最新刊期
卷 48 ， 期 4 ， 2020
- 摘要：Video object segmentation (VOS) is a research hotspot in the field of computer vision. Traditional VOS based on deep learning fine-tunes the deep network online, which leads to long time-consuming segmentation and is difficult to meet real-time requirements. Therefore, we propose a fast VOS method. First, the weight-shared siamese encoder subnet maps the reference stream and the target stream to the same feature space; so that the same objects have similar features. Then,the global feature extraction subnet matches the features similar to the given object to locate the object. Finally, the decoder subnet restores the object features and gets edge information by connecting the low-level features of target stream to output the mask. Experiments on public benchmark datasets show that our method improves the speed significantly and achieves good performance.关键词：video object segmentation;computer vision;deep learning;siamese network;feature space112|319|4更新时间：2025-07-08
- 摘要：In order to address the issue of motion boundary blurring caused by the complex scenes, large displacement and motion occlusion, this paper proposes a binocular scene flow estimation method based on semantic segmentation. Firstly, by using the image semantic information, we classify the image regions into several categories with semantic labels through convolutional neural networks. Then we plan the motion models of various image regions according to the different semantic categories and compute the optical flow and disparity under the prior knowledge of semantic information. Secondly, we apply the superpixel segmentation to the input image and couple the optical flow and disparity information via least squares method to solve the motion parameters of each superpixel patch. Finally, we add the boundary information of semantic segmentation constraint to the optimization energy function, and estimate the scene flow by updating the mappings of pixels-to-superpixel and superpixel-to-plane. We evaluate the proposed approach and some state-of-the-art methods on the KITTI 2015 database to conduct a comparison experiment. The experimental results demonstrate that our method has high accuracy and good robustness, and especially has significant benefit of boundary preserving in the areas of complex scene, motion occlusion and motion boundary.关键词：semantic segmentation;scene flow;deep learning;binocular stereo matching;least squares method;superpixel segmentation;motion occlusion;edge protection75|164|2更新时间：2025-07-08
- 摘要：In order to assist doctors in planning treatment and review programs for non-small cell lung cancer (NSCLC) patients, a prognostic survival analysis method based on CT radiomics was proposed. First, we segmented the tumor areas in the lung CT images. Then, we extracted and optimized the radiomics features. Finally, the optimized features and the patients’ prognosis survival were taken as input, and the prognostic analysis model was constructed by using machine learning method to predict the prognosis survival time range of the patients. The data of 124 NSCLC patients were selected and the clinical significance of 3-year survival was used as the predictive limit to predict the prognosis survival time range. The experimental results showed the prediction accuracy of the model reached 91.9%, which could effectively assist doctors to carry out more accurate assessment and develop more personalized treatment and review programs for NSCLC patients.关键词：computed tomography;non-small cell lung cancer;radiomics features;prognostic analysis35|103|1更新时间：2025-07-08
- 摘要：To solve the problem of face recognition under unlimited conditions, a simple network structure named Inception Module Incorporated Siamese Convolutional Neural Networks (IMISCNN) was designed, which was suitable for small-scale data sets. On the basis of making full use of the Siamese structure to effectively reduce external interference and avoid over-fitting, inception module was incorporated to the Siamese network to extract richer features. Furthermore, a cyclical learning rate strategy was adopted to accelerate the convergence of the model. Simulation results on the CASIA-webface and Extended Yale B standard face database showed that the recognition accuracy of IMISCNN was significantly improved compared with other face recognition algorithms.关键词：face recognition;siamese convolutional neural networks;inception module;cyclical learning rate73|158|5更新时间：2025-07-08
- 摘要：3D object recognition with shape changes is a challenging task. The irregular data structure of the mesh model prevents the operation of the conventional convolution, which brings difficulties to feature extraction of the 3D non-rigid objects. In this paper, we propose a method of mesh convolution for 3D non-rigid objects to extract shape features and use them for classification. Firstly, we obtain the distribution of typical patch shapes by the mesh convolution. Then, we model the spatial co-occurrence relationship by Markov chains to complete the global feature description. Finally, we use the support vector machine to classify the 3D objects. Our method adopts the continuous polynomial function as the convolution kernel for the irregular data structure, and learn the kernel by an unsupervised learning method. Experimental results on the standard non-rigid 3D model datasets show our method can effectively extract the features and achieve classification accuracy of 92.88% on SHREC10 and 96.54% on SHREC15,respectively.关键词：non-rigid 3D model;mesh convolution;3D shape feature;support vector machine48|119|0更新时间：2025-07-08
