Table of Contents

18 January 2022, Volume 42 Issue 6 Previous Issue    Next Issue
---
Contents
Image Processing and Computer Vision
Computer Graphics and Virtual Reality
Digital Design and Manufacture
BIM/CIM
Industrial Design
Total Contents


For Selected:

Download Citations EndNote Ris BibTeX

Toggle Thumbnails

Contents

Select

Table of Contents for Issue 6, 2021 2021, 42(6): 1-1.  Abstract ( 57 )   PDF (231KB) ( 87 )   Related Articles | Metrics

Image Processing and Computer Vision

Select

Salient instance segmentation via attention residual multi-scale feature enhancement   SHI Cai-juan, CHEN Hou-ru, GE Lu-lu, WANG Zi-wen 2021, 42(6): 883-890.  DOI: 10.11996/JG.j.2095-302X.2021060883 Abstract ( 231 )   PDF (2943KB) ( 200 )   Salient instance segmentation is to segment the most noticeable instance object in the image. However, there remain some problems in the existing methods of salient instance segmentation. For example, the small salient instances are difficult to be detected and segmented, and the segmentation accuracy is insufficient for large salient instances. Therefore, to solve these two problems, a new salient instance segmentation model, namely the attention residual multi-scale feature enhancement network (ARMFE), has been proposed. ARMFE includes two modules, i.e. the attention residual network module and the multi-scale feature enhancement module. The attention residual network module combines the residual network with the spatial attention sub-module and the channel attention sub-module to enhance the features. The multi-scale feature enhancement module can further enhance the information fusion for features with large scale span based on the feature pyramid. Therefore, the proposed ARMFE model can make full use of the complementary information of multi-scales features by attention residual multi-scale feature enhancement, and then simultaneously improve the accuracy of detecting and segmenting large instance objects and small instance objects. The proposed ARMFE model has been tested on the salient instance segmentation dataset Salient Instance Saliency-1K (SIS-1K), and the segmentation accuracy and speed have been improved. This indicates that our proposed model outperforms other existing salient instance segmentation algorithms, such as MSRNet and S4Net.  Related Articles | Metrics

Select

Image encryption algorithm based on cellular neural network and parallel compressed sensing JIANG Dong-hua , LIU Li-dong , WANG Xing-yuan , RONG Xian-wei 2021, 42(6): 891-898.  DOI: 10.11996/JG.j.2095-302X.2021060891 Abstract ( 169 )   PDF (3121KB) ( 202 )    A high-security non-visual image encryption algorithm based on cellular neural network (CNN) and parallel compressed sensing (CS) was proposed, aiming to improve the information transmission efficiency and reduce the storage space of existing encryption algorithms. First, the wavelet coefficients of plain image were processed by thresholding and index confusion, and compressed by the key-controlled partial Hadama matrix in parallel. Next, the Fisher-Yates confusion and modular arithmetic were performed. Then the partial encrypted image was segmented and randomly hidden into the alpha channel of remaining encrypted image by the least significant bit (LSB) embedding algorithm, thereby generating the final noise-like cipher image. In this scheme, the pseudo-random sequences generated by CNN with hyperchaotic properties were employed to construct the scrambling, diffusion, and key-controlled measurement matrix. Eventually, a series of security analyses indicated that the proposed image encryption algorithm is of high efficiency and security in transmission.  Related Articles | Metrics

Select

Attribute and graph semantic reinforcement based zero-shot learning for image recognition WANG Yu-jin, XIE Cheng, YU Bei-bei, XIANG Hong-xin, LIU Qing 2021, 42(6): 899-907.  DOI: 10.11996/JG.j.2095-302X.2021060899 Abstract ( 125 )   PDF (1996KB) ( 92 )   Zero-shot learning (ZSL) is an important branch of transfer learning in the field of image recognition. The main learning method is to train the mapping relationship between the semantic attributes of the visible category and the visual attributes without using the unseen category, and use this mapping relationship to identify the unseen category samples, which is a hot spot in the current image recognition field. For the existing ZSL model, there remains the information asymmetry between the semantic attributes and the visual attributes, and the semantic information cannot well describe visual information, leading to the problem of domain shift. In the process of synthesizing unseen semantic attributes into visual attributes, part of the visual feature information was not synthesized, which affected the recognition accuracy. In order to solve the problem of the lack of unseen semantic features and synthesis of unseen visual features, this paper designed a ZSL model that combined attribute and graph semantic to improve the zero-shot learning’s accuracy. In the learning process of the model, the knowledge graph was employed to associate visual features, while considering the attribute connection among samples, the semantic information of the seen and unseen samples was enhanced, and the adversarial learning process was utilized to strengthen the synthesis of visual features. The method shows good experimental results through experiments on four typical data sets, and the model can synthesize more detailed visual features, and its performance is superior to the existing ZSL methods.  Related Articles | Metrics

