Table of Contents

30 June 2022, Volume 43 Issue 3 Previous Issue   
---
Cover
Contents
Image Processing and Computer Vision
Computer Graphics and Virtual Reality
Digital Design and Manufacture
BIM/CIM
Industrial Design
Published as


For Selected:

Download Citations EndNote Ris BibTeX

Toggle Thumbnails

Cover

Select

Cover of issue 3, 2022 2022, 43(3): 0.  Abstract ( 74 )   PDF (1592KB) ( 107 )   Related Articles | Metrics

Contents

Select

Table of Contents for Issue 3, 2022 2022, 43(3): 1-1.  Abstract ( 56 )   PDF (232KB) ( 60 )   Related Articles | Metrics

Image Processing and Computer Vision

Select

Homography estimation for multimodal coin images DENG Zhuang-lin, ZHANG Shao-bing, CHENG Miao, HE Lian 2022, 43(3): 361-369.  DOI: 10.11996/JG.j.2095-302X.2022030361 Abstract ( 165 )   PDF (8273KB) ( 109 )   Registration of coin images under different illuminant is the predecessor of coin surface defect detection. However, the traditional multimodal registration method based on mutual information is slow and low accuracy, and the existing image registration methods realized by homography estimation based on deep learning only work in single-mode tasks. A homography estimation method based on deep learning for multimodal coin images is proposed in this paper, and image registration can be realized with the estimated homography. First, the homography estimation layer is used to estimate the homography between the pair of input images, and the homography is used for perspective transformation of the image to be registered; Then, the image translation layer is used to translate the pair of images to the same domain, and this layer can be removed in inference so as to reduce the inference time; Finally, train the network with the loss calculated using the pair of images in the same domain. Experiments show that the average distance error of the proposed method on the test set is 3.417 pixels, which is 38.71% lower than the traditional multimodal registration method based on mutual information. The inference time of the proposed method is 17.74 ms, which is much less than 6368.49 ms of the traditional multimodal registration method based on mutual information. Related Articles | Metrics

Select

3D spatial scale estimation based on two vanishing points and local scale ZHOU Xiao-hui, DING Xiao-feng, XIONG Yun-hui, PENG Chang-xin 2022, 43(3): 370-376.  DOI: 10.11996/JG.j.2095-302X.2022030370 Abstract ( 140 )   PDF (1792KB) ( 80 )   3D spatial scale estimation is of great significance to 3D reconstruction, and is in demand in the real world in the case of a single image. Generally, the camera must be calibrated before the scale estimation. According to the perspective principle of the monocular image, a method based on two vanishing points and local scale information was proposed to calibrate the camera, thereby obtaining the estimation of the 3D scale information in the monocular image. Firstly, two orthogonal groups of parallel lines were selected from the monocular image to produce the coordinates of two vanishing points. Then the rotation matrix between the world coordinate system and the camera coordinate system was yielded using the vanishing point coordinate and the focal length information. The translation vector was also acquired through the property of vanishing point and the known local scale information. Finally, the 3D worldcoordinates corresponding to the pixels in the 2D image were restored to calculate the 3D scale information. Experimental results show that this method can effectively estimate the scale of building objects in a single image. Related Articles | Metrics

Select

Product surface defect detection and segmentation based on anomaly detection WANG Su-qin, REN Qi, SHI Min, ZHU Deng-ming 2022, 43(3): 377-386.  DOI: 10.11996/JG.j.2095-302X.2022030377 Abstract ( 253 )   PDF (1183KB) ( 185 )   In industrial manufacturing, it is difficult to obtain defective samples and the defects are in diverse forms. Anomaly detection, which only trains positive samples, is being increasingly applied to defect detection on product surfaces. Anomaly detection generally determines whether the product has defects by evaluating the anomaly score of the product image, while unable to describe the locations of the defects. The latest anomaly segmentation method has been improved, but the segmentation of the defective area is not accurate enough. Based on the anomaly detection method, normalization flow was employed to judge whether the product surface was defective, and multi-scale feature fusion and alignment were adopted to initially locate the defects. Combined with the gradient and maximum information entropy, the watershed algorithm was used to optimize the initial positioning results to obtain the defect segmentation mask. The detection and segmentation results on the three surface defect datasets of Lisheng Board, KolektorSDD, and AITEX are superior to other similar methods. In addition, good detection and segmentation accuracy can also be achieved on few-shots. Related Articles | Metrics

