Search

Semantics-aware anomaly detection based on log has attracted much attention. However, the existing methods based on the weighted aggregation of all word vectors might lose the semantic relationship of word order and cannot maintain the unique representation, and the methods based on word order-preserving by concatenating all word vectors might lead to a high computation time cost. To solve these issues and further improve the sequential anomaly detection, this paper proposes a network anomaly detection scheme LogNADS by designing a novel log semantics representation method and an adaptive sequence data construction method. It first discards the useless words and then selects theme words to hold the log abstraction and maintain a low time cost as well. Subsequently, it concatenates theme words’ vectors based on the original word order to maintain the unique representation and avoid the word order loss. Furthermore, to better detect the sequential anomalies, we utilize the sliding window scheme and design a method to compute the optimal window size for constructing the log sequence self-adaptively, and then LSTM is built to extract timing characteristics of the log sequences. Experimental results conducted on the public benchmark HDFS dataset and BGL dataset demonstrate the effectiveness of LogNADS through comparing with other state-of-the-art methods in the detection accuracy and time cost. Moreover, the statistical significance tests prove the superior performance.

Encryption and information hiding are widely used in security applications requiring copyright protection, over-communication, and tamper detection. Quantum steganography, using quantum images and audio as information carriers, is a fascinating hybrid of classical and quantum informatics. While existing steganography techniques offer high invisibility, current embedding capacities remain insufficient. As such, a high capacity quantum steganography algorithm, based on image interpolation, is proposed in this study. In this approach, quantum interpolation is conducted by maximizing difference values between neighboring pixels to generate a cover image. Secret message embedding and extraction steps are then developed for the resulting quantum images, in addition to corresponding quantum circuit networks. A series of simulation experiments were conducted to demonstrate the implementation and assess the performance of the proposed technique, the efficiency of which was evaluated using several performance metrics and the visual quality of the stego-images. The resulting information capacity was much higher than that of existing quantum steganography algorithms. This suggests the proposed strategy to be a promising trade-off between invisibility and capacity, providing new motivation to investigate image steganography using quantum computing resources.

The X-ray radiation from computed tomography (CT) brought us the potential risk. Simply decreasing the dose makes the CT images noisy and diagnostic performance compromised. Here, we develop a novel denoising low-dose CT image method. Our framework is based on an improved generative adversarial network coupling with the hybrid loss function, including the adversarial loss, perceptual loss, sharpness loss, and structural similarity loss. Among the loss function terms, perceptual loss and structural similarity loss are made use of to preserve textural details, and sharpness loss can make reconstruction images clear. The adversarial loss can sharp the boundary regions. The results of experiments show the proposed method can effectively remove noise and artifacts better than the state-of-the-art methods in the aspects of the visual effect, the quantitative measurements, and the texture details.

The real-time rendering of large-scale terrain scenes has been a challenge and is a popular research topic in the fields of 3D games and virtual reality. To fully utilize GPU features, this paper proposes a real-time rendering algorithm for large-scale regular grid terrain based on GPU Tessellation. The CPU processes a coarse-grained quadtree to generate terrain patches and then passes the patches to the GPU after view culling. Then, the GPU refines the terrain by using GPU Tessellation and displacement mapping. The most important advantage of our method is that during the tessellation stage, a new rule that the distance, screen space projection error and variance of the height, as the standard of terrain roughness, is designed. During the tessellation stage, the distance, screen space projection error and variance of the height, as the standard of terrain roughness, are considered. The experimental results show that our algorithm can greatly reduce the CPU processing time and main memory space requirement, and fully utilize the new features of the GPU; In conclusion, the proposed approach can render large-scale terrain with high-precision in real-time.

The WeChat Official Account (WCOA) is the most influential self-media platform in China, and the follower economy brought by the extension of self-media has profoundly affected its survival and development. Therefore, it is of vital importance to understand the followers' behavioural patterns and the law of quantity change of followers. To study this problem, we collected the operation data of two WCOAs. First, an evolutionary Followers-Susceptible-View-Forward-Removed (F-SVFR) model is proposed to describe the trends in the behavioural state of the followers of WCOA after accepting the pushed message. Second, we proposed an equation to simulate the number of newly increased followers and find that the number of newly increased followers of WCOA followed a similar pattern that grew rapidly in the early stages and kept a relatively low and steady rate of increase rate later on. Finally, the popularity formula is defined for the WCOA to serve as a predictor of online popularity, which demonstrated that popularity is positively correlated with the number of followers. Our study provides account operators with practical guidance, which has significance for understanding the popularity of increased official accounts and the law of quantity growth of followers.