Select

Cross-modal chat cartoon emoticon image synthesis based on knowledge meta-model  LI Xiao-rui, XIE Cheng, LI Bin, LIU Qing, HU Jian-long 2021, 42(6): 908-916.  DOI: 10.11996/JG.j.2095-302X.2021060908 Abstract ( 110 )   PDF (2434KB) ( 79 )   The traditional chat cartoon emoticon technologies are mainly based on the predefined chat cartoon emoticon library. Through the semantic description of users, the “semantic-to-visual” cross-modal retrieval is carried out to match the appropriate emoticon. However, the number of predefined emoticon samples in the library is limited and fixed. In the actual chat scenarios, the emoticon is often mismatched or there is no match at all. In view of this problem, this research focused on synthesizing new chat cartoon emoticon rather than retrieval. A new method of cross-modal chat cartoon emoticon synthesis based on knowledge meta-model was designed. According to the semantic description of users, the corresponding chat cartoon emoticons were synthesized immediately. The method established the inner semantic logic relation of chat cartoon emoticon through the knowledge meta-model, and enhanced the semantic consistency of chat cartoon emoticon synthesis. Through the multi-generator model, the corresponding partial chat cartoon emoticons were synthesized from each meta-knowledge point, and then integrated into a complete cartoon emoticon by the joint generator, which greatly reduced the training sample demand. In the test of public chat cartoon emoticon synthesis data set, the method has achieved better semantic consistency, and it is comparable with the existing methods in the quality of synthesized image.   Related Articles | Metrics

Select

Research on concrete image classification based on three-channel separation feature fusion and support vector machine  ZHANG Cheng , HOU Yu-chao , JIAO Yu-qian , BAI Yan-ping , LI Jian-jun 2021, 42(6): 917-923.  DOI: 10.11996/JG.j.2095-302X.2021060917 Abstract ( 65 )   PDF (2390KB) ( 64 )   The different mix ratios of concrete determine the performance of the material. The research on the classification of concrete images with various mix ratios and particle sizes is conducive to the efficient recycling of industrial waste concrete. In order to improve the classification effect, a new feature extraction module—image texture feature aided deep learning feature (ITFA-DLF) was proposed. This module employed convolution on the R, G, and B channels reconstructed from image separation and reconstruction. Convolutional neural network (CNN) extracted the color features of the three-channel image, utilized the multi-block local binary pattern (MB-LBP) to extract the texture features of the three-channel image, and merged the two features and input them into the support vector machine (SVM) optimized by the grid search (GS) algorithm for classification. Experiments with concrete images were adopted to compare various classification methods. It is concluded that the model proposed can produce the best effect. The recognition rate of nine types of images has reached more than 92%, and the classification time was shortened while ensuring the classification accuracy, and the classification efficiency of the concrete image was improved, which verified the effectiveness of the proposed method.  Related Articles | Metrics