Select

Offline handwriting mathematical symbol recognition based on improved YOLOv5s FANG Hong-bo, WAN Guang, CHEN Zhong-hui, HUANG Yi-wei, ZHANG Wen-yong, XIE Ben-liang 2022, 43(3): 387-395.  DOI: 10.11996/JG.j.2095-302X.2022030387 Abstract ( 284 )   PDF (1661KB) ( 174 )   Offline mathematical symbol recognition is the premise of offline mathematical expression recognition. The existing offline symbol recognition methods can only recognize symbols, but is of no help to other steps of offline expression recognition, even restricting expression recognition. Thus, an improved YOLOv5s offline symbol recognition method was proposed. Firstly, considering the small size of symbolic image, generative adversarial network (GAN) was employed to enhance the data. Secondly, from the point of view of symbolic categories, the spatial attention mechanism was introduced to YOLOv5s model, and the global maximum and global mean were pooled to enlarge the differences between categories. Finally, from the point of view of the symbol itself, the bidirectional long-short-term memory network (BiLSTM) was utilized to process the symbol feature matrix, so that the symbol feature could possess the upper and lower related information. Experimental results show that the improved YOLOv5s achieves better offline symbol recognition, with a recognition rate of 92.47%. Compared with other methods, the proposed method is effective and robust. At the same time, it can effectively avoid the problem of error accumulation in offline mathematical expression recognition and provide an effective basis for expression structure analysis. Related Articles | Metrics

Select

Embedded substation instrument detection algorithm based on improved YOLOv4 FAN Xin-nan, HUANG Wei-sheng, SHI Peng-fei, XIN Yuan-xue, ZHU Feng-ting, ZHOU Run-kang 2022, 43(3): 396-403.  DOI: 10.11996/JG.j.2095-302X.2022030396 Abstract ( 139 )   PDF (3989KB) ( 151 )   With the rapid development of robotics technology, intelligent robots are widely used in substation inspections. Aiming at the problem that the current target detection algorithms have too many parameters and the performance of embedded devices is limited. It is difficult to achieve real-time detection on the embedded platform. A n improved YOLOv4 embedded substation instrument detection algorithm is proposed. The algorithm is based on YOLOv4 and uses MobileNetV3 as the backbone feature extraction network. It reduces the amount of calculation and increases the detection speed while ensuring that the model can effectively extract features. At the same time, the convolution operation in the path aggregation network (PANet) is replaced with a depthwise separable convolution after feature extraction; the training strategy of transfer learning is used to overcome the difficult problem of model training. Finally, the improved model is optimized by TensorRT to achieve fast and efficient deployment reasoning. The improved algorithm is tested on the embedded NVIDIA Jetson Nano, and the experimental results show that the detection speed is increased by 2 times to 15 FPS at the expense of less accuracy. This provides the possibility for real-time instrument detection in edge computing scenarios. Related Articles | Metrics

Select

Scene flow prediction with simulated real scenarios MEI Hai-yi, ZHU Xiang-yu, LEI Zhen, GAO Rui, MA Xi-bo 2022, 43(3): 404-412.  DOI: 10.11996/JG.j.2095-302X.2022030404 Abstract ( 134 )   PDF (5276KB) ( 82 )   Artificial intelligence is stepping into the age of cognition, the ability of cognizing and inferring the physical world for machines needs to be improved. Recent works about exploring the physical properties of objects and predicting the motion of objects are mostly constrained by simple objects and scenes. We attempted to predict the scene flow of objects in simulated scenarios to extend common sense cognizing. First, due to the lack of data in the related field, a dataset called ModernCity based on simulated scenarios is proposed, which contains the street scene of modern cities designed from the perspective of cognizing common sense, and provides RGB images, depth maps, scene flow, and semantic segmentations. In addition, we design an object descriptor decoder (ODD) to predict the scene flow through the properties of the objects. The model we proposed is proved to have the ability to predict future motion accurately through the properties of objects in simulated scenarios by experiments. The comparison experiment with other SOTA models demonstrates the performance of the model and the reliability of the ModernCity dataset. Related Articles | Metrics