Traffic flow prediction is one of the fundamental components in Intelligent Transportation Systems (ITS). Many traffic flow prediction models have been developed, but with limitation of noise sensitivity, which will result in poor generalization. Fused Lasso, also known as total variation denoising, penalizes L1-norm on the model coefficients and pairwise differences between neighboring coefficients, has been widely used to analyze highly correlated features with a natural order, as is the case with traffic flow. It denoises data by encouraging both sparsity of coefficients and their differences, and estimates the coefficients of highly correlated variables to be equal to each other. However, for traffic data, the same coefficients will lead to overexpression of features, and losing the trend of time series of traffic flow. In this work, we propose a Fused Ridge multi-task learning (FR-MTL) model for multi-road traffic flow prediction. It introduces Fused Ridge for traffic data denoising, imposes penalty on L2-norm of the coefficients and their differences. The penalty of L2-norm proportionally shrinks coefficients, and generates smooth coefficient vectors with non-sparsity. It has both capability of trend preservation and denoising. In addition, we jointly consider multi-task learning (MTL) for training shared spatiotemporal information among traffic roads. The experiments on real traffic data show the advantages of the proposed model over other four regularized baseline models, and on traffic data with Gaussian noise and missing data, the FR-MTL model demonstrates potential and promising capability with satisfying accuracy and effectiveness.

Three-dimensional (3D) registration plays a pivotal step in augmented reality (AR) systems. Traditional 3D registration methods have the disadvantages of poor accuracy and robustness. This paper proposes a novel registration method, we believe, for AR systems based on the AKAZE and Tanimoto similarity measurement method. In this paper, the image feature points are extracted and matched by combining the AKAZE algorithm and the Tanimoto similarity measurement method. Then, the camera pose is estimated by calculating the constraint relationship of the feature points. Finally, the 3D registration and real-time tracking of the virtual objects are realized by the Lucas–Kanade (LK) optical flow tracking algorithm. We use Tanimoto to determine the similarity of feature points to improve the matching accuracy of the AKAZE algorithm, and this method not only retains the advantages of strong scale adaptation but also has the advantage of high-precision matching. Experiments show that the method proposed in this paper has the benefits of high registration accuracy, low time consumption, and strong robustness. Under the premise of ensuring accuracy, when the marker is rotated or blocked, it can be accurately registered. In addition, when the external environment changes, for example, the light intensity or the size of the parallax, the registration can still be stable.

In order to solve the problem that the shape change of flexible object is very complex and the projection on the flexible object seriously affects the viewing effect of the observer, a method of correcting the geometric distortion of flexible object based on texture mapping was proposed. By building the rules on the surface of the projection area of the largest calculating model, solved the camera and projector, projection object coordinate system to solve the problem of the conversion relationship between the three, implements the projection images aligned with the flexible projection shape of the object. The experimental results show that this method can reduce the limitation of projection device and simplify the operation flow of projection correction. To achieve a better projection mapping effect on the deformation of flexible objects.

For spinal curvature measurements, because of the anatomical complexity of the spine CT image, developing an automated method to avoid manual landmark is a challenging task. In this study, we propose an intelligent framework that integrates the cascade AdaBoost classifier and region-based distance regularized level set evolution (DRLSE) with the vertebral centroid measurement. First, the histogram-of-oriented-gradients based cascade gentle AdaBoost classifier is used to detect automatically and localize vertebral bodies from computer tomography (CT) spinal images. Considering these vertebral pathological images enables us to produce a diverse training dataset. Then, the DRLSE method introduces the local region information to converge the vertebral boundary quickly. The located bounding box is regarded as an accurate initial contour. This avoids the negative impact of manual initialization. Finally, we perform vertebral centroid extraction and spinal curve fitting. The spinal curvature angle is determined by calculating the angle between two tangents to the curve. We verified the effectiveness of the proposed method on 10 spine CT volumes. Quantitative comparison against the ground-truth centroids yielded a detection accuracy rate of 98.3% and a mean centroid location error of 1.15 mm. The comparative results with existing methods demonstrate that the proposed method can accurately detect and segment vertebral bodies. Furthermore, the spinal curvature can be automatically measured without manual landmark.

We present the modified-Bessel-superposed Gaussian (MBSG) orbital-angular-momentum (OAM) modes, which can evolve into a field featuring a tailored Gaussian amplitude profile after propagating to a target receiving plane. Analytical expressions for the MBSG OAM modes at a source plane are developed in both the space and wavevector domains. The space- and wavevector-domain representations of the MBSG OAM modes at the source plane are found to share a mathematically formal similarity. We propose to utilize this appealing property for designing a Fourier-transform-based approach to forming the MBSG OAM modes at the source plane. Unlike other commonly reported OAM modes, the receiving-plane field of MBSG OAM modes always exhibits a customized non-ring-shaped Gaussian intensity pattern. This behavior can be exploited, e.g., to benefit the projective phase-flattening OAM measurement and to reduce the turbulence-induced OAM-mode scattering in relevant practical applications.