Select

Implementation of meat classification system for autonomous robotic cutting  MA Huan, JI Jing-jing, LIU Jia-hao, LIU Yu-ting 2021, 42(6): 924-930.  DOI: 10.11996/JG.j.2095-302X.2021060924 Abstract ( 188 )   PDF (3201KB) ( 119 )   To solve the worldwide problems in the traditional poultry meat cutting process, including high labor costs, high safety risks, and other global problems, a robot autonomous cutting production line system was designed by integrating the key technologies such as accurate perception, rapid cutting, and autonomous deboning. To meet the requirements of efficient automatic classification and packaging for chicken breasts and wings (including wing tip, middle joint, and root) produced in the autonomous cutting process, a new recognition method combining image processing, convolutional neural network (CNN) classification, and the hardware/software collaborative framework was proposed, aiming to achieve the function integration and real-time requirements of image acquisition, processing, and detection. Firstly, the meat area was extracted to distinguish chicken breast and wing tip; secondly, the wing middle and root were classified based on CNN technology; finally, the recognition speed via software/hardware cooperation was estimated by parameter and computational efficiency analysis in the recognition algorithm. With the meat identification system platform built, the motion blur of the conveyor belt at full speed was analyzed, and the data set was expanded by data enhancement. In order to reduce the amount of computation, only the image data of R channel was used as the input of neural network. The results show that the recognition accuracy of chicken breast and wing tip can reach 100%, and that of wing middle and root can reach 98.7%, with recognition speed of 0.047 seconds, which could meet the research and development needs of efficient sorting of 10,000 poultries per hour for future work.  Related Articles | Metrics

Select

Age recognition method based on low resolution input image ZHU Xi-mei , LI Rui, 2021, 42(6): 931-940.  DOI: 10.11996/JG.j.2095-302X.2021060931 Abstract ( 67 )   PDF (3584KB) ( 136 )    If the accessed facial image is of low resolution, facial wrinkles and other characteristics of the information would often be lost, undermining the performance of age identification. In view of the existing age identification method lacking this research field and in order to solve this problem, this paper proposed an age identification method for low-resolution images by reconstructing the input low-resolution face images using conditional generative adversarial net (CGAN), and then identifying the age using the deep learning method. Firstly, a comparative experiment on image reconstruction was carried out, and then the results of age recognition were compared on different face image data sets. The experimental comparison with other deep learning methods on signal noise ratio, peak signal noise ratio, and mean absolute error shows the effectiveness of the proposed method in image reconstruction and age recognition. In addition, the time complexity of the proposed method was also analyzed.  Related Articles | Metrics

Select

A eulerian video magnification based cable tension identification method for bridge structures  ZHANG Yu-hang , SU Cheng, DENG Yi-chuan, 2021, 42(6): 941-947.  DOI: 10.11996/JG.j.2095-302X.2021060941 Abstract ( 129 )   PDF (914KB) ( 93 )   Cables, booms, and ties are important load-bearing elements of cable-bearing bridges. However, the traditional cable tension identification methods have the disadvantages of complicated operation, high cost, and low efficiency. This paper proposed a cable tension identification method for bridge structure based on Eulerian video magnification. The digital image of cable vibration collected by digital camera was processed by the phase-based Eulerian motion magnification, after which the image sequence after motion magnification was obtained. The edge image sequence was obtained by edge recognition algorithm based on Canny operator, and the displacement data of the mark point was extracted by edge positioning. Finally, the cable force was identified by frequency method. The applicability of the phase-based Eulerian motion magnification algorithm and the edge recognition algorithm based on Canny operator was discussed in the context of the arch bridge boom cable force test. Compared with the traditional acceleration sensor-based cable force testing method, the proposed cable tension identification method for bridge structure based on a motion magnification algorithm has the features of easy operation, low cost, and high efficiency.  Related Articles | Metrics

Select

Underwater image enhancement algorithm using gated fusion generative adversarial network  LIN Sen , LIU Xu 2021, 42(6): 948-956.  DOI: 10.11996/JG.j.2095-302X.2021060948 Abstract ( 200 )   PDF (24652KB) ( 127 )   An underwater image enhancement algorithm using gated fusion generative adversarial network was proposed for solving problems of underwater image color distortion, low contrast, and heavy fogging. The crucial element of this algorithm is that it recruited a generator to pixel-by-pixel restore image feature details and synthesized a clear image through gated fusion. First of all, to increase the variety of image feature learning by the network, several parallel sub-networks were employed to learn different kinds of spatial feature knowledge of the same image. The image features learned from different sub-networks were fused utilizing gated fusion. The generator and the discriminator were used for mutual games, and the network was repeatedly trained to obtain enhanced underwater images. Finally, using the EUVP dataset and the U45 testset, this paper performed a series of comparative experiments. The algorithm’s key point matching was 19 points higher than the raw image, according to the experimental results. The average UCIQE value was 0.664 7, while the average UIQM value was 5.723 7. It can achieve improvements over other classic and latest algorithms, demonstrating the algorithm’s extraordinary performance.  Related Articles | Metrics

