Search

Extracting geometric descriptors in 3D vision is the first step. It plays an important role in 3D registration, 3D reconstruction, and other applications. The success of many 3D tasks is closely related to whether the geometric descriptor has accurate characteristics. Today, the main methods are divided into manual production and neural network learning. The applicability of descriptors is limited to a low-level point, corner, edge, and fixed neighborhood features. For this, we use the class attention of the point cloud. In order to extract class attention, the graph clustering approach is utilized. It can collect points with similar structures and divide regions dynamically. While maintaining rotation invariance, features can enhance their fit to the original data. Point attention and edge attention are used to describe the structural characteristics of point clouds. We combine the three attentions indicated before to improve the features obtained by the PointNet decoder. This feature can dynamically reflect the structure of the point cloud, which includes both soft shape information and rich detail information. Finally, the 3D descriptors are extracted with the FoldingNet decoder. Our method is validated on both indoor and outdoor datasets. The accuracy of the final result is improved by two percentage points.

To solve the problem of low detection accuracy due to the loss of detailed information when extracting pavement crack features in traditional U-shaped networks, a pavement crack detection method based on multiscale attention and hesitant fuzzy set (HFS) is proposed. First, the encoding-decoding structure is used to construct a pavement crack segmentation network, ResNeXt50 is used to extract features in the encoding stage, and a multiscale feature fusion module (MFF) is designed to obtain multiscale context information. Second, in the decoding stage, a high-efficiency dual attention module (EDA) is used to enhance the ability of capturing details of the cracks while suppressing background noise. Finally, the membership degree of the crack is calculated based on the advantages of the HFS in multiattribute decision-making to obtain the similarity of the crack, and the binary image after segmentation is judged by the hesitation fuzzy measure. The experiment was conducted on the public road crack dataset Crack500. In terms of segmentation performance, the evaluation indexes Intersection over Union (IoU), Precision, and Dice coefficients of the proposed network reached 55.56%, 74.26%, and 67.43%, respectively; in terms of classification performance, for transversal and longitudinal cracks, the classification accuracy was 84% ± 0.5%, while the block and the alligator were both 78% ± 0.5%. The experimental results prove that the crack details detected by the proposed method are more abundant, and the image detection effect of complex topological structures and small cracks are better.

Video summarization for educational scenarios aims to extract and locate the most meaningful frames from the original video based on the main contents of the lecture video. Aiming at the defect of existing computer vision-based lecture video summarization methods that tend to target specific scenes, a summarization method based on content detection and tracking is proposed. Firstly, DBNet is introduced to detect the contents such as text and mathematical formulas in the static frames of these videos, which is combined with the convolutional block attention module (CBAM) to improve the detection precision. Then, frame-by-frame data association of content instances is performed using Kalman filtering, the Hungarian algorithm, and appearance feature vectors to build a tracker. Finally, video segmentation and key frame location extraction are performed according to the content instance lifelines and content deletion events constructed by the tracker, and the extracted key frame groups are used as the final video summary result. Experimenting on a variety of scenarios of lecture video, the average precision of content detection is 89.1%; the average recall of summary results is 92.1%.

The easy availability and usability of photo-editing tools have increased the number of forgery attacks, primarily splicing attacks, thereby increasing cybercrimes. Because of an existing image-splicing tamper-detection algorithm based on deep learning with high model complexity and weak robustness, a multiscale fusion lightweight model for image-splicing tamper detection is proposed. For the above problems and to improve MobileNetV2, the structural block of the classification part of the original network structure was removed, the stride of the sixth largest structural block of the network was changed to 1, the dilated convolution was used instead of downsampling, and the features extracted from the second and third large structural blocks in the network were downsampled with maximal pooling; then, the constraint on the backbone network was increased by jumping connections. Combined with the pyramid pooling module, the acquired feature layers were divided into regions of different sizes for average pooling; then, all feature layers were fused. The experimental results show that it had a low number of parameters and required a small amount of computation, achieving 91.0% and 96.4% precision on CASIA and COLUMB, respectively, and 83.2% and 88.1% F-measure on CASIA and COLUMB, respectively.