Mutually complementary bases are crucial to secure quantum key distribution (QKD). By aiming at free-space QKD with use of spatially structured photons, the effect of turbulence on the mutual unbiasedness of two complementary bases relevant to photonic orbital-angular-momentum (OAM) modes is modeled theoretically, with novel insightful expressions and physical explanations yielded. For two complementary bases constructed from Laguerre-Gaussian (LG) modes with a fixed radial index of zero, it is shown that the superposed joint-two-LG-mode (JTLGM) correlation function plays a fundamental role in determining the turbulence-induced mutual-unbiasedness decay. The mutual information between the sent and detected photonic states is used as a metric to quantify the degree of turbulence-induced mutual-unbiasedness decay. It is found that the degree of turbulence-induced mutual-unbiasedness decay depends on the scaled atmospheric coherence width, anisotropy of turbulence and contents of the complementary bases, and it can become non-negligible in certain cases.

In the cloud infrastructure, the co-resident attack is a critical security threat. Through virtualization technology provided by Cloud Service Provider, tenants’ virtual machines (VMs) are possible to be allocated on the same host. Multi-tenant environment provides malicious tenants an opportunity to launch the co-resident attack and steal other tenants’ information by side channels. To prevent this type of attack, previous works mostly pay attention to eliminating side channels and few of them study VM deployment strategy. Hence, we focus on deploying VMs with a secure and effective allocation strategy to reduce the probability of VM co-residence. A novel VM allocation strategy is proposed with three optimization objectives including security, load balancing and energy consumption. Finally, we implement our VM allocation strategy and prove its effectiveness on the simulation platform CloudSim.

Background Imbalanced data classification is an inevitable problem in medical intelligent diagnosis. Most of real-world biomedical datasets are usually along with limited samples and high-dimensional feature. This seriously affects the classification performance of the model and causes erroneous guidance for the diagnosis of diseases. Exploring an effective classification method for imbalanced and limited biomedical dataset is a challenging task. Methods In this paper, we propose a novel multilayer extreme learning machine (ELM) classification model combined with dynamic generative adversarial net (GAN) to tackle limited and imbalanced biomedical data. Firstly, principal component analysis is utilized to remove irrelevant and redundant features. Meanwhile, more meaningful pathological features are extracted. After that, dynamic GAN is designed to generate the realistic-looking minority class samples, thereby balancing the class distribution and avoiding overfitting effectively. Finally, a self-adaptive multilayer ELM is proposed to classify the balanced dataset. The analytic expression for the numbers of hidden layer and node is determined by quantitatively establishing the relationship between the change of imbalance ratio and the hyper-parameters of the model. Reducing interactive parameters adjustment makes the classification model more robust. Results To evaluate the classification performance of the proposed method, numerical experiments are conducted on four real-world biomedical datasets. The proposed method can generate authentic minority class samples and self-adaptively select the optimal parameters of learning model. By comparing with W-ELM, SMOTE-ELM, and H-ELM methods, the quantitative experimental results demonstrate that our method can achieve better classification performance and higher computational efficiency in terms of ROC, AUC, G-mean, and F-measure metrics. Conclusions Our study provides an effective solution for imbalanced biomedical data classification under the condition of limited samples and high-dimensional feature. The proposed method could offer a theoretical basis for computer-aided diagnosis. It has the potential to be applied in biomedical clinical practice.

By analyzing the numerical representation of amplitude values in audio signals and integrating the time component, a representation for audio signals on quantum computers, FRQA, is proposed. The FRQA representation is a normalized state that facilitates basic audio signal operations targeting these parameters. The preparation and retrieval for FRQA are discussed and, based on the FRQA state, we realize the circuits to accomplish basic audio signal operations such as signal addition, signal inversion, signal delay, and signal reversal. These operations can be employed as the major components to build advanced operations for particular applications in the quantum computing domain., 9 pages, 13 figures, 1 table

Computer-Aided Diagnosis (CAD) benefits from its early diagnosis and accurate treatment of lung diseases. Accurate segmentation of lung fields is an important component in CAD for lung health, which facilitates subsequent analysis. However, most of the existing algorithms for lung fields segmentation are unable to ensure appearance and spatial consistency due to the varied boundaries and poor contrasts. In this paper, we propose a novel and hybrid method for lung fields segmentation by integrating Dense-U-Net network and a fully connected conditional random field (CRF). In order to realize the reuse of image features, the structure of densely-connected is added to the decoder, which ensures the object with varied shapes and sizes can be extracted without adding more parameters. To make full use of the mutual information among pixels of the original image, a fully connected CRF algorithm is adopted to further optimize the preliminary segmentation results according to the intensity and position of each pixel. Compared with some previous popular methods on JSRT dataset, the proposed method in this paper shows higher Jaccard index and Dice-Coefficient.

