Search

In some computed tomography (CT) applications, limited-angle projections are used for image reconstruction, and traditional reconstruction methods, such as filtered back-projection (FBP) or simultaneous algebraic reconstruction technique (SART), cannot reconstruct high-quality CT images without prior knowledge assistance. For limited-angle CT reconstruction, total variation minimization (TVM) method is not conductive to recovering image structures. Image reconstruction methods based on anisotropic total variation (ATV) and reweighted anisotropic total variation (RwATV) can significantly reduce the shading artifacts using prior knowledge of the scanning angular range and image sparsity. However, using the knowledge of image sparsity does not further improve image quality in some applications. In this paper, we propose a new reconstruction method based on anisotropic relative total variation (ARTV) for limited-angle CT reconstruction. In ARTV, the windowed inherent variation (WIV) indicates the strength of structure information and WIV values are adaptively determined by local structure information. In limited-angle CT, particular scanning angular range urges us into exerting different strengths on different directions of ARTV. Experiments on FORBILD HEAD phantom, a thoracic image and real projections of a walnut are performed to test the new method. Experimental results show that the destroyed structures are recovered to some extent, and we acquire higher image quality compared to some existing limited-angle CT reconstruction methods.

Recent studies have shown that both the quantity and quality of the driving data are crucial for the success of deep learning based driver identification solutions. This is not only true for the training of deep neural networks, but also for their deployment. In the deployment phase the models are applied to identify new drivers that are different from the drivers in the training phase. New data is required to re-train the models and adjust them to fit the new drivers. While the focus of recent studies has been mostly on the training phase, this study focuses on the deployment phase. Inspired by face and fingerprint recognition solutions, in this paper we propose a solution that does not require model retraining in the deployment phase. Instead, only a small amount of example driving data from each driver is required. More specifically, we propose an approach that utilises the triplet loss function to learn a mapping from vehicular sensor data to an embedding space, where the distance between two embeddings represents the drivers' similarity. The resulting network can be directly used for identifying new drivers without re-training.

In this paper, a fault diagnosis method of cascaded NPC inverter based on single sensor is proposed. By detecting the DC side current, the specific harmonic content is extracted by Fourier decomposition as the fault feature, and the BP neural network is used for fault diagnosis. The method can install only one current sensor to identify and locate all the faults of all modules. It is important to reduce system hardware resources. Finally, the validity of the diagnostic method was verified through MATLAB/Simulink simulation and experiment.

Deep learning has shown unique advantages in multiple applications. However, medical image data such as 3D brain MRI scans contain private information of subjects and cannot be shared and used without special permits. This makes it difficult for aggregated analysis of medical images from multiple sites. Federated learning provides an approach to aggregate learned features from different sites without transferring raw data, ensuring the security of subject information. Traditional federated learning algorithms such as FEDAVG has been proved to be effective, but the performance is always poor due to high data heterogeneity across sites. Moreover, federated learning for 3D medical images has rarely been explored. Here we proposed a Guide-Weighted Federated Deep Learning framework for multisite 3D brain MRI images. We used weighted aggregation on gradients and then used the guide gradients to update federated models to deal with cross-site heterogeneity. The framework was validated on the publicly accessible multi-site brain MRI database of Autism Spectrum Disorder (ASD) - ABIDE I&II. Results showed that federated models outperform local models with improvements in accuracy of 0.92-4.02%. As the first federated learning framework for 3D MRI data, the proposed framework is expected to facilitate imaging data sharing across sites and accelerate the Medical Imaging Alliance's formation. In this way, a robust federated model with better performance can then be generated for future diagnostic classification tasks and assisted in the diagnosing and screening of neuropsychiatric disorders.

In order to meet the needs of modern society for agricultural crop environmental monitoring, a multidimensional and multi angle monitoring system for crop environment is designed. By connecting the corresponding data acquisition sensors, the system collects and processes the data, and sends it to the network module. Finally, it sends it to the information processing center for precise analysis and processing of the data, and displays it visually. Finally, through the verification and testing of the system function, it fully proves that the system can effectively improve the supervision efficiency of crop environment, has high application value, and achieves the expected effect.