Few-shot text classification aims to recognize new classes with only a few labeled text instances. Previous studies mainly utilized text semantic features to model the instance-level relation among partial samples. However, the single relation information makes it difficult for many models to address complicated natural language tasks. In this paper, we propose a novel hesitant fuzzy graph neural network (HFGNN) model that explores the multi-attribute relations between samples. We combine HFGNN with the Prototypical Network to achieve few-shot text classification. In HFGNN, multiple relations between texts, including instance-level and distribution-level relations, are discovered through dual graph neural networks and fused by hesitant fuzzy set (HFS) theory. In addition, we design a linear function that maps the fused relations to a more reasonable range in HFGNN. The final relations are used to aggregate the information of neighbor instance nodes in the graph to construct more discriminative instance features. Experimental results demonstrate that the classification accuracy of the HFGNN-based Prototypical Network (HFGNN-Proto) on the ARSC, FewRel 5-way 5-shot, and FewRel 10-way 5-shot datasets reaches 88.36%, 94.45%, and 89.40%, respectively, exceeding existing state-of-the-art few-shot learning methods.

Vehicle reidentification has important applications in intelligent monitoring systems. However, due to many factors, such as inaccurate vehicle image detection and viewing angle changes, distinguishing features cannot be effectively obtained when the vehicle is reidentified. To improve the recognition ability and robustness of vehicle reidentification, this study proposes a new multiattention part alignment network (MAPANet). The network uses different channels in the feature map to perceive different characteristics of the image clustering of the channels and achieves fine-grained attention to the vehicle. It can automatically locate the distinguishing subregions in the vehicle image and avoid the need for a large number of additional manual pretreatment steps. Moreover, an unsupervised reranking method based on multiple metrics is proposed. The k-reciprocal encoding algorithm can optimize the performance of the sorted list in the reordering problem, recalculate the interclass and intraclass distances of vehicle pictures, and improve sorting results. The experiments in this paper are carried out on the VeRi-776 and VehicleID datasets, and the mean average precision (mAP) results on the two datasets are 72.83% and 75.25%, respectively.

In recent years, the rapid growth of multimodal information has become an important factor affecting the results of sentiment analysis. However, a few state-of-the-art works take into account the multimodal features and sentiment fuzziness. To this end, a fuzzy method is proposed for assessing sentiment intensity in this paper. Firstly, based on the visual-text conversion network (CNN-LSTM), as well as sentiment optimization through SentiBank and SentiBridge, the visual features are normalized to the text features. At the same time, the emotional features of the extracted audio will be predicted by the random forest algorithm. Subsequently, the sentiment characteristics are processed by dual hesitant fuzzification to form positive and negative sentiment intensity factors. Finally, a classification method, that is, MD-HFCE (multilayer dual hesitant fuzzy comprehensive evaluation), fuzzy comprehensive evaluation method improved by Mamdani fuzzy reasoning, is proposed to realize the multifeature fuzzy sentiment classification based on the comprehensive sentiment dictionary. The classification results are applicable to the topics of sentiment monitoring. The experimental results show that the proposed algorithm can effectively realize feature integration and improve the average sentiment classification accuracy of multimodal blogs to 82.2%.

Aiming at the problem that the mathematical expressions in unstructured text fields of documents are hard to be extracted automatically, rapidly and effectively, a method based on Hidden Markov Model (HMM) is proposed. Firstly, this method trained the HMM model through employing the symbol combination features of mathematical expressions. Then, some preprocessing works such as removing labels and filtering words were carried out. Finally, the preprocessed text was converted into an observation sequence as the input of the HMM model to determine which is the mathematical expression and extracts it. The experimental results show that the proposed method can effectively extract the mathematical expressions from the text fields of documents, and also has the relatively high accuracy rate and recall rate.