Computer-Aided Diagnosis (CAD) benefits to early diagnosis and accurate treatment of lung diseases. As a preprocessing of CAD-based chest radiograph analysis, reliable lung segmentation is a prerequisite step which affects the precision of lesion recognition and classification. The techniques of deep learning have been widely applied for learning task-adaptive features in image segmentation. However, most existing lung fields segmentation methods based on deep learning are unable to ensure appearance and spatial consistency of the lung fields due to the varied boundaries and poor contrasts. In this study, we propose a novel method for lung fields segmentation by integrating U-Net network and a fully connected conditional random field (CRF). In the first step, we train the U-Net network designed in this paper to provide a preliminary probability to each pixel in images. Secondly, a fully connected CRF algorithm is used in this paper to optimize the coarse segmentation according to the intensity and position information of each pixel in images. Comparison with some previous methods on JSRT dataset, the proposed method in this paper shows higher Dice-Coefficient and Jaccard index.

Internet addiction refers to excessive Internet use in daily life. Its negative impact on college students calls for timely discovery of its reasons and correct guidance. However, present research methods are mainly based on the questionnaire, which can be affected by non-randomly selected samples and a low response rate. Thanks to the development of the smart campus, students' behavior can be recorded as data, thus whether Internet addiction is correlated with daily behavior can be analyzed quantitatively. In this paper, we extracted five features and proposed an Internet Addiction Rating Scale (IARS) to quantify students' Internet addiction level based on the Internet login data, and found that Internet addiction is positively correlated with consumption amount, negatively correlated with consumption speed and academic performance, but has little relationship with self-discipline. Our work throws some light on the relationship between students' Internet addiction and daily behavior through unobtrusive data analysis, helping college management staff detect abnormal learning status timely and reasonably.

Optical spatial-mode reception has a physical nature quite different from that of the traditional optical power-in-the-bucket (PIB) reception. The former belongs to coherent reception scheme while the latter pertains to incoherent reception scheme. Under weak-turbulence conditions, the statistical correlation between turbulence-impacted optical signals collected by a pair of adjacent spatial-mode receivers is mathematically formulated in terms of a new theoretical framework that takes into account the distinctive nature of the spatial-mode reception. The aperture Fresnel number, coherence Fresnel number, separation Fresnel number and mode Fresnel number are identified as fundamental determinative parameters in evaluation of the correlation coefficient. With the help of the obtained formulations, two analytical asymptotic formulae for the correlation coefficient are further derived under the conditions that the aperture Fresnel number is much smaller than the coherence Fresnel number and separation Fresnel number, respectively. Despite the use of asymptotic approximations in the theoretical derivation, it is found that the two asymptotic formulae indeed have utility in many situations of practical interest to us. Furthermore, Monte-Carlo-simulation-based calculations are carried out to examine the accuracy of employing the two asymptotic formulae to approximate the correlation coefficient. It is shown that the ranges of applicability of these two asymptotic formulae are mainly determined by the coherence Fresnel number and the ratio of the separation Fresnel number to the aperture Fresnel number, basically irrespective of the mode Fresnel number.

Multimodal medical images are useful for observing tissue structure clearly in clinical practice. To integrate multimodal information, multimodal registration is significant. The entropy-based registration applies a structure descriptor set to replace the original multimodal image and compute similarity to express the correlation of images. The accuracy and converging rate of the registration depend on this set. We propose a new method, logarithmic fuzzy entropy function, to compute the descriptor set. It is obvious that the proposed method can increase the upper bound value from log(r) to log(r) + ∆(r) so that a more representative structural descriptor set is formed. The experiment results show that our method has faster converging rate and wider quantified range in multimodal medical images registration.

The multi-model image registration is widely used in medical clinical diagnosis. However, the feature descriptor pairs are hard to detect between different modalities, which is a common dilemma faced in multi-modal image registration. In this paper, we detect feature descriptor pairs in the structure representation by convolutional neural network (CNN). On the one hand, structure representation can transform different modalities into a third-type modality, and it shows the potential information which do not appear at multi-modal medical image commonly. On the other hand, the matched feature points pairs which are computed by CNN can move back from structure representation to original image. Finally, the effective feature points are matched by 2NN algorithm. Extensive experiments on group-wise registration prove that this algorithm overcome the dilemma in extracting feature descriptor pairs among the multi-modal medical images.

The identification of students with financial difficulties is one of the main problems in campus data research. Effective and timely identification not only provides convenience to campus administrators but also helps students who are really in financial hardship. The popular using of smart cards makes it possible to identify students with financial difficulties through big data. In this paper, we collect behavioural records from undergraduate students’ smart cards and propose five features by which to associate with students’ poverty level. Based on these features, we proposed the Apriori Balanced Algorithm (ABA) to mine the relationship of poverty level with students’ daily behaviour. Association rules show that students’ poverty level is most closely related to their academic performance, followed by consumption level, diligence level, and life regularity. Finally, we adopted the semisupervised K-means algorithm to more accurately find out students with financial difficulties. Tested by classical classification algorithms, our method has a higher identification rate, which is helpful for university administrators discover students in real financial hardship effectively.