Computer Graphics and Virtual Reality

Select

A fast and free interactive picking method for complex surfaces  XIONG Yu-long, LI Wei-shi 2021, 42(6): 957-962.  DOI: 10.11996/JG.j.2095-302X.2021060957 Abstract ( 56 )   PDF (889KB) ( 63 )   As an intuitive way for human-computer interaction, the interactive picking of 3D models is widely used in various fields, such as geometric modeling, 3D games, and finite element analysis. To address the problems of low picking efficiency and low degree of picking freedom of complex surface models, a fast and free interactive picking method for complex surface model was proposed. This method employed a brush-like tool to discretize the moving path of the brush in the screen, applied the single point picking method based on BVH structure to pick up the complex surface for each discrete point, and then utilized two hash structures to reduce the response time of the algorithm, and finally the algorithm was verified.  Related Articles | Metrics

Select

A novel wire frame generation for triangle meshes QIN Yu , CAO Li, WU Yao , LI Lin, 2021, 42(6): 963-969.  DOI: 10.11996/JG.j.2095-302X.2021060963 Abstract ( 250 )   PDF (997KB) ( 134 )   Extracting wire frame from 3D models is a challenge. Existing methods are typically based on the analysis of local shape properties, such as surface curvatures and angles between faces, which are generally sensitive to small features in the model. In order to solve this problem, we proposed a wire frame extraction method based on geometric approximation of 3D shapes. This method employed a well-established variational geometric segmentation algorithm to derive a complete set of descriptive feature curves. Firstly, the model was divided into several patches based on the variational geometry approximation method. Secondly, the internal characteristic curves of all patches were extracted, and the short characteristic curves were filtered. Then, the boundary curves of patches were smoothed. Finally, the patch boundary curves and characteristic curves were merged, and the closed wireframe network was obtained by extending the curves. The advantage of the proposed method was that descriptive wire frames can capture the global structures of the 3D shapes using a reliable feature filtering mechanism that was inherently incorporated in the geometric approximation step. Experiments on various kinds of meshes have been carried out and the results demonstrate that our method is superior to existing approaches in terms of correctness and completeness of the extracted wire frame.  Related Articles | Metrics

Select

Application in sheet metal forming based on h-adaptive finite element method  LIU Xi-zhou, WANG Cheng-jing, WANG Hu 2021, 42(6): 970-978.  DOI: 10.11996/JG.j.2095-302X.2021060970 Abstract ( 67 )   PDF (1722KB) ( 72 )   In the finite element simulation analysis of the sheet metal forming process, it is difficult to accurately analyze the stress concentration and strain gradient areas where the field variables change drastically. How to balance the relationship between accuracy and efficiency becomes the key to stamping simulation. Therefore, based on the related theory of nonlinear finite element large deformation and the key technology of mesh adaptation for dynamic simulation, an algorithm for sheet metal forming under the adaptive analysis mode was established. In order to improve the calculation accuracy, an energy error criterion based on the unit strain energy increment and a geometric error criterion based on the geometric characteristics of sheet metal forming were proposed. Combining these two types of error criteria, an error judgment criterion based on an adaptive analysis algorithm was established. In order to enhance the calculation efficiency, by introducing a damper factor, a damper subcycling algorithm was proposed, which divided the adaptive sheet element unit into several regions according to the time step, and each region was integrated separately according to the time step. The results show that the algorithm improves the accuracy and efficiency of finite element simulation of sheet metal forming.  Related Articles | Metrics

Select

Vector map curve simplification algorithm based on progressive-iterative approximation  ZHOU Chen , CHEN Wei, LIU Yuan, 2021, 42(6): 979-986.  DOI: 10.11996/JG.j.2095-302X.2021060979 Abstract ( 68 )   PDF (882KB) ( 90 )   Vector map simplification plays an important role in the research on terrain simulation, cartographic generalization, and so on. As it is difficult to balance the overall shape and local feature point accuracy of the simplified curve with the existing algorithms, a vector map simplification method based on progressive iterative approximation (PIA) with B-spline curve was proposed. First, select the feature point sequence that can maintain the contour of the curve with the largest amount of information, and use it as the initial control point sequence to obtain the corresponding nonuniform cubic B-spline curve. Secondly, it obtained a series of curves that were gradually fitting the real one by iteratively adjusting the control points according to the bias between the fitted curve and the feature points until the accuracy requirements were met. The experiments result show that the PIA method can not only keep the overall geometry of the map curve, but also achieve high-precision approximation at feature points while meeting the global bias requirements.  Related Articles | Metrics