In recent years, a growing number of math contents are available on the Web. When conventional search engines deal with mathematical expressions, the two-dimen- sion-al structure of mathematical expressions is lost, which results in a low performance of math retrieval. While the retrieval technology specifically designed for mathematical expressions is not mature currently. Aiming at these problems, an improved mathematical expression indexing and matching method was proposed through employing full text index method to deal with the two-dimensional structure of mathematical expressions. Firstly, through the fully consideration of LaTeX formulae’ characteristics, a feature representation method of mathematical expressions and a clustering method of feature keywords were put forward. Then, an improved inter-relevant successive trees index model was applied to the construction of the mathematical expression index, in which the cluster algorithm of mathematical expression features was employed to solve the problem of the quantity growth of the trees in processing large amount of formulae. Finally, the matching algorithms of mathematical expressions were given which provide four query modes called exact matching, compatible matching, sub-expression matching and fuzzy matching. In browser/server mode, 110027 formulae were used as experimental samples. The index file size was 29.02 Mb. The average time of retrieval was 1.092 seconds. The experimental result shows the effectiveness of the method.

We present distinct antiblockade features of strongly interacting 64D 5/2 Rydberg atoms employing a two-color excitation scheme. The first color (pulse A) is set to resonantly excite a few seed Rydberg atoms, each of which establishes a blockade region due to the long-range multipole interactions. The second color (pulse B) is blue detuned so that the multipole-interaction-induced shifts of certain atoms are well compensated to result in the antiblockade effect. We find in particular that a few seed atoms can lead to a remarkable difference of the Rydberg excitation in the presence of pulse B for a wide range of blue detuning. Relevant dynamics of this antiblockade excitation is also investigated by varying the pulse-B duration for a fixed blue detuning, further confirming the facilitation process of Rydberg excitation accompanied by a saturation effect. These experimental results can be well recovered by theoretical simulations based on a multilevel two-body model.

In order to simulate the hesitancy and uncertainty associated with impression or vagueness, a decision maker may give her/his judgments by means of hesitant fuzzy preference relations in the process of decision making. The study of their consistency becomes a very important aspect to avoid a misleading solution. This paper defines the concept of additive consistent hesitant fuzzy preference relations. The characterizations of additive consistent hesitant fuzzy preference relations are studied in detail. Owing to the limitations of the experts’ professional knowledge and experience, the provided preferences in a hesitant fuzzy preference relation are usually incomplete. Consequently, this paper introduces the concepts of incomplete hesitant fuzzy preference relation, acceptable incomplete hesitant fuzzy preference relation, and additive consistent incomplete hesitant fuzzy preference relation. Then, two estimation procedures are developed to estimate the missing information in an expert's incomplete hesitant fuzzy preference relation. The first procedure is used to construct an additive consistent hesitant fuzzy preference relation from the lowest possible number, (n − 1), of pairwise comparisons. The second one is designed for the estimation of missing elements of the acceptable incomplete hesitant fuzzy preference relations with more known judgments. Moreover, an algorithm is given to solve the multi-criteria group decision making problem with incomplete hesitant fuzzy preference relations. Finally, a numerical example is provided to illustrate the solution processes of the developed algorithm and to verify its effectiveness and practicality.

In this paper, we develop a decision support model that simultaneously addresses the consistency and consensus for group decision making based on hesitant fuzzy preference relations. The concepts of a consistency index and a consensus index are introduced. Two convergent algorithms are proposed in the developed support model. The first algorithm is used to convert an unacceptable hesitant fuzzy preference relation to an acceptable one. The second algorithm is utilized to help the group reach a predefined consensus level. The main characteristic of the developed model is that it makes each hesitant fuzzy preference relation maintain acceptable consistency when the predefined consensus level is achieved. Several illustrative examples are given to illustrate the effectiveness and practicality of the developed model.

Hesitant fuzzy sets, permitting the membership of an element to be a set of several possible values, can be used as an efficient mathematical tool for modeling people's hesitancy in daily life. The aim of this paper is to present a consensus support model for group decision making with hesitant fuzzy information. This model is composed of two processes: a consensus process and a selection process. The consensus process is carried out to reach a high level of consensus among experts' opinions before applying a selection process. We first aggregate the hesitant fuzzy decision matrix into a group decision matrix by using the additive aggregation (AA) operator. Then the consensus measure is used to design a feedback mechanism that generates advice to the experts on how they should change their preferences to obtain a solution with a high consensus degree. In the selection process, based on the consentaneous group decision matrix, the additive weighted aggregation (AWA) operator is utilized to derive the overall attribute values of alternatives, by which the most desirable alternative can be found out. Finally, a practical example is proposed to illustrate the application of the proposed model.