The topological structure of the spatial information network changes drastically, the resources on the space are limited, and the communication delay is long which bring great challenges to the construction of the spatial information network. In the analysis of the existing satellite network routing model, based on the DTN protocol, a low-orbit satellite network routing model on account of the price game is proposed based on the characteristics of satellite network. The model uses the price game to save the remaining storage space. The success of data forwarding is an important evidence for the routing node status as a satellite network for routing, which enhances the practicability of the routing algorithm. Simulation experiments show that compared with classical models such as Epidemic and FC, the proposed routing model has better performance in terms of delay, success rate, network overhead, etc., and the comprehensive performance is better.

Deep learning has been widely used for identifying anomaly network traffic. It trains supervised classifiers on a pre-screened numerical traffic feature dataset in the most cases, so the classification effectiveness depends heavily on feature representation. There is no unified feature representation method, and the current feature representation methods cannot profile traffic precisely. Therefore, how to design a traffic feature representation method to profile traffic is challenging. We propose a Network Anomaly Detection Scheme based on data Representation (NADSR). Data representation method converts raw network traffic into images by treating every numerical feature value as an image pixel and then creating a circulant pixel matrix for a traffic sample. It retains the traffic feature’s spatial structure instead of padding empty pixels with constant values while directly reshaping a long feature vector into a pixel matrix. Experimental results verify the effectiveness of the proposed NADSR. It improves the overall detection accuracy compared with state-of-the-art methods, and also provides reference to solve security-related classification problems.

In the past few decades, anti-aliasing method has been widely investigated in the field of computer graphics. Compared with the traditional anti-aliasing method, the anti-aliasing method with temporal coherence can avoid the computational cost of rendering additional samples. However, most of the existing temporal coherence anti-aliasing methods have limitations in performing anti-aliasing of specular materials. In this paper, we propose an optimized anti-aliasing method based on virtual hit point, which can avoid sampling error of specular surface. We use ray-tracing pipeline to render the image and display the results. Experimental results show that our method can achieve anti-aliasing effect with real-time performance.

To improve the performance of cloud-based rendering systems, it not only requires a good load balancing algorithm, but also needs less cost for data transfer. This paper proposes an internal data compression method in the cloud, which reduces the cost of data transfer between worker nodes on the basis of Hadoop distributed framework. In this method, the coordinate system transformation and simple vector addition are used to reduce the storage space of each pixel's primary ray-object intersection point and unit normal vector. We implemented the experiment of data transfer in a Hadoop ray tracing cluster with gigabit switch. First, we compared the cost under different types of data transfer. Secondly, we measured the execution time of the compression method. Finally, we analyze the reason for the error between the decompressed data and the origin data. The results show that the cost of data transfer in our method is approximately 66% of that of the three-dimensional coordinate representation method, and the calculation error is less than 10-6. Thus, our proposed method can effectively improve the transfer efficiency of cloud-based rendering system.

The instantaneous transmission coefficient, i.e., instantaneous transmittance, of a turbulent optical orbital-angular-momentum (OAM) channel is mathematically formulated as a weighted integration and is found to range between 0 and 1. Common probability distribution models for optical irradiance fluctuations with a support from 0 to ∞ are not strictly proper for statistical description of the fluctuating transmission coefficient. The novel dual Johnson S B distribution is proposed to model the statistical behavior of the fluctuating transmission coefficient. Its applicability is verified by making comparisons between the histograms of transmission-coefficient samples generated by Monte Carlo simulations and the corresponding fitted probability density functions; the values for its four independent control parameters under different conditions are obtained by the fit of the dual Johnson S B distribution to relevant simulated transmission-coefficient samples. It is found that each of the four independent control parameters of the dual Johnson S B distribution can be considered as a function of three quantities, viz., the OAM index, the Fried’s atmospheric coherence width, and the ratio of the root-mean-square (RMS) OAM-beam radius to the Fried’s atmospheric coherence width. The results demonstrate that the statistical distribution of the fluctuating transmission coefficient depends less on the first two quantities than on the last one. Finding a model for direct mapping from these three quantities to the four control parameters of the dual Johnson S B distribution deserves future study.

In clinical diagnosis, automatic fracture detection can reduce misdiagnosis which caused by fatigue and inexperience of radiologists, and provide support for reducing patient suffering and preventing disease progression. This paper proposes a SK-DenseNet model to detect femur fracture based on DenseNet network. It uses SK module to adjust the size of receptive fields adaptively, adopts Focal loss function to update parameter, and uses Grad-CAM method to visualize the femoral detection results. It can improve the interpretability of the fracture detection model. This paper used the femoral dataset to verify the performance of the mode. Our model achieves an accuracy of 0.9117, with the kappa coefficient is 0.8228. The experimental results show that the performance of the proposed model is better than the traditional deep learning model.