Select

Peak-point similarity fitting-based GPR hyperbola extraction method ZHAO Wei-wei, YUAN Da, JIANG Xin-bo, ZHANG Ya-sen 2022, 43(3): 414-424.  DOI: 10.11996/JG.j.2095-302X.2022030414 Abstract ( 65 )   PDF (3725KB) ( 57 )   In ground-penetrating radar applications, hyperbolic waves are the key morphological features for subsurface target identification, as well as for the acquisition of location, size, and other important parameters. Due to the influence of complex subsurface clutter factors, hyperbolic waves tend to be morphologically blurred, chaotic, and discontinuous, leading to high complexity of hyperbolic wave extraction and difficulty of uniform modeling. To improve the robustness of hyperbolic wave extraction, a hyperbolic wave extraction method based on peak point similarity fitting (PSFE) was proposed. For the time-varying characteristics of hyperbolic waves, especially the problem of hyperbolic waveform breakage in images, a waveform clustering model was constructed to obtain the set of peaks of interest using the similarity of subwave regions. Through the effective separation of the clutter waves from the target hyperbolic waves using the fitting, the dependence of the algorithm on the image quality was reduced, thus enhancing the robustness of hyperbolic wave extraction. Comparative experiments were conducted on simulated and real datasets to verify the performance of the PSFE algorithm for hyperbolic wave extraction for different types of images. The experiments show that the algorithm is of high feasibility and robustness in complex background noises and the clutter interference environment. Related Articles | Metrics

Select

Research on mural line enhancement based on block PCA and endmember extraction MAO Jin-cheng, LYU Shu-qiang, HOU Miao-le, WANG Wan-fu 2022, 43(3): 425-433.  DOI: 10.11996/JG.j.2095-302X.2022030425 Abstract ( 98 )   PDF (10867KB) ( 68 )   Linear feature is an important element in murals. However, natural or human factors tend to make it difficult for human eyes to distinguish some blurred lines of the murals. Therefore, a linear feature enhancement method using hyperspectral image block principal component analysis (PCA) and image unmixing was proposed. Firstly, the support vector machine (SVM) was employed to classify the hyperspectral composite image of the mural, the result of which could help produce the mural label data. In doing so, the block data of the homogeneous area of the hyperspectral image could be acquired. Secondly, vertex component analysis (VCA) was performed on each segmented image to obtain a candidate endmember set. The final endmember set was determined by constructing a projection matrix and merging similar endmembers. Then, the non-negative least squares unmixing was used to obtain the line abundance map. Finally, the first principal component image of the block principal component analysis was normalized, and band calculation was performed with the line abundance map to obtain the linear feature enhanced image. They were fused with the true color composite image to obtain the linear feature fusion image. Taking some hyperspectral images of murals in Qutan Temple, Qinghai Province, China as an example, the results show that the algorithm can enhance the linear features in the murals, which is superior to the PCA enhancement method. Related Articles | Metrics