Digital Design and Manufacture

Select

Optimization of milling parameters for thin-walled parts based on improved chaotic particle swarm optimization algorithm LIU Si-meng, WANG Gang-feng, SUO Xue-feng 2021, 42(6): 987-994.  DOI: 10.11996/JG.j.2095-302X.2021060987 Abstract ( 86 )   PDF (1227KB) ( 163 )   In order to improve milling precision and processing efficiency of thin-walled frame structural parts, a method for optimizing the milling processing parameters of thin-walled frame structural parts was proposed. Aiming at the problems of standard particle swarm algorithm that are easy to fall into local optimal solutions and cannot adjust weight coefficients adaptively, this method combined chaos optimization algorithm and multi-objective particle swarm optimization algorithm to establish the optimization target based on milling force and material removal rate per unit time. The four factors of milling were taken as optimization variables, and the spindle speed, feed rate, milling depth, and surface roughness were taken as constraints. The machining error of each optimization solution was calculated accurately by finite element simulation, and the results were fed back to the optimization algorithm in time, so as to find the optimal machining parameter combination. Taking typical thin-walled structure sidewall milling as an example, experimental parameters, standard particle swarm optimization parameters, and optimization results of the algorithm proposed in this paper were used for simulation respectively, and the simulation results were analyzed and compared, which proves the effectiveness of the proposed method.  Related Articles | Metrics

BIM/CIM

Select

Research on levels and classification of city information model WANG Yong-hai , YAO Ling , CHEN Shun-qing , BAO Shi-tai, 2021, 42(6): 995-1001.  DOI: 10.11996/JG.j.2095-302X.2021060995 Abstract ( 547 )   PDF (944KB) ( 225 )   The research is aimed at key characteristics of City Information Model (CIM) on level and classification. CIM could provide multi-scale 3D data base and scientific decision support for city planning, construction, management and operation, which becomes a part of 14th five - year national economic and social development planning in China. CIM has the complex characteristics of multi-scale, multi-specialty and multi-industry cross-integration, which is difficult to guide the integrated application and sharing information of models. The establishment of a unified levels and classification of CIM will effectively improve the situation. By comparing and analyzing the existing standard systems of City three-dimensional model, CityGML classification and Building information model (BIM), comprehensive concept and levels of CIM are deeply explored. This paper proposed a gradually refined levels of CIM from 1st to 7th, which includes terrain surface model, frame model, standard model, fine model, functional level model, component-level model, and part-level model. The content, feature and precision of seven-level models were described and preliminarily verified. Based on BIM classification, this paper used faceted classification method to define and expand CIM classification., including the five dimensions of result, process, resource, character and application. This research explored the levels and classification characteristics of CIM, which contribute to the creation, procedure, display, sharing and application of CIM.  Related Articles | Metrics

Select

Review on intelligent performance evaluation of construction supervisors  ZHENG Wei , LIN Jia-rui , YANG Cheng , YAN Ke-xiao , CHENG Yu 2021, 42(6): 1002-1010.  DOI: 10.11996/JG.j.2095-302X.2021061002 Abstract ( 139 )   PDF (499KB) ( 107 )   The project supervision system is key to ensuring the project quality in the construction field. Supervisors are the essential parts of project supervision, and evaluating their performance is of great significance to ensuring the quality of a construction project. Currently, related research still focuses on the common goal, content, and procedure of performance evaluation, and is highly dependent on manual operation. Meanwhile, the performance indicators are usually rigid and unitary, leading to inappropriate and unclear results. The development of advanced information technology has brought new opportunities and methods for the intelligent performance evaluation of supervisors. This paper aims to systematically review the performance evaluation methods and the underlying intelligent technologies, and to propose new directions and trends for intelligent performance evaluation. This paper reviewed the current status of performance evaluation of supervisors assisted by intelligent supervision technology from two aspects, namely, the perspective of technical methods, and the perspective of application. Key evaluation methods and theories involved were summarized, and application of corresponding methods based on information technology were discussed. Based on prior work, discussion and analysis were conducted to identify existing problems and challenges. To tackle these problems, the developments of simplified performance evaluation methods, and more flexible and objective indicators were suggested. Thus, this paper shed light on the future performance evaluation of supervisors under the development of intelligent construction with modern comprehensive evaluation method combined with artificial intelligence, block chain technology, and other emerging information technologies.  Related Articles | Metrics