Multi-step traffic flow prediction extends short-term single-step prediction to long-term prediction, which is more significant in many basic application in intelligent transportation systems, such as traffic planning. A main problem of multi-step prediction is that the error accumulation as steps increase, resulting in prediction performance degradation. In this work, combining recursive and multi-output strategies, we proposed a deep learning model, named MARNN, for multi-step traffic flow prediction. Specifically, we jointly consider recurrent neural network as dynamic neural network for simulating the dynamic characteristics in traffic time series as recursive strategy does and multi-output strategy for decreasing the accumulated error as step increases. In addition, we introduce attention mechanism for adaptively seeking correlated important information among traffic time series to improve prediction performance. The experiments on real traffic data show the advantages of MARNN model over other four baseline models, demonstrating the potential and promising capability of the proposed model on multi-step traffic flow prediction.

Traffic flow prediction has been regarded as one of the key problems in Intelligent Transportation Systems. Neural networks are widely leveraged in traffic prediction task, but with the limitation of (1) single task learning, which would ignore shared information among traffic network; and (2) initialization problem that would directly affect the training efficiency. In this work, we propose a deep neural network based multitask learning approach for traffic flow prediction, called MSAE, which incorporates stacked autoencoders (SAE) for the neural network initialization, and jointly and adaptively predicts network-scale traffic flow with shared information. The experiments on real traffic data indicate that the proposed model, jointly considering shared information among different traffic flow prediction tasks and neural network initialization, is capable and promising of dealing with complex traffic flow forecasting with satisfying accuracy and effectiveness.

Lumbar vertebral fracture seriously endangers the health of people, which has a higher mortality. Due to the tiny difference among various fracture features in CT images, multiple vertebral fractures classification has a great challenge for computer-aided diagnosis system. To solve this problem, this paper proposes a multiclass PSVM ensemble method with multi-feature selection to recognize lumbar vertebral fractures from spine CT images. In the proposed method, firstly, the active contour model is utilized to segment lumbar vertebral bodies. It is helpful for the subsequent feature extraction. Secondly, different image features are extracted, including 3 geometric shape features, 3 texture features, and 5 height ratios. The importance of these features is analyzed and ranked by using infinite feature selection method, thus selecting different feature subsets. Finally, three multiclass probability SVMs with binary tree structure are trained on three datasets. The weighted voting strategy is used for the final decision fusion. To validate the effectiveness of the proposed method, probability SVM, K-nearest neighbor, and decision tree as base classifiers are compared with or without feature selection. Experimental results on 25 spine CT volumes demonstrate that the advantage of the proposed method compared to other classifiers, both in terms of the classification accuracy and Cohen’s kappa coefficient.

In the low earth orbit (LEO) satellite networks, a single satellite may cover areas with various population quantity, economic conditions and time zones. In this case, networks have time-varying traffic load and may face with unbalanced load distribution. Moreover, software defined networking (SDN) has dynamic monitoring function that can be implemented in the satellite network to observe the real-time status of satellite links and nodes, and quickly update routing tables. However, high link utilization is likely to cause local network congestion, resulting in increased delay. To solve this problem, we propose a load balancing algorithm under the delay constraint. The experimental results show that our proposed scheme can reasonably allocate link bandwidth under the delay requirement and achieve lower end-to-end delay and higher system throughput.

Load balancing determines the performance of the sort-first rendering clusters. A dynamic load balancing algorithm based on per-pixel rendering cost is proposed. Firstly, a 3D scene is rendered by ray tracing. At the same time, the number of intersection points between rays and scene objects in a pixel is obtained. Then, a rendering-cost map is created according to these numbers in a frame. Based on the render history, the rendering-cost map in a previous frame is used to estimate the real distribution of the ray tracing rendering cost in a future frame. Finally, a load assignment strategy is proposed to make rendering cost more uniform among rendering nodes. The experimental results show that the proposed algorithm can render continuously the next frame by subdividing future frames in advance. It is better than the compared algorithm in terms of load balancing and frame rate.

Liver tumour is a general term for tumour lesions occurring in the liver, which is the third leading cause of tumour death in the world. Most liver tumours are small in size, variable in shape and unfixed in position, and the existing neural network is not fully utilized in the segmentation of liver tumours. In order to make up for the problem of traditional tumour segmentation, improve the precision and effect of tumour treatment. The multi-scale Attention Module U-NET (MAM U-NET) was proposed by analyzing CT image data of liver tumours. The MAM U-NET model consists of a newly designed Combination module and an MAM attention module. Each MAM attention module is composed of two parts: spatial attention model and channel attention model, which enable the model to learn more extensive and rich context-dependent information and spatial information. Finally, a comparison experiment was conducted in the LITS data set to verify the performance of the model. Liver segmentation achieved the optimal result, and the average Dice coefficient reached 96.8282%. The segmentation results of liver tumours were second only to nnU-Net, and the average Dice coefficient was 60.5991%, which proved that the MAM U-NET model had better performance in the segmentation of liver tumours.