Select

High definition reconstruction of black and white cartoon based on recurrent alignment network LI Hua-en, ZHAO Yang, CHEN Yuan, ZHANG Xiao-juan 2022, 43(3): 434-442.  DOI: 10.11996/JG.j.2095-302X.2022030434 Abstract ( 70 )   PDF (6079KB) ( 53 )   Various quality problems would arise in the process of vintage cartoons digitization, for example, a mixture of scratches and stains, reduction of resolution, and complex noises. The enhancement of old cartoon videos is a special sub-problem of video enhancement, which is barely researched. Hence, a multi-frame recurrent alignment network for black and white cartoon video reconstruction was proposed to enhance video quality. The recurrent neural network was employed to fully exploit the temporal redundancy among the neighboring frames, and extract historical information to remove scratches and stains, thus solving the difficult problems of continuous scratches and stains. Deformable convolution was applied in a coarse-to-fine manner to frame alignment at the feature level, which improved the capability of extracting the related inter-frame information in large motion scenes. The pyramid network with residual dense connections on multiple scales was introduced as the basic network unit to facilitate information aggregation. Experiments were conducted on multiple real vintage cartoon datasets and degraded datasets, which validated the performance of the proposed method. Meanwhile, such objective evaluation metrics as peak signal-to-noise ratio (PSNR) was adopted to measure the quality of the reconstructed cartoons. The test data confirms that the enhancing network can fully exploit the temporal redundancy among neighboring frames and quickly remove scratches and spots. The comparative experimental results show that our method outperforms several state-of-the-art approaches. The subjective experimental results demonstrate that the reconstructed cartoons can meet the needs of modern visual quality. Related Articles | Metrics

Computer Graphics and Virtual Reality

Select

Combinatorial quadratic Phillips q-Bézier curves with monotone curvature LIANG Ji-na, XIE Bin, HAN Li-wen 2022, 43(3): 443-452.  DOI: 10.11996/JG.j.2095-302X.2022030443 Abstract ( 65 )   PDF (681KB) ( 69 )   Phillips q-Bézier curves are a class of generalized Bézier curves containing q-integers. The research was conducted on the curvature monotonicity condition of quadratic Phillips q-Bézier curve from two aspects of algebra and geometry. Based on this, the following two curves were constructed: a quadratic Phillips q-Bézier curve with monotonous curvature and a combined quadratic Phillips q-Bézier curve with decreasing curvature. Firstly, through the coordinate representation of curve curvature, this paper explored the condition of monotonic curvature in algebraic form. By defining the curvature decreasing (or increasing) bounding circle, the geometric sufficient and necessary conditions were given to enable decreasing (or increasing) curvature for quadratic Phillips q-Bézier curves. In the case of the shape parameter q=1, Phillips q-Bézier curves would degenerate into classical Bézier curves. Thus, the curvature monotonicity conditions of quadratic Phillips q-Bézier curves include the results of classical quadratic Bézier curves. Secondly, the paper examined the G 2 smooth condition of quadratic Phillips q-Bézier curves and the influence of parameters on the stitching curve. Thirdly, for the quadratic Phillips q-Bézier curve with given initial and final control vertices, the appropriate intermediate control vertex was selected, the range of shape parameters was obtained in the case of decreasing (or increasing) curvature, and a quadratic Phillips q-Bézier curve with decreasing (or increasing) curvature was constructed. Furthermore, a combined quadratic Phillips q-Bézier curve was constructed, which could satisfy both G 2 smooth condition and decreasing curvature. Finally, using the combined quadratic Phillips q-Bézier curve with decreasing curvature, the transition curve between two circles with inclusion relationship was constructed. The numerical examples highlight the advantages and flexibility of the combinatorial quadratic Phillips q-Bézier curve in modeling. Related Articles | Metrics

Select

Denoise method for meshes with developable features GUI Jie, CAO Li, BO Peng-bo, KOO Siu-kong 2022, 43(3): 453-460.  DOI: 10.11996/JG.j.2095-302X.2022030453 Abstract ( 70 )   PDF (2595KB) ( 63 )   Developable features can be commonly found in various meshes from different datasets. A novel denoising method was proposed for meshes with developable features. The developable features can be reflected by the sum of vertex interior angles of the mesh and well processed by L 0 computation. According to the definition of the developability of the mesh, the existing L 0 denoising algorithm was improved, and the sum of the internal angles formed in the neighborhood of a certain point was constrained to obtain the denoising effect conforming to the developable features. Compared with the existing methods, the denoising method combined with the developable features of the model can denoise more effectively while maintaining the original shape and the developable features of the model. The proposed method is particularly superior in the case of processing a large number of model data in multiple meshes datasets. Related Articles | Metrics