Select

Research on closed-loop management system of BIM course teaching in universities  ZHAO Xue-feng, HOU Xiao, LIU Zhan-sheng, LI Meng-xuan 2021, 42(6): 1011-1017.  DOI: 10.11996/JG.j.2095-302X.2021061011 Abstract ( 182 )   PDF (852KB) ( 108 )   With the development of building information modeling (BIM) and the introduction of national policies, BIM technology has been gradually incorporated into the teaching system of universities. However, the existing BIM courses and teaching methods in universities cannot meet the requirements of BIM personnel training. Based on the analysis of the existing problems of BIM course, this paper first constructed a closed loop of BIM course teaching content composed of professional knowledge, software knowledge, software operation knowledge, and project application knowledge. Secondly, according to the closed-loop management concept, this paper put forward the closed-loop teaching process of BIM course, which included pre-class review and introduction, traditional teaching and learning, class practice and discussion, and post-class practice release submission and instant feedback with the help of online platform, and proposed the objective answer method of subjective questions in the learning effect feedback link. Then, inspired by Gilbert’s behavioral engineering model, the resource scene and application scene were established in key links, and the closed loop of BIM teaching scene was constructed. Finally, the closed-loop management system was applied in BIM courses offered in a university, and the experience and lessons were summarized, which can provide reference for BIM teaching reform in universities in China.  Related Articles | Metrics

Industrial Design

Select

Research on multi-stage network coupling decision-making method of product styling design YANG Yan-pu, LAN Chen-xin, LEI Zi-jing, WANG Xin-rui, GONG Zheng 2021, 42(6): 1018-1026.  DOI: 10.11996/JG.j.2095-302X.2021061018 Abstract ( 70 )   PDF (951KB) ( 109 )   To comprehensively consider multi-stage decision-making information of product styling design, the complex network theory was introduced and a multi-stage network coupling decision-making process of product styling design was proposed. The relationship of multi-stage decision-making information of product styling design was modeled with coupled association. Based on the consistency analysis of decision-makers’ scores, the adjacency matrix of design decision-making network was identified. The decision-makers’ weights were obtained by calculating changes in network cohesion. Through analyzing the stability of decision-making network cohesion and the consistency of decision-makers’ opinions, the number of decision-making rounds was determined. To compute the weights of each stage in the decision-making process of product styling design, a programming function was constructed, and the aggregation of multi-stage decision-making information was realized through linear coupling. Taking the decision-making of numerical control grinder styling design as an example, the effectiveness of the method was verified. Results show that the proposed method can help realize the aggregation of multi-stage decision-making information of product styling design and improve the quality of design decision-making in an overall and scientific way.  Related Articles | Metrics

Select

A team situation awareness-based approach to automotive HMI evaluation and design YOU Fang , XIE Yu-kun , YUE Tian-yang , YU Ai-qun , WANG Jian-min , ZHANG Hui 2021, 42(6): 1027-1034.  DOI: 10.11996/JG.j.2095-302X.2021061027 Abstract ( 213 )   PDF (1391KB) ( 151 )   With the development of intelligent and networked vehicle technologies, the automotive human-machine interface (HMI) has become important to the drivers’ user experience as a medium between human and vehicle. In order to improve the teamwork between human and vehicle by optimizing the human-machine interface, the evaluation and design method of automotive human-machine interface was studied. Twenty-six drivers participated in the evaluation of driving simulator with FM as the multi-path interface paradigm. The evaluation method was based on the team situation awareness theory and user experience elements. A multi-dimensional evaluation system containing driving indicators, behavioral indicators, and subjective indicators was constructed. The association between situation awareness and user experience elements was further examined by analyzing task stages and channels before the experiment, and the results of the five levels of user experience data of the automotive HMI were analyzed after the experiment, thus locating the occurrence of design problems and deducing the corresponding design optimization directions. Finally, the design optimization solution was obtained, which can help optimize the design of automotive HMIs and enhance user experience.  Related Articles | Metrics