Virtual binocular stereoscopic camera models that can be used in generation of stereoscopic images or videos of virtual 3D scenes by use of computer graphics techniques are presented and analyzed in detail. It is found that the paralleloptical-axis model is most appropriate for synthesizing stereoscopic images or videos of virtual 3D scenes. Mathematical formulae for calculating the camera position and ‘lookat’ position of both the left and right virtual cameras therein are developed. Light-source visibility filtering with a depth variation constraint is proposed to speed up the ray-tracing-based soft shadow rendering of virtual 3D scenes and meanwhile to eliminate the drawback of generating fake shadows associated with the original light-source visibility filtering method; dual light-source visibility filtering scheme is suggested to make shadows reflected in a mirror surface correctly feature penumbra. The stereoscopic video experimentally generated by use of our method shows pleasing visual realism and impressive stereoscopic effects.

A framework to synchronise audio and visual contents to realise talking quantum movies is proposed in this study. This framework segments a complete movie into three components, i.e. frame, audio, and time. A multi-channel representation for quantum images (MCQI) is used to encode the still image that make up the movie frames. In addition, the amplitude content of the flexible representation of quantum audio (FRQA) signal is recorded at each instant of time. The synchronisation of audio and frame in the movie is accomplished through a quantum sequence of time. The feasibility of the proposed framework is demonstrated by a simple simulation.

Common randomness arising from turbulence-induced signal fading in reciprocal optical wireless channels is a beneficial resource that can be used to generate secret keys shared by two legitimate parties. The concept of optical wireless channels using common-transverse-spatial-mode coupling (CTSMC) that can maintain perfect fading reciprocity in atmospheric turbulence is first developed in a general manner. Subsequently, by performing Monte Carlo simulations, the Johnson SB probability distribution is demonstrated to be appropriate for statistical description of turbulence-induced signal fading in an optical wireless channel constructed by use of two identical CTSMC transceivers, and the nature of correlation between signal fadings detected by two contiguous reception spatial modes is further quantitatively characterized, revealing that rapid spatial decorrelation between signal fadings observed by a legitimate party and an eavesdropper holds for scenarios of practical interest. Finally, the information theoretic capacity for generating secret keys from CTSMC-based optical wireless channels is theoretically formulated and quantitatively examined under different conditions, manifesting that the turbulence strength and average electrical signal-to-noise ratio have a noticeable combined impact on the secret key capacity, especially in the far-field case.

In medical intelligent diagnosis, most of the real-world datasets have the class-imbalance problem and some strong correlation features. In this paper, a novel classification model with hierarchical feature representation is proposed to tackle small and imbalanced biomedicine datasets. The main idea of the proposed method is to integrate extreme learning machine-autoencoder (ELM-AE) into the weighted ELM (W-ELM) model. ELM-AE with norm optimization is utilized to extract more effective information from raw data, thereby forming a hierarchical and compact feature representation. Afterwards, random projections of learned feature results view as inputs of the W-ELM. An adaptive weighting scheme is designed to reduce the misclassified rate of the minority class by assigning a larger weight to minority samples. The classification performance of the proposed method is evaluated on two biomedical datasets from the UCI repository. The experimental results show that the proposed method cannot only effectively solve the class-imbalanced problem with small biomedical datasets, but also obtain a higher and more stable performance than other state-of-the-art classification methods.

Characteristics of LEO satellite networks, like dynamically changed topological structures, limited on-board resources, and longer communication delay, have brought new challenges to the construction of satellite networks. By analyzing existing routing models for satellite networks, this paper proposes an auction-gaming-based routing model for LEO satellite networks, based on the DTN protocol and against such characteristics. By making use of an auction model, it takes space propagation loss, residual storage space of a node, and routing hop counts as important bases for routing selection. Analysis shows that besides the routing function, this model also plays an active role in avoiding “selfish” satellite nodes, as well as in relieving network congestion.

A method to detect a moving target in multi-channel quantum video is proposed based on multiple measurements on the video strip. The proposed method is capable of detecting the location of the moving target in each frame of the quantum video thereby ensuring that the motion trail of the object is easily and efficiently retrieved. Three experiments, i.e. moving target detection (MTD) of a pixel, MTD of an object in complex shape, and MTD of a pixel whose color is conterminous with that of its background, are implemented to demonstrate the feasibility of the proposal. This study presents a modest attempt to focus on the moving target detection and its applications in quantum video.