Select

3D model wireframe extraction method based on medial axis expression CAO Li, WU Yao, XU Yi-ke 2022, 43(3): 461-468.  DOI: 10.11996/JG.j.2095-302X.2022030461 Abstract ( 138 )   PDF (1974KB) ( 93 )   The wireframe of 3D models is widely employed in mesh retrieval, mesh simplification, and mesh reconstruction. The complexity of the existent wireframe generation methods makes it necessary to analyze the geometric features of the mesh model and calculate the surface mesh of the model to obtain the wireframe. This entails a large amount of calculation and sometimes failures to complete the wireframe. To address the problems, a 3D model wireframe extraction method based on axis representation was proposed. Firstly, the axis representation information of the 3D model was extracted, and the corner points of the axis were projected onto the surface of the 3D model. Then, according to the topological relationship of each region, the suitable corner connection relationship was selected, and the projection points were connected to form the wireframe of the model region. After the connection operation, the errors in the projection process were analyzed and corrected. Finally, the region wireframes were merged to obtain the complete wireframe of the 3D model. Through experiments on and comparisons of reconstruction errors of representative 3D mesh models in several model databases, the average re-construction quality of the proposed method is about 10% higher than the three methods, outperforming the existing wireframe extraction methods in reconstruction quality and wireframe information integrity. Related Articles | Metrics

Select

ST-Rec3D: a structure and target-aware 3D reconstruction BAI Jing, MENG Qing-liang, XU Hao, FAN You-fu, YANG Zhan-yuan 2022, 43(3): 469-477.  DOI: 10.11996/JG.j.2095-302X.2022030469 Abstract ( 128 )   PDF (2124KB) ( 93 )   Image-based 3D reconstruction is the process of producing 3D representations of an object based on its single or multiple images. Existing methods for 3D reconstruction can directly learn to transform image features into 3D representations, using encoder-decoder structure, combined with binary cross entropy function and its deformation. However, the encoder cannot extract enough information from images to reconstruct high-quality 3D shapes, resulting in inaccurate Geometric details of reconstructed 3D objects. The loss functions based on the binary cross entropy function underperforms in target perception when the voxel distribution is imbalanced, leading to problems of incompleteness such as fractures and missing in the reconstruction results. To address these problems, a structure and target-aware 3D object reconstruction framework was proposed for single-view and multi-view 3D reconstruction, named ST-Rec3D. Combined with attention mechanism, we designed an encoder with a spatial perception structure, namely structure-aware encoder. In doing so, the spatial structure information could be fully captured in the input image and the local details of the reconstructed object could be effectively perceived. The utilization of IoU loss in the 3D voxel reconstruction, in the case of uneven voxel distribution, could accurately perceive the target object to ensure the integrity and accuracy of the reconstructed object. Experimental results demonstrate that ST-Rec3D can give a significant boost to reconstruction quality and outperform state-of-the-art methods on the ShapeNet and Pix3D. Related Articles | Metrics