Select

Non-random population method for intelligent optimization of color matching design  LIU Xiao-jian, PENG Cheng, LV Yun-yun 2021, 42(6): 1035-1042.  DOI: 10.11996/JG.j.2095-302X.2021061035 Abstract ( 64 )   PDF (4069KB) ( 59 )   The optimization method of color matching design often faces convergence difficulties during the formation of the final scheme, because the designer’s evaluation usually enters an unstable state, which makes it difficult to accurately choose among the slightly different schemes. Thus, after the optimization method has completed the large-scale search, the final convergence stage generally still needs the designer to do the manual fine-tuning, which significantly lowers the efficiency of the whole optimization process. A non-random population generation technique based on continuous interpolation was developed for the convergence stage of interactive genetic algorithm to meet the demand of color image reproduction, which can assist designers to realize rapid fine-tuning and output the final design. A prototype system was developed based on graphic design software, in which two kinds of variation operations based on RGB and HSB color spaces and a multi-scheme fusion of crossover operation were given as three kinds of non-random population generation methods. It enabled designers to have a more intuitive visual perception on the color difference, to select schemes more quickly, and to improve the interactive experience of designers as users. The application is verified by the case of color matching design of e-sports chairs.  Related Articles | Metrics

Select

Research on human-machine interface evaluation method based on QFD-PUGH  LI Hui , WANG Shi-ying , LI Juan-li 2021, 42(6): 1043-1050.  DOI: 10.11996/JG.j.2095-302X.2021061043 Abstract ( 181 )   PDF (673KB) ( 126 )   The human-machine interface design and optimization consists of the transformation from user demand to design demand and the formation of the best program based on design demand. Evaluation is significant for design optimization, while in the existing studies on evaluation method, the above two parts are performed in different stages, leading to inefficiency in data sharing. For this problem, an evaluation method of human-machine interface based on quality function deployment (QFD) and the PUGH decision matrix was put forward. Firstly, the Analytic Hierarchy Process was adopted to produce the weight of user demand, and corresponding user needs were mapped to relevant design requirements based on interface product features. Then a house of quality model was built based on QFD. According to the matrix of the degree of correlation between user requirements and design requirements, the weight was obtained, and the user’s subjective description was transformed into the design requirements. Finally, the design requirements and weights derived from the QFD were employed as the criteria for PUGH decision-making evaluation, the best scheme was chosen, thus completing a comprehensive evaluation of interface design. Taking the interface of the household treadmill as an example, it shows the effectiveness of the proposed method, and provides decision-making method and optimization ideas for the whole process of human-machine interface implementation.  Related Articles | Metrics

Select

Research on product family ontology modeling image mining method based on big data  WANG Peng, ZHU Wei-long 2021, 42(6): 1051-1060.  DOI: 10.11996/JG.j.2095-302X.2021061051 Abstract ( 161 )   PDF (3529KB) ( 99 )   The research is aimed at improving the accuracy of product modeling image formation and enhancing the standardization of image vocabulary extraction and knowledge reusability in the research on perceptual engineering. This research started with the concepts of homogenousness, pan-ethnicity, and heterogeneousness, and defined the model image ontology of the target product family; then, with the help of word2vec tools and principal component analysis (PCA), the relevance association and dimensionality reduction extraction of the product family image vocabulary was respectively completed, thereby constructing an image vocabulary mining mechanism. This mechanism can assist designers to more efficiently and accurately mine the image of the target product using network big data resources, which can to a certain extent solve the vagueness of the image mining method in traditional perceptual engineering. Finally, combining the mapping relationship between the image vocabulary and the modeling features, the Protege tool was employed to construct the ontology model of the product family modeling image, which logically characterized the modeling image knowledge of the target product, and provided reference for the next-generation products to inherit and develop the original product family modeling image. Taking the modeling image of Mazda MX-5 car series as an example, the ontology model was constructed, and the conceptual design of the front face of the child product was undertaken.  Related Articles | Metrics

Total Contents

Select

Total Contents of  2021 2021, 42(6): 1061-1064.  Abstract ( 90 )   PDF (349KB) ( 42 )   Related Articles | Metrics