Purpose: Accurate detection of pulmonary nodules remains a technical challenge in computer-aided diagnosis systems because some nodules may adhere to the blood vessels or the lung wall, which have low contrast compared to the surrounding tissues. In this paper, the analysis of typical shape features of candidate nodules based on a shape constraint Chan–Vese (CV) model combined with calculation of the number of blood branches adhered to nodule candidates is proposed to reduce false positive (FP) nodules from candidate nodules. Methods: The proposed scheme consists of three major stages: (1) Segmentation of lung parenchyma from computed tomography images. (2) Extraction of candidate nodules. (3) Reduction of FP nodules. A gray level enhancement combined with a spherical shape enhancement filter is introduced to extract the candidate nodules and their sphere-like contour regions. FPs are removed by analysis of the typical shape features of nodule candidates based on the CV model using spherical constraint and by investigating the number of blood branches adhered to the candidate nodules. The constrained shapes of CV model are automatically achieved from the extracted candidate nodules. Results: The detection performance was evaluated on 127 nodules of 103 cases including three types of challenging nodules, which are juxta-pleuralmore » nodules, juxta-vascular nodules, and ground glass opacity nodules. The free-receiver operating characteristic (FROC) curve shows that the proposed method is able to detect 88% of all the nodules in the data set with 4 FPs per case. Conclusions: Evaluation shows that the authors’ method is feasible and effective for detection of three types of nodules in this study.« less

The mathematical formula extraction is the prerequisite of formula structure analysis, recognition and retrieval. This paper studies the formula extraction method for the printed Chinese scientific and technical document images, proposes a criterion based on connected component run-length feature to estimate formulae in text lines, and then improves the formula location method based on rules. The connected component run-length's change regularity was analyzed firstly for all symbols in a text line. Then Change-rate threshold was set to estimate whether there is formula in this line. Finally, improved formula extraction method was given. The experimental results on the samples collected from printed Chinese scientific and technical documents showed that the proposed method is effective in estimate the embedded formula, and improves the accuracy of the formula location.

Facing the demand of providing retrieval result with rich semantic information for users in math searching, mathematical formulas of LaTeX are usually converted to Content MathML. For the problem of ambiguity formulas in the process of conversion, a method of semantic disambiguation for mathematics formulas which is based on the operator context was proposed. At first, ambiguity operator was found according to the ambiguity operator mapping table. Then, the ambiguity operator context is got through the array traversal. At last, the specific meaning was conjectured according to the ambiguity operator context. The experimental results show that compared with the type-system the method can make up for its disadvantages in simple formulas aspect and gets a higher average accuracy. In practical application, this method can effectively solve the problem of ambiguity formulas in the process of conversion.

The aim of this study was to propose an algorithm for three-dimensional projection onto convex sets (3D POCS) to achieve super resolution reconstruction of 3D lung computer tomography (CT) images, and to introduce multi-resolution mixed display mode to make 3D visualization of pulmonary nodules. Firstly, we built the low resolution 3D images which have spatial displacement in sub pixel level between each other and generate the reference image. Then, we mapped the low resolution images into the high resolution reference image using 3D motion estimation and revised the reference image based on the consistency constraint convex sets to reconstruct the 3D high resolution images iteratively. Finally, we displayed the different resolution images simultaneously. We then estimated the performance of provided method on 5 image sets and compared them with those of 3 interpolation reconstruction methods. The experiments showed that the performance of 3D POCS algorithm was better than that of 3 interpolation reconstruction methods in two aspects, i.e., subjective and objective aspects, and mixed display mode is suitable to the 3D visualization of high resolution of pulmonary nodules.