Select

Algorithm of rendering shadows with combined use of shadow map and deep partitioned shadow volumes WANG Wen-liang, CHEN Chun-yi, HU Xiao-juan, YU Hai-yang, TIAN Ye 2022, 43(3): 478-485.  DOI: 10.11996/JG.j.2095-302X.2022030478 Abstract ( 96 )   PDF (1806KB) ( 75 )   The shadow map algorithm can render hard shadows easily and quickly, but the hard shadow rendered by this algorithm will appear aliased in the edge area of the shadows. Affected by this, the soft shadows rendered based on the shadow map algorithm may still appear aliased in the shadow's small penumbra areas. Therefore, to render soft shadows without aliasing, it is necessary to calculate the visibility of shading points at the edge of the shadow from the point light source precisely. While deep partitioned shadow volumes algorithm can accurately calculate the visibility of the shading point to the point light source, they are less efficient than shadow map algorithms and cannot render soft shadows. In response to the above problems, we propose a shadow rendering algorithm that combines a shadow map algorithm with a deep partitioned shadow volumes algorithm. For the shading points in the shadow edge area, the deep partitioned shadow volumes algorithm is used to calculate the visibility of the shaded point to the point light source accurately, for other shading points, we use the shadow map algorithm to calculate the visibility of the shaded point to the point light source quickly. At last, we store the visibility value of the shading point in the visibility map and filter to achieve the rendering of soft shadows without aliasing. Related Articles | Metrics

Select

Visualization of ocean flow field based on unstructured triangular mesh LI Zhong-wei, XU Bin, LI Yong, GONG Kai-xuan, LIU Ge-ge 2022, 43(3): 486-495.  DOI: 10.11996/JG.j.2095-302X.2022030486 Abstract ( 146 )   PDF (10701KB) ( 99 )   The existing two-dimensional flow field visualizations are all based on structured grid flow field data. This paper proposed a strategy of ocean flow field visualization based on unstructured triangular grid, in which the flow field is expressed by streamline. The main challenge faced by streamline visualization lies in the placement of seed point, namely the initial point of streamline. To meet this challenge, a feature-guided seed point placement strategy was designed based on unstructured triangular grid. Thus, streamline initial points can be placed reasonably to facilitate the expression of flow field characteristics. A hierarchical clustering algorithm based on grid density was also designed. The grid density attribute was introduced to cluster the streamline, which was placed based on the centroid of the cluster. The visualization effect of streamline was enhanced on the premise of preserving the multi-density of FVCOM schema data. The experimental results show that the above methods can effectively retain the characteristics of the flow field, and achieve a good visualization effect for such areas as bay and riverway using the boundary fitting of FVCOM model. On this basis, taking advantage of Cesium engine, a dynamic flow field visualization application was developed based on streamline clustering data. This marks the first attempt to apply FVCOM mode data to dynamic particle flow field, and can produce a good visualization effect. Related Articles | Metrics

Select

Immersive WYTIWYG network visual analytics method WANG Song, LIU Liang, CAI Ting, ZHAO Wei-xin, WU Ya-dong 2022, 43(3): 496-503.  DOI: 10.11996/JG.j.2095-302X.2022030496 Abstract ( 94 )   PDF (4922KB) ( 73 )   Immersive network visualization possesses natural advantages in terms of spatial immersion, user engagement, and multi-dimensional perception. Inspired by users’ interaction with real-world objects, this paper proposed an immersive network visual analytics method to excavate network characteristics and association patterns based on the concept of What You Touch is What You Get (WYTIWYG). Firstly, a gesture comfort evaluation model was proposed to guide the design of gesture action, and a window state model was introduced to optimize the stability of gesture recognition. In addition, based on the binding of network analysis interaction requirements with gesture semantics, an immersive network gesture interaction paradigm was proposed. Similar to how interactions are grasped in the real world, users can employ natural interactive gestures to perform operations such as moving, highlighting, layout dimension transformation, and edge bundling in an immersive environment. Finally, case studies can verify the effectiveness of the proposed method. Related Articles | Metrics