- 摘要：In order to analyze and predict the performance of CUDA program kernel and guide parallel program design and performance optimization, a performance prediction framework is proposed. This paper starts with the GPU programming model and hardware architecture details, with warp as the research unit. By integrating hardware and software factors closely related to GPU program time, high-level performance-related features such as device parallel space idle degree (DPSID), number of streaming multiprocessor warp (NSMW) are defined. Based on the above features, a framework for evaluating the execution time of kernel functions under different problem sizes and execution configurations is built for thread load balancing GPU programs. The principle of optimizing configuration parameters of kernel function execution is put forward to guide optimizing program performance. The experimental results show that the average prediction accuracy of the framework is 89% and 94% in the two scenarios, respectively.关键词：performance prediction;thread warp;device active warps;parallel effect;performance features;execution configuration optimization54|100|0更新时间：2025-07-08
- 摘要：In wireless networks-on-chip, wireless communication congestion and fault have a severe impact on the communication efficiency of the entire network-on-chip. Therefore, this paper proposes a fault-tolerant routing algorithm for wireless communication congestion and faults. Firstly, a wireless communication congestion and fault aware model is designed. The model can get the congestion and fault information of the wireless node communication pair, encode it and send it to the routers in subnet. Then the router in the subnet determines whether the data packet uses wireless transmission according to the received wireless node communication pair status information. Experiments show that the proposed scheme can guarantee lower network delay and higher network throughput under smaller additional area overhead and power consumption than the comparison schemes, and exhibit good performance to tolerant permanent faults of wireless node communication pairs.关键词：wireless network-on-chip;wireless node communication pair;fault tolerance;congestion control;routing algorithm48|104|0更新时间：2025-07-08
- 摘要：A MMSE feature extraction method based on MEMD was proposed to analyze multi-modal signals and evaluate the static balance ability of human body. First, the human multi-mode signal was collected. It was adaptively decomposed by multi-empirical mode from which a series of (IMFs) were obtained. The best IMF components were selected according to the T-test and correlation coefficients which was used for signal reconstruction. The multivariate multi-scale entropy algorithm was used to extract the features. Finally, K-means and support vector machine were used to compare with this paper’s methods about dealing with human body static balance problem, which was used to evaluate the optimal feature extraction method. Results shows that MMSE based on MEMD and support vector machine are optimal for feature extraction and classification in this paper.关键词：assessment of static equilibrium;multimodal signal;MEMD;MMSE54|74|5更新时间：2025-07-08
- 摘要：For the problem of data security in the cloud storage system, a self-healing Byzantine Quorum system for the cloud storage security is proposed. In this system, the virtual machines are used as back-end storage devices to construct the virtual storage node. The diverse operating systems, dynamic migration and rapid generation mechanism of virtual machines are introduced to build dynamic and heterogeneous storage system architecture. On the basis of the Byzantine fault tolerance threshold, the concept of self-healing threshold is presented and several security protocols are devised to achieve automated anomaly detection and storage node recovery. The experimental results show that the proposed cloud storage system is greatly robust and can effectively improve the security of stored data.关键词：cloud storage;data security;Byzantine Quorum system;self-healing mechanism33|77|0更新时间：2025-07-08
- 摘要：Incremental attribute reduction is a data mining method for dynamic environment. The incremental attribute reduction algorithm already proposed is only applicable to symbolic information systems. However, there are few related studies on mixed information systems, which promotes the construction of the related incremental attribute reduction algorithm under the mixed information system. The discernibility degree is an important method used for designing attribute reduction. In this paper, the traditional discernibility degree is generalized under the mixed information system, and the concept of neighborhood discernibility degree is presented. Then,the incremental learning of neighborhood discernibility degree is studied respectively when objects increase or objects decrease under the mixed information system. Finally, according to this incremental learning, the corresponding incremental attribute reduction algorithms are proposed, respectively. The related experimental results on the UCI data set show that the proposed incremental attribute reduction can update the reduction results more quickly than the non incremental attribute reduction.关键词：rough set;mixed data;discernibility degree;neighborhood relation;incremental learning;attribute reduction62|88|5更新时间：2025-07-08