Concerning the problems of the comparative translation is not accurate in machine translation and the useful information is unable to match in information retrieval, a liheci word sense disambiguation method is adopted and a classifier model is established using Support Vector Machines(SVM). To improve the accuracy of the liheci word sense disambiguation, it extracts not only local words(LW), local part-of-speeches(POS), local words and part-of-speeches (LWP) but also the middle insert part of the separated form as disambiguation features according to the characteristics of liheci, When the text features are converted to feature vectors for improving on boolean weight method, we can fix feature weights of some type in turn and change the other two types' to verify the disambiguation effect of three kinds of features, respectively. The results show that the effect of LW, LWP is higher than POS. Setting higher feature weights to LW and LWP, the disambiguation accuracy can effectively improve.

Accurate detection of pulmonary nodules remains a technical challenge in computer-aided diagnosis systems because some nodules may adhere to the blood vessels or the lung wall, which have low contrast compared to the surrounding tissues. In this paper, the analysis of typical shape features of candidate nodules based on a shape constraint Chan-Vese (CV) model combined with calculation of the number of blood branches adhered to nodule candidates is proposed to reduce false positive (FP) nodules from candidate nodules.The proposed scheme consists of three major stages: (1) Segmentation of lung parenchyma from computed tomography images. (2) Extraction of candidate nodules. (3) Reduction of FP nodules. A gray level enhancement combined with a spherical shape enhancement filter is introduced to extract the candidate nodules and their sphere-like contour regions. FPs are removed by analysis of the typical shape features of nodule candidates based on the CV model using spherical constraint and by investigating the number of blood branches adhered to the candidate nodules. The constrained shapes of CV model are automatically achieved from the extracted candidate nodules.The detection performance was evaluated on 127 nodules of 103 cases including three types of challenging nodules, which are juxta-pleural nodules, juxta-vascular nodules, and ground glass opacity nodules. The free-receiver operating characteristic (FROC) curve shows that the proposed method is able to detect 88% of all the nodules in the data set with 4 FPs per case.Evaluation shows that the authors' method is feasible and effective for detection of three types of nodules in this study.

Left ventricular (LV) epicardium segmentation in cardiac magnetic resonance images (MRIs) is still a challenging task, where the a-priori knowledge like those that incorporate the heart shape model is usually used to derive reasonable segmentation results. In this paper, we propose a sparse group composition (SGC) approach to model multiple shapes simultaneously, which extends conventional sparsity-based single shape prior modeling to incorporate a-priori spatial constraint information among multiple shapes on-the-fly. Multiple interrelated shapes (shapes of epi- and endo-cardium of myocardium in the case of LV epicardium segmentation) are regarded as a group, and sparse linear composition of training groups is computed to approximate the input group. A framework of iterative procedure of refinement based on SGC and segmentation based on deformation model is utilized for LV epicardium segmentation, in which an improved shape-constraint gradient Chan-Vese model (GCV) acted as deformation model. Compared with the standard sparsity-based single shape prior modeling, the refinement procedure has strong robust for relative gross and not much sparse errors in the input shape and the initial epicardium location can be estimated without complicated landmark detection due to modeling spatial constraint information among multiple shapes effectively. Proposed method was validated on 45 cardiac cine-MR clinical datasets and the results were compared with expert contours. The average perpendicular distance (APD) error of contours is 1.50 ± 0.29 mm, and the dice metric (DM) is 0.96 ± 0.01. Compared to the state-of-the-art methods, our proposed approach appealed competitive segmentation performance and improved robustness.

We propose a simple but effective tracking algorithm for non-rigid objects with geometric appearance changes. The discriminative features of the object are adaptively selected according to their descriptive ability. To adapt to the geometric changes, we use a deformable rectangle to represent the object, and use Markov Chain Monte Carlo-based Particle Filter (MCMC-PF)to estimate the state of the object in a restricted four dimensional space. Experimental results show that the proposed tracking algorithm has ideal performance.