A method for video encryption and decryption on quantum computers is proposed based on color information transformations on each frame encoding the content of the encoding the content of the video. The proposed method provides a flexible operation to encrypt quantum video by means of the quantum measurement in order to enhance the security of the video. To validate the proposed approach, a tetris tile-matching puzzle game video is utilized in the experimental simulations. The results obtained suggest that the proposed method enhances the security and speed of quantum video encryption and decryption, both properties required for secure transmission and sharing of video content in quantum communication.

Sensory information is multimodal; through audiovisual interaction, task-irrelevant auditory stimuli tend to speed response times and increase visual perception accuracy. However, mechanisms underlying these performance enhancements have remained unclear. We hypothesize that task-irrelevant auditory stimuli might provide reliable temporal and spatial cues for visual target discrimination and behavioral response enhancement. Using signal detection theory, the present study investigated the effects of spatiotemporal relationships on auditory facilitation of visual target discrimination. Three experiments were conducted where an auditory stimulus maintained reliable temporal and/or spatial relationships with visual target stimuli. Results showed that perception sensitivity ( d') to visual target stimuli was enhanced only when a task-irrelevant auditory stimulus maintained reliable spatiotemporal relationships with a visual target stimulus. When only reliable spatial or temporal information was contained, perception sensitivity was not enhanced. These results suggest that reliable spatiotemporal relationships between visual and auditory signals are required for audiovisual integration during a visual discrimination task, most likely due to a spread of attention. These results also indicate that auditory facilitation of visual target discrimination follows from late-stage cognitive processes rather than early stage sensory processes.

Two distinct methods based on which two different quadratic-form expressions for the two-source spherical wave structure function (WSF) can be derived are reviewed. The validity of closed-form expressions for the beam-wave cross-spectral density function (CSDF) due to generalized atmospheric turbulence featuring an arbitrary spectral index ranging from 3 to 4, developed based on the quadratic two-source spherical WSFs, is examined in detail. New formulations for the conditions under which the said closed-form expressions for the beam-wave CSDF are strictly valid are developed and several novel interesting findings are elucidated. In particular, the closed-form beam-wave CSDF derived based on the small-separation asymptotic two-source spherical WSF can be considered a rigorous asymptotic solution under the strong-turbulence condition only when the separation distance between the two observation points is much smaller than the inner scale of turbulence; moreover it is also a rigorous asymptotic solution when a certain relation among the initial beam radius, initial transverse coherence width and inner scale holds, regardless of the turbulence strength and spectral index. On the other hand, the accuracy of the closed-form beam-wave CSDF derived based on the large-separation-approximation two-source spherical WSF depends on the spectral index, and a spectral index closer to 4 results in better accuracy.

Both symmetric and asymmetric color image encryption have advantages and disadvantages. In order to combine their advantages and try to overcome their disadvantages, chaos synchronization is used to avoid the key transmission for the proposed semi-symmetric image encryption scheme. Our scheme is a hybrid chaotic encryption algorithm, and it consists of a scrambling stage and a diffusion stage. The control law and the update rule of function projective synchronization between the 3-cell quantum cellular neural networks (QCNN) response system and the 6th-order cellular neural network (CNN) drive system are formulated. Since the function projective synchronization is used to synchronize the response system and drive system, Alice and Bob got the key by two different chaotic systems independently and avoid the key transmission by some extra security links, which prevents security key leakage during the transmission. Both numerical simulations and security analyses such as information entropy analysis, differential attack are conducted to verify the feasibility, security, and efficiency of the proposed scheme.

Expressions for the correlation coefficient between light-flux fluctuations of two waves counter-propagating along a common path in weak turbulence are developed. Only the aperture and inner-scale Fresnel parameters are needed for evaluation of the correlation coefficient if the turbulence spectrum has no path dependence, and of the path weighting functions for the cross-covariance and variances of normalized light-flux fluctuations if the turbulence spectrum is dependent on path locations. Under the condition that atmospheric turbulence is statistically homogeneous over a path, although good correlation between light-flux fluctuations of two counter-propagating spherical waves may be achieved for a relatively small aperture Fresnel parameter or relatively large inner-scale Fresnel parameter, the correlation coefficient between light-flux fluctuations of two counter-propagating plane waves is always lower than 1 obviously. When the aperture Fresnel parameter becomes larger than the inner-scale Fresnel parameter, the inner scale of turbulence tends to play an unimportant role in determining the correlation coefficient.

Shadow mapping is commonly used in real-time rendering. In this paper, we presented an accurate and efficient method of soft shadows generation from planar area lights. First this method generated a depth map from light’s view, and analyzed the depth-discontinuities areas as well as shadow boundaries. Then these areas were described as binary values in the texture map called binary light-visibility map, and a parallel convolution filtering algorithm based on GPU was enforced to smooth out the boundaries with a box filter. Experiments show that our algorithm is an effective shadow map based method that produces perceptually accurate soft shadows in real time with more details of shadow boundaries compared with the previous works.