Select

Spatiotemporal fusion network for hand gesture recognition and virtual signature system LI Yang-ke, SONG Quan-bo, ZHOU Yuan-feng 2022, 43(3): 504-512.  DOI: 10.11996/JG.j.2095-302X.2022030504 Abstract ( 51 )   PDF (4839KB) ( 51 )   Due to the coronavirus pandemic, the non-touch personal signature can reduce the risk of infection to a certain extent, which is of great significance to our daily life. Therefore, a simple and efficient spatiotemporal fusion network was proposed to realize skeleton-based dynamic hand gesture recognition, based on which a virtual signature system was developed. The spatiotemporal fusion network is mainly composed of spatiotemporal fusion modules based on the attention mechanism, and its key idea is to synchronously realize the extraction and fusion of spatiotemporal features using an incremental method. This network adopts different spatiotemporal coding features as inputs, and employs the double sliding window mechanism for post-processing in practical applications, thus ensuring more stable and robust results. Extensive comparative experiments on two benchmark datasets demonstrate that the proposed method outperforms the state-of-the-art single-stream network. Besides, the virtual signature system performs well with a single normal RGB camera, which not only greatly reduces the complexity of the interaction system, but also provides a more convenient and secure approach to personal signature. Related Articles | Metrics

Digital Design and Manufacture

Select

Study on the penetration resistance of steel/aluminum composite target with isoplanar density ZHANG Chen, ZHANG Xiao-ping, LIU Su-su, QU Chang, ZHANG Fu-bao, ZHONG Jian-lin, CAO Yan-feng 2022, 43(3): 513-521.  DOI: 10.11996/JG.j.2095-302X.2022030513 Abstract ( 121 )   PDF (2693KB) ( 53 )   The anti-penetration performance of metal target plate has always been a focus of research. In order to clarify the influence of structural parameters on the vertical penetration resistance of steel/aluminum composite structures, numerical calculation models of steel and aluminum composite structures were established and verified by ballistic impact tests. Based on the numerical model, the influences of the combination form of steel-aluminum, al-steel and target spacing on the penetration resistance of the composite target with the same thickness were analyzed. The simulation results show that under the same thickness, the combination form of steel plate in front was better than the target plate with aluminum plate in front, but the different combination form of laminar spacing has little influence on its anti-penetration performance. In addition, the influence of the thickness ratio of steel to aluminum on the penetration resistance of the composite target plate at different initial velocities was compared. The penetration resistance of the composite target plate decreases first and then increases with the increase of the thickness ratio, and finally reaches a stable value. There is a low point with the worst penetration resistance during the change process, which should be avoided as much as possible in structural design. Finally, the trajectory limit formula of musketeer projectile penetrating steel-aluminum composite structure was obtained by fitting the simulation data according to THE R-I formula. The research results can provide reference for the design of anti-penetration performance of composite target plate and the protection of personnel and materials. Related Articles | Metrics

BIM/CIM

Select

Identification of the plane irregularity of structures based on BIM and deep learning JIANG Liu, SHI Jian-yong, FU Gong-yi, PAN Ze-yu, WANG Chao-yu 2022, 43(3): 522-529.  DOI: 10.11996/JG.j.2095-302X.2022030522 Abstract ( 184 )   PDF (830KB) ( 165 )   Compliance checking on earthquake-resistance is essential for architectures, especially for high-rise buildings. Current checking methods rely heavily on human efforts. In particular, as one of the critical checking contents, the identification of the plane irregularity of structures is time-consuming and error pone, because the building plane designs are becoming increasingly complex. The identification of the plane irregularity of structures could be regarded as a plane classification problem where the regular planes are identified as normal samples and the irregular ones as abnormal samples. Considering the unbalanced distribution of regular and irregular planes in construction projects and adopting the idea of an anomaly detection model, a methodology for the identification of the plane irregularity was proposed based on Building Information Modeling (BIM) and deep learning. Firstly, the building plane of BIM model was obtained by Boolean intersection operation between geometric objects. After image processing, the building plane could be converted into a building plane contour map. Finally, the trained anomaly detection model was executed on the contour map to yield the identification results. The experimental results show that in comparison with the traditional image classification models, the new one following the idea of an anomaly detection model can increase the identification rate of irregular building planes by 15%, more readily meeting the needs of practical applications. Related Articles | Metrics