PDF documents are the important information resources for a mathematical expression retrieval system. As a major component of PDF documents, the image objects must be converted to coded form with the help of character recognition and document analysis technology firstly for content based searching. Therefore, the quality of these images becomes the key factor which decides the correctness in this conversion process. Considering the characteristics of PDF images and mathematical expressions, a preprocessing and analyzing method was proposed which includes the modules of PDF image extraction, graying, binarization, denoising, skew correction and layout parameter detection. The features of mathematical expressions were adequately considered to avoid the information loss in image converting process and the adverse interference both to the analysis and correction process resulted from formulas. The experimental results show that the method is effective in improving the accuracy and efficiency of document image recognition, analysis and retrieval. DOI : http://dx.doi.org/10.11591/telkomnika.v12i6.5440

PDF document gains its popularity in information storage and exchange. With more and more documents, especially the scientific documents, available in PDF format, extracting mathematical expressions in PDF documents becomes an important issue in the field of mathematical expression recognition and retrieval. In this paper, we proposed a method of extracting mathematical components directly from PDF documents rather than cooperating indirectly with corresponding images converted from PDF files. Compared with traditional image-based method, the proposed method makes full use of the internal information of PDF documents such as font size, baseline, glyph bounding box and so on to extract the mathematical characters and their geometric information. The experimental result shows the method could meet the needs of the following processing of mathematical expressions such as formula structural analysis, reconstruction and retrieval, and has a higher efficiency than traditional image-based ways.

There are various kinds of methods on activated regions detection, including model-driven method and data-driven method, univariate method and multivariate method, frequency domain analysis and time-domain analysis etc. We investigated the problems of principal component analysis applied to activated regions detection,an autocovariance based principal component analysis method was proposed. Firstly,the time series were converted to the autocovariance series, and then the principal component analysis was employed. Meanwhile, the tactic of principal component selection was discussed. The validity of the proposed method was illustrated by experiments on a synthetic dataset and a real dataset. It was shown that the error rate of the new approach was lower compared with the principal component analysis itself.

The performance of the cooling water system strongly affects the energy efficiency of centralized chiller plant. However, many cooling water systems are not well operated. Energy diagnosis is necessary to enhance the system efficiency. The existing energy diagnosis methods rely on expert experience and are time-consuming and expensive. Therefore, this chapter presents an indicator-based diagnosis method for the cooling water systems. In this method, four indicators are defined to quantitatively evaluate the performance of the system. These indicators can be calculated by a standard process. By reading the indicators, building managers and operators can easily find the energy-waste problems and estimate the energy-saving potential. This method has been applied in a large-scale commercial building in Hong Kong. This application is presented in detail to show this method.

Towards the needs of retrieving mathematical formulas, this paper details the semantic descriptors of MathML by utilizing the extensibility of XML and the characteristics of mathematical formulas through the analysis of the existing mathematical formula marking languages. We define a description attribute on the basis of MathML for conversion and retrieval of mathematical formulas in various formats, and also realize the conversion between the extended MathML and LaTeX. Experiments show that this method can strengthen the capability to describe the semantic information of formulas and improve the precision and efficiency of conversion and retrieval of mathematical formulas.

In this paper, the theoretical analysis, field testing, ANSYS finite element analysis software was adopted to analyze the process of projectile penetration. Which established the body of missile and concrete target to obtain the acceleration distribution of missile. So as to provide theoretical support of missile structure, damage assessment, verify the feasibility of solving problems in high-speed penetration response analysis based on ANSYS/LS-DYNA non-linear dynamic finite element analysis.

Electric circuit courses are the major basic subjects about hardware curriculum of computer majors in undergraduate college. The teaching quality of these courses would influence the students’ learning effect of the following hardware curriculum. In this paper, several problems exist in the teaching and learning activities of electric circuit courses are introduced firstly. And then, some teaching and learning methods are discussed in order to improve the students’ learning quality of electric circuit courses.

Touching symbols cannot be segmented accurately which is one of the main reasons for false recognition of typeset mathematical expressions. To solve this problem, a technique which is based on contour feature is presented for the segmentation of touching symbols in mathematical expressions. Partitioning paths can be constructed ground on contour feature and the ratio of the width and the height, and then touching symbols are segmented. Experimental results show the effectiveness on the accuracy improvement of mathematical expression recognition.

At present, the research on the analysis of structure in mathematical expressions recognition is rare. For this, a method of structural analysis of mathematical expressions which combine baseline and operator range with syntax analysis is presented in the printed mathematical expression recognition. The experiments indicate that this method can obtain favorable results.