statistical analysis. Named Entity Recognition (NER) is the key task of automatic knowledge graph construction, and currently mainly concentrates on medical, financial, and military fields. In the realm of construction safety, there has been an absence of relevant research on NER. In this paper, five concepts in this field were defined, and the entity labeling specifications were clarified. The improved Bidirectional Encoder Representations from Transformers (BERT) pre-trained language model was employed to obtain dynamic word vectors, and the Bidirectional Long Short-Term Memory-Conditional Random Field (BiLSTM-CRF) model was utilized to gain the optimal entity tag sequence, thus proposing the NER model for the field of construction safety. In order to train and verify the proposed model, 1,000 accident reports on construction safety were collected, sorted, and annotated as an experimental corpus. Experiments show that compared with traditional models, the proposed model can yield a better recognition effect in texts on construction safety accident.  Related Articles | Metrics

Select

A model adaptive method for Chinese word segmentation using IFC-based building information model  ZHANG Xin , ZHOU Xiao-ping, WANG Jia, 2021, 42(2): 316-324.  DOI: 10.11996/JG.j.2095-302X.2021020316 Abstract ( 156 )   PDF (1340KB) ( 161 )   The building information model (BIM) has become an effective solution to information technology applications in the construction industry. With the continuous increase of BIM data, natural language processing (NLP) has been introduced into BIM applications in many studies to effectively utilize BIM data. In the Chinese language environment, due to the absence of terminology features in the building field, Chinese word segmentation cannot be efficiently adapted in BIM application. By analyzing the currently popular industry foundation class (IFC) files in BIM data format, this study extracted BIM model features from IFC files and added them together with architectural terminology characteristics into the statistical word segmentation model, thus improving the adaptability of Chinese word segmentation in the building field. The experimental results show that compared with the original conditional random fields (CRF)based word segmentation model, on the domain test set, the F-measure increased by 1.26%, and F-measure still increased by 0.10% with BIM model features added alone, indicating that appending BIM model features to the segmentation model can effectively improve the performance of Chinese word segmentation in the building field. Meanwhile, on the model test set, compared with the case of architectural terminology characteristics being appended alone, after BIM model features were appended, the precision rate increased from 46.97% to 87.74%, the recall rate from 67.60% to 94.77%, and the F-measure from 55.43% to 91.12% (by 35.69%), thereby effectively boosting the BIM model adaptability of Chinese word segmentation in the building field.  Related Articles | Metrics

Industrial Design

Select

Usability evaluation and application of multi-joint active and passive trainer based on user experience REN Ying-li, WU Shi-jin 2021, 42(2): 325-331.  DOI: 10.11996/JG.j.2095-302X.2021020325 Abstract ( 103 )   PDF (1538KB) ( 157 )   In order to improve the usability of multi-joint active and passive trainer, the usability evaluation index system of multi-joint active and passive trainer was constructed based on the user experience theory, and the usability evaluation model of multi-joint active and passive trainer was constructed through the fuzzy analytic hierarchy process (FAHP) and the membership quantitative processing method. Taking a multi-joint active and passive trainer produced by some company as an example, the evaluation index was determined by expert evaluation and the weight value of each index factor was calculated. Then the evaluation grade of each usability evaluation index was generated after the product was operated by the scoring team, and the final evaluation result was yielded based on the usability evaluation model of multi-joint active and passive trainer. The results show that the product usability can be evaluated comprehensively and accurately using the model, and then the usability problems can be identified and improvements can be proposed, thus informing the upgrading and innovation of products.  Related Articles | Metrics

Select

Research on evolutionary design of interactive interface for knowledge reuse and intelligent innovation  JING Chun-hui, ZHI Jin-yi 2021, 42(2): 332-338.  DOI: 10.11996/JG.j.2095-302X.2021020332 Abstract ( 110 )   PDF (2532KB) ( 109 )   In order to break away from the traditional design method depending on the designer’s personal experience and subjective cognition, to reuse design knowledge, and to provide innovative inspiration for artificial design with automatic intelligence, the evolutionary gene model of interactive interface was constructed based on the reuse knowledge rules and boundaries summarized in the process of interactive interface design. Then, the genetic algorithm was employed to construct the evolutionary modules of the interactive interface coding, such as genetic operators, knowledge reuse constraint rules base, and interactive semantic evaluation, forming an interactive interface evolutionary design method for knowledge reuse and automation intelligent innovation. Finally, on the basis of the method, forty-five effective offspring individuals were generated from the ten generations of evolution using five common login interactive interfaces as the initial population. The effectiveness of the method was verified by design practice. The results show that the design knowledge can be summarized and reused through design rules, and the automatic intelligent aided design can be undertaken through constraint evolution, thereby elevating the level of design knowledge reuse and computer-aided intelligent innovation design. Related Articles | Metrics

Total to Discuss

Select

Vol.42，No.2，2021 2021, 42(2): 339.  Abstract ( 37 )   PDF (57806KB) ( 209 )   Related Articles | Metrics