Select

Disaster tweets classification method based on pretrained BERT model LIN Jia-rui, CHENG Zhi-gang, HAN Yu, YIN Yun-peng 2022, 43(3): 530-536.  DOI: 10.11996/JG.j.2095-302X.2022030530 Abstract ( 260 )   PDF (578KB) ( 147 )   Social media has become an important medium for the release and dissemination of disaster information, the effective identification and utilization of which is of great significance to disaster emergency management. Given the shortcomings of the traditional text classification model, a disaster tweet classification method was proposed based on the pre-trained model of bidirectional encoder representations from transformers (BERT). After data cleaning and preprocessing, this study constructed a text classification model based on long short-term memory-convolutional neural network (LSTM-CNN) through comparative analysis, based on BERT. Experiments on the tweet datasets of the Kaggle competition platform showed that the proposed classification model outperforms the traditional Naive Bayesian classification model and the common fine-tuning model, with the recognition rate up to 85%. This study could shed significant light on enhancing the identification accuracy of real disaster information and the efficiency of disaster emergency response. Related Articles | Metrics

Industrial Design

Select

Multistage decision-making method of product industrial design by integrating Bayesian network and prospect theory YANG Yan-pu, LEI Zi-jing, LAN Chen-xin, WANG Xin-rui, GONG Zheng 2022, 43(3): 537-547.  DOI: 10.11996/JG.j.2095-302X.2022030537 Abstract ( 88 )   PDF (1283KB) ( 62 )   In view of the uncertainty in the decision-making process of product industrial design and the difficulty of accurately describing the result of overall decision-making through a single design decision-making stage, a three-parameter interval gray number was introduced to describe the opinions of decision makers, and a Bayesian network (BN) model was constructed to learn the users’ decision-making information about the existing mature products in the market. In doing so, the state distribution probability of the target product design schemes on each index could be obtained. To reflect the psychological behavior of decision-makers’ perception of the relative gains and losses about design schemes, the prospect theory (PT) and BN were integrated to construct the prospect functions of product industrial design schemes in different decision-making stages. In addition, an optimization model was built based on the cognitive progression assumption to calculate the weights of multistage decision-making information in product industrial design. The comprehensive prospect values were computed to help identify the pros and cons of the product industrial design schemes. The effectiveness of the method was verified through the case study of the multistage decision-making information fusing of numerical control grinder industrial design. Results show that the proposed method can help introduce the multistage opinion preference of users to estimate the probability distribution of design decision-making indexes, realize decision-making information fusion with prospect values of product industrial design schemes, and improve the quality of design decision-making in an overall and scientific way. Related Articles | Metrics

Select

Cobweb grey target decision-making model of multi-Kansei image in product form ZHANG Shu-tao, WANG Shi-jie, LIU Shi-feng, LI Wei-xing 2022, 43(3): 548-557.  DOI: 10.11996/JG.j.2095-302X.2022030548 Abstract ( 103 )   PDF (1129KB) ( 67 )   In view of the difficulty in decision-making for multi-Kansei image design scheme in concept design, a cobweb grey target decision-making method was proposed for multi-Kansei image in product form. Firstly, the cognitive data of the design subjects was obtained by Kansei engineering, and the comprehensive evaluation model based on the cognitive data of the design subjects was constructed using the entropy weight method and game theory. Meanwhile, the weight relationship of each Kansei image was determined based on the comprehensive evaluation data of the design subjects of each image. Secondly, multiple evolution schemes were artificially selected from the product form evolution system, and the spider chart was used to represent the Kansei image relationship of each evolution scheme. A cobweb grey target decision-making model of multi-Kansei image was constructed to calculate the decision-making coefficient, and the relative optimal scheme that conformed to the cognition of the design subjects was obtained by comparing and sorting them. Finally, the feasibility of the decision-making model was verified by applying the grey correlation analysis method. The model can help designers quickly and accurately determine the multi-Kansei image scheme in the design decision-making stage, which provides a new theory and method for the multi-Kansei image decision-making of product scheme. Related Articles | Metrics

Published as

Select

Published as 3, 2022 2022, 43(3): 558-558.  Abstract ( 67 )   PDF (73758KB) ( 454 )   Related Articles | Metrics