In this article, a tunable linear-to-circular polarization converter (LTCPC) in terahertz (THz) regime using the graphene transmissive metasurface is proposed, which is composed of two resonant layers containing the metal and graphene resonators separated by a dielectric spacer. The linearly-polarized wave with normal incidence can be transformed to the circularly-polarized wave. The operating band can be dynamically regulated in THz band by electrically controlling the Fermi energy ( $E_{\mathrm {f}}$ ) of the graphene sheets rather than reforming the structures. The optimized result of axial ratio (AR) band which is less than 3 dB is located at 2.64-3.29 THz (the relative bandwidth is 21.92%) in the case of $E_{\mathrm {f}}=0.1$ eV. The physical characteristics of graphene are explored and the relevant operational results of the presented LTCPC are elucidated in this article. Compared with the conventional LTCPC, our design provides a more effective implementation method for wide applications, and it offers a further step in graphene controllable devices.

With the advancement of multimedia technologies and driven by the ever-increasing user interests in the variety of multimedia applications, the content-rich multimedia streaming services tend to be the most attractive service in future Internet. The multipath TCP (MPTCP), which uses multiple paths for parallel transmission and bandwidth aggregation, is considered to be the most potential transfer mechanism to satisfy the specific requirements of multimedia transmission in a multi-homed wireless network environment. However, the fully-reliable transmission nature of MPTCP can cause an unnecessary retransmission of expired multimedia data, block message from handing over to upper layer, and thus degrade the user-perceived multimedia service quality. In this paper, we propose a receiver-centric partial-reliable multipath transport solution (referred to as $rec$ PR-MPTCP) to support real-time Internet multimedia applications. In the $rec$ PR-MPTCP, a partial reliability extension is presented which runs at receiver side in order to offer a partial reliability multipathing service for multimedia applications. In addition, a new one-way delay-based bandwidth aggregation method is included into the $rec$ PR-MPTCP in order to reduce the receive buffer blocking problem in MPTCP and increase multimedia services performance. Simulation results show that $rec$ PR-MPTCP outperforms the current MPTCP solutions in terms of throughput performance and user-perceived quality of service.

Computed tomography (CT) has its irreplaceable function in nondestructive testing and medical diagnosis. In some practical CT imaging applications, the limited-angle scanning is common due to X-ray’s potential harm to human and the limitation of the scanning conditions. Under these circumstances, analytic reconstruction algorithms, like filtered backprojection (FBP), will not obtain satisfactory results because of lacking the projection data. Iterative reconstruction (IR) methods that can incorporate prior knowledge have attracted attention in many fields, and wavelet frame-based regularization reconstruction algorithms have proven to be a useful means to reduce slope artifacts and noise for limited-angle CT. However, with the obtained projection data of the scanned object further reduces, the edge structures and the details of the reconstructed image worsen. For the sake of improving the quality of the reconstructed image from the limited-angle projection data, a guided image filtering (GIF)-based limited-angle CT reconstruction algorithm using wavelet frame was proposed. In each iteration of the proposed algorithm, the reconstructed result constrained by the wavelet frame was used as the guidance image to transfer the important features it contains to the reconstructed result of SART method by GIF. Furthermore, some simulated experiments and real data tests were conducted to evaluate the feasibility and validity of the proposed algorithm, and the qualitative and quantitative indexes indicated that the proposed algorithm was superior to other iterative reconstruction algorithms in artifacts reduction, noise suppression, and structure preservation.

How to efficiently and accurately locate people and objects in the indoor environment has always been a hotspot in indoor localization research. Channel State Information (CSI) fingerprint is a very effective method for indoor localization. At present most CSI information collection tasks are done by humans, which is labor-intensive and inefficient. To overcome the disadvantages of manual operation, researchers begin to use UAV to complete CSI information collection tasks in recent years. At present, the UAV localization fingerprint information collection algorithm has the following problems: (1) When planning the UAV path, the energy consumption of turning, ascending, descent, etc. are not taken into consideration, which leads to large energy consumption for completing tasks; (2) Data collection only at the two-dimensional level instead of three-dimension results in errors in data analysis. To address these problems, we model the three-dimensional indoor environment, the action, and energy consumption of the UAV, then convert the path and action planning problem into a Generalized Traveling Salesman Problems (GTSP). An energy-efficient UAV path and action planning algorithm (EEPAP) is proposed to collect three-dimension CSI information. Finally, through experiments, we evaluated the performance of EEPAP and proved that it outperforms the existing algorithm in terms of energy consumption.