- 摘要：In energy-limited communication systems, energy efficiency is a key factor that affects the performance of the systems. In this paper, we consider a wireless powered hybrid non-orthogonal multiple access network consisting of a base station and multiple clustered users. In this network, the base station transfers energy to the users in a wireless manner, and the users utilize the harvested energy to transmit information to the base station. To simplify the complexity of the base station’s information receiver, a hybrid multiple access scheme is adopted by the users, which are partitioned into multiple clusters. The users in the same cluster transmit in the non-orthogonal multiple access manner, while the users from different clusters transmit in the time division multiple access manner. To maximize the energy efficiency of the network, we jointly optimize the time length of wireless energy transfer from the base station to the users and the time length of information transmission from the users to the base station, as well as the transmit powers of the users. The formulated problem is non-convex and thus is difficult to solve. To tackle this problem, we first find the structure of its optimal solution, and then propose an efficient iterative algorithm to solve it based on the fractional programming technique. Simulation results show that the energy efficiency performance of the proposed algorithm significantly outperforms that of two benchmark schemes, namely the throughput maximization scheme and the fixed time allocation scheme.关键词：non-orthogonal multiple access;energy-efficiency;fractional programming;resource allocation69|453|1更新时间：2025-07-08
- 摘要：In consideration of the sensor control for multi-target tracking, this paper proposes the corresponding sensor control strategy using Gaussian mixture multi-Bernoulli filter based on the FInite Set Statistics (FISST) theory. First, this paper gives the implementation of the Cubature Kalman Gaussian Mixture Cardinality Balanced Multi-target Multi-Bernoulli Filter (CK-GMCBMeMBerF), and extracts the Gaussian mixture component to approximate multi-Bernoulli density. In addition, we study the solution of the Cauchy-Schwarz divergence between the two Gaussian mixture distributions, and derive the information gain corresponding to the change of multi-target probability density. Then, the corresponding sensor control strategy is proposed. Moreover, a detailed Gaussian Mixture (GM) implementation of the posterior expected number of targets (PENT) criteria is given based on CK-GMCBMeMBerF, and the corresponding sensor control strategy is studied with GM-PENT as the evaluation criteria. Finally, simulation results verify the effectiveness of these proposed algorithms.关键词：multi-target tracking;sensor control;finite set statistics;Gaussian mixture;information gain62|254|4更新时间：2025-07-08
- 摘要：In view of the shortcomings of the existing method of identification of radiant sources, a novel feature extraction method based on I/Q mismatch is proposed, which can be detected, reproducible and invariable by using the I/Q mismatch fingerprint of the source. This method is a fingerprint feature extraction method based on signal space. It needs to estimate the signal-to-noise ratio of the signal. Through comprehensive consideration, the SNR estimation method based on eigenvalue decomposition is adopted. The experimental simulation results show that, under the same signal to noise ratio,the radio frequency fingerprint feature extraction method based on I/Q mismatch is superior to the dual spectrum and Hilbert-Huang transform signal feature extraction methods.关键词：emitter recognition;I/Q imbalance;fingerprint characteristics;SNR estimation148|181|5更新时间：2025-07-08
- 摘要：In order to improve the data reconstruction accuracy and alleviate the influence of packet loss over unreliable links on the Compressive Sensing (CS) data gathering in sensor networks, we propose a Packet Loss Matching Data Gathering Algorithm Based on Compressive Sensing (CS-MDGA) in this paper. This proposed algorithm establishes the correlation effect of the network data with the CS technique. We further design the Sparse Observation Matrix based on Packet Loss Matching (SPLM) in this paper. In addition, we prove that the designed observation matrix satisfies the Restricted Isometry Property (RIP) with a probability arbitrarily close to 1, which can guarantee the reliable delivery of the multi-path routing data among different nodes. The simulation results show that the relative reconstruction error of this proposed algorithm is still lower than 5% even when the packet loss rate of the link is as high as 60%. Therefore, it is verified that this proposed algorithm not only exhibits high reconstruction accuracy, but also effectively alleviates the influence of packet losses over unreliable links on the CS-based data collection.关键词：sensor networks;compressive sensing;data gathering;correlation effect;sparse observation matrix88|131|3更新时间：2025-07-08
- 摘要：Aiming at the weak signal detection problem under strong noise background, the weak signal detection principle based on Duffing oscillator is analyzed. Combining the complementary ensemble empirical mode decomposition (CEEMD) method with the variable-scale Duffing oscillator, a new weak signal detection method is proposed. The complex noisy signal is decomposed into different intrinsic mode functions (IMF) by using CEEMD. Through the Duffing system bifurcation diagram and its changes, the critical threshold of the phase trajectory change is found, and the information detection of the noisy signal is realized. The results show that the joint detection method can not only immune noise well, but also effectively detect multi-frequency periodic signals with signal-to-noise ratio as low as -73dB.关键词：weak signal detection;chaos;bifurcation diagram;complementary ensemble empirical mode decomposition53|78|3更新时间：2025-07-08
- 摘要：Linear phases are often a requisite in many signal filtering applications. All-pass digital filters are the main device to realize the linear phases. An iterative reweighted minimax method is proposed for the design of all-pass digital filters. The method minimizes the maximum weighted phase deviation from some linear phase, and reduces the maximum group-delay deviation from some constant group delay through updating the phase-error weight function iteratively by virtue of the group-delay deviation. To demonstrate its advantages, the proposed method is applied in the designs of phase equilizers and linear-phase Hilbert transformers. Simulation results show that the proposed algorithm has good convergence properties, and is very effective in realizing the nearly linear phase and nearly constant group delay of the filtering system.关键词：all-pass filter;phase equalizer;Hilbert transformer;linear phase60|85|0更新时间：2025-07-08
- 摘要：Due to the massive parameters and complex structure, deep learning networks are usually trained in a long time with large-scale training samples. In this paper, we propose a spatial-spectral convolutional dense network (SSCDenseNet) which mainly targets limited samples for hyperspectral image classification. Three novel strategies are proposed to construct the proposed network. First, a spatial-spectral separable convolution method is adopted to make up a hidden layer unit with a spectral one-dimensional convolutional layer and a spatial two-dimensional convolutional layer; then the deep network is constructed by stacking multiple units. Second, we use batch normalization before each hidden layer unit to reduce covariance drift of data and accelerate the network training procedure. Finally, a direct connection between every two units is adopted to reuse hierarchical features, and solve the problem of gradient vanishing. The comprehensive evaluation of experiments on different datasets such as Indian Pines, Pavia University and Salinas are conducted to show the performance of the SSCDenseNet, and the results show that the proposed method outperforms several state-of-the-art deep learning based methods in terms of classification performance, especially under small-scale samples.关键词：hyperspectral image;supervised classification;deep learning;dense network;spectral-spatial convolution149|132|11更新时间：2025-07-08
- 摘要：In this paper, we design and propose SemtoSql+, a semantic deep network query model based on demand aggregation. At the same time, it is a network to address the complex and cross-domain Text-to-SQL generation task. Based on LSTM and Word2Vec embedding technology, the corpus is trained as the input word vector of the model. Combined with the dependency graph method, the problem of SQL statement generation transforms into slot filling. SemtoSql+divides complex tasks into four levels and constructs by the need of aggregation, using the attention mechanism to effectively avoid the order problem in the traditional model and using a random masked mechanism to enhance the model.关键词：natural semantic processing;complex events;semantic web;deep learning83|93|2更新时间：2025-07-08
- 摘要：Domain Models shape the Domain Layer of Web Applications.Anemic Domain Models (ADMs) are Domain Models holding only data. States of ADMs are maintained by classes in other layers, causing the latter bloodshot. However, there lacks research revealing the significance of impact that anemia and bloodshot of layers pose on maintainability. To quantify the significance, this paper assesses intensity of 3 Code Smells (Feature Envy, Blob and Data Class) as evaluation standards. Through an experiment conducted on 91 Java projects and multiple releases of 10 Java Web applications, this paper concludes that over 75% of the projects are affected. As the impact persists, correlations of the intensities exist among different classes of a project as well as same classes in different releases of a project.关键词：software maintainability;web application;layered architecture;code smell;domain modeling125|148|1更新时间：2025-07-08
- 摘要：Aiming at the high energy-efficiency implementation of cryptographic algorithm, this paper proposed a coarse-grained reconfigurable cryptographic logic array structure named PVHArray. Based on the research of cryptographic algorithm operation and control structure features, adopted the reconfigurable array structure design method, this paper proposed the coarse-grained reconfigurable cryptographic logic array structure and its parametric model, which is mainly composed of pipeline variable coarse-grained reconfigurable computing units, hierarchical interconnected network and periodic-oriented distributed control network. In order to improve the energy-efficiency of the reconfigurable cryptographic logic array, this paper combined the cryptographic algorithm mapping results to determine the model parameters, and constructed a high energy-efficiency PVHArray structure with a size of 4×4. The chip area of PVHArray is 12.25mm2 based on 55nm CMOS technology, and at the same time, cryptographic algorithm mapping is performed for PVHArray. The experimental results show that the proposed high-efficiency PVHArray structure can effectively support the mapping of block, stream and hash cipher algorithm. In the cipher block chaining (CBC) mode, compared with state-of-the-art reconfigurable cryptographic logic array REMUS_LPP, the performance per unit area has increased by 12.9% and the performance per unit power has increased by 13.9%.关键词：pipeline variable;hierarchical;periodic-oriented;high-efficiency;array47|93|1更新时间：2025-07-08
- 摘要：Image attribute annotation is a refined method of image annotation.It can narrow the "semantic gap" between cognition and features. However, a single feature is used to characterize images and the deep-level semantics are not fully explored. So annotations cannot depict images comprehensively. The traditional effective range based gene selection algorithm is modified to complete feature fusion. And transfer learning strategy is designed to complete material annotation. The cross-modal semantics among features are mined by the discriminant correlation analysis algorithm. So the relative attribute model is optimized to complete deep-level semantics (practical attributes) annotation. Experimental results demonstrate: Material attributes annotation accuracy reaches 63.11%, which is improved by 1.97% compared with baseline. Practical attributes annotation accuracy reaches 59.15%, which is improved by 2.85% compared with baseline. The proposed hierarchical annotation mechanism can more comprehensively depict images.关键词：image annotation;effective range based gene selection;relative attribute;transfer learning;cross-modal semantics;discriminant correlation analysis69|81|2更新时间：2025-07-08
- 摘要：This paper proposes a context modeling framework based on multi-stream architecture and LSTM, which aims to overcome two difficulties for group behavior recognition. One is to fuse information from multiple visual cues in complex scenes, the other is to model situational characters to get the long-term temporal context in the video. In addition, decision fusion is performed on the behavior recognition results based on global information and local information to determine the final group behavior attributes. The algorithm achieved 93.2% and 95.7% average recognition rates on CAD1 and CAD2 respectively.关键词：group behavior recognition;fusion of multiple visual cues;interactive context modeling;global-local model;long short-term memory network136|85|3更新时间：2025-07-08
- 摘要：Graphs are increasingly used in data management, knowledge discovery and information services. As an important strategy of graph analysis and applications, graph embedding has become one of the subjects with great attention in artificial intelligence. Starting from the challenges faced in graph embedding studies, this paper introduces the principal methods based on matrix decomposition, random walk and deep learning. Then, we introduce general test datasets, evaluation criteria as well as typical applications widely used in graph embedding. Finally, we summarize the trend and future research issues of graph embedding.关键词：graph model;graph embedding method;graph embedding application;test datasets;evaluation metrics106|142|17更新时间：2025-07-08
- 摘要：Modeling of image complexity is a research that builds a computer to sense the visual complexity of images in a way that is similar to the majority of people. The research is an interdisciplinary subject and is of great significance in the field of image engineering. In this article, the main methods of evaluating image complexity have been carried out by a comprehensive analysis. The paper reviews the different application fields of image complexity. Then the paper introduces the evaluation methods for visual complexity in detail from information theory, theory of image compression, image analysis and image modeling method of complexity. Especially, we focus on descriptions of the image features used in the complexity evaluation method based on the image features. In terms of image complexity assessment, we illustrate the classification and regression methods for evaluating complexity of images. This work also proposes the existing problems and challenges in the study of image complexity.关键词：visual complexity;affective perception;complexity evaluations;feature extraction;classification and regression76|100|3更新时间：2025-07-08
- 摘要：By making the pseudocodewords more costly,the penalized decoding method based on alternating direction method of multipliers (ADMM) can improve the decoding performance for low-density parity-check (LDPC) codes at low signal-to-noise ratios and also has low decoding complexity.Reducing the number of Euclidean projection in ADMM penalized decoding,selecting the appropriate message scheduling strategy and designing effective penalty function are three important methods to increase the ADMM penalized decoding speed.In order to increase the ADMM penalized decoding speed further,by using the method proposed by Wei et al to reduce the number of Euclidean projections,this paper designs two kinds of ADMM penalized decoding methods with the horizontal layered scheduling and the vertical layered scheduling strategy for LDPC codes based on the I-l1-PF penalty function.Simulation results show that the designed methods not only have better decoding performance but also significantly reduce the average number of iterations and the average decoding time compared with the existing ADMM penalized decoding methods.关键词：low-density parity-check (LDPC) codes;alternating direction method of multipliers (ADMM);penalty function;penalized decoding;layered scheduling61|113|1更新时间：2025-07-08
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备09064830号-19
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
0