China’s economic transition from quantitative growth to qualitative growth, improving internal management efficiency and managerial expertise is becoming the top priority for power transmission companies. Standard cost system is an important tool for companies to deliver on that goal. This paper looks into the development and application of the activity-based standard cost system for grid companies’ operation and maintenance(O&M), analyzes and assesses the resource consumed and costs incurred in the O&M process using a standard cost system based on actual procedures and practices. The effectiveness analysis shows that the establishment of activity-based O&M standard cost system can improve the budget management for the company and lays a systematical groundwork for further application facilitating corporate finance management.

This paper studies the whole process of teaching in higher vocational colleges. The whole process of teaching and its management are the core parts of school teaching and management. We have studied the teaching process theories at home and abroad and defined the teaching process in higher vocational colleges. The research group analyzed the input and output influencing variables in the whole process of higher vocational teaching, and noticed that the operation and management of higher vocational teaching is to manage the variation of the whole process of teaching. According to the 80/20 rule, 20% of the important influencing variables in the whole process of higher vocational teaching are well controlled, so as to add value to teaching activities and teaching processes, and make students’ learning and teaching results be integrated with knowledge and practice.

With the development of education informati-zation in China, the student profiles are becoming digitized. However, currently different digital profiles of a student are still stored dispersedly in multiple databases, which results in information silos and vulnerability to security and privacy threats. Aiming at these deficiencies, this paper proposes a secure storage and sharing scheme of student digital profile based on blockchain technology. With the characteristics of decentralization, security, credibility, and tamper-resistance of the blockchain, the consensus of network nodes is achieved through the Delegated Byzantine Fault Tolerance (DBFT) protocol, and the sharing of student digital profile information is achieved through the access control mechanism. The public information of educational digital information is preserved on the blockchain, and the student profiles are encrypted and stored in a database or cloud, which facilitates the effective storage of personal digital profiles and information sharing among various systems. The proposed scheme utilizes blockchain technology for decentralized and tamper-proof management, and provides a solution to information silos and security issues of student digital profiles.

The housing provident fund system is the state's incentive policy to solve the housing problems of low-income families, and more and more companies are beginning to pay provident funds. In the past, there is little research on the evaluation method of the company provident fund payment status. To effectively evaluate the willingness and ability of the company to pay the provident fund, this paper designs an evaluation method of the company's payment status based on historical data of the provident fund. This method can be used as a basis for future provident fund evaluations. First, construct a feature set from multi-source data, and use Spearman correlation coefficient to extract the optimal feature subset. Then, according to the two features that have the greatest impact on the provident fund, the provident fund payment rating is calculated. Finally, we select random samples as the training set, and train in models such as LightGBM Random Forest XGBoost DeepGBM. The results show that LightGBM can be better applied to the evaluation of historical provident fund data.

Recent studies on machine learning based driver identification have shown that leveraging deep neural networks to learn latent features from vehicle sensor data boosts the performance of the models to high levels of accuracies. However, models produced by deep neural networks are difficult to explain and the interpretability of their results is limited. Consequently, the reliability of the learned models heavily depends on the amount, quality and diversity of the training dataset as well as on the validation scenarios. In this work, we evaluate state-of-the-art deep learning networks for driver identification using a very rich dataset of more than 395,000 kilometres of different driving scenarios and environmental conditions (e.g. route, vehicle, traffic, weather) collected over two years. It turns out that the neural networks achieve high accuracy levels when training and testing on the same type of driving conditions. However, accuracy drops and importance of individual signals varies when testing on different driving conditions, although all best practices like stratification and cross validation have been applied. Our findings suggest that relying solely on the vehicle sensor data without taking the contextual information about driving conditions into account is not a practical solution.

This paper aims to solve the difficulties of Alzheimer’s disease (AD) diagnosis by constructing a classification framework based on a one-dimensional convolutional neural network (CNN), which provides finer clinical cues and additional performance gain for AD diagnosis. For the selected MRI image data, it is divided into three types according to the degree of disease for further discussion, Mild Cognitive Impairment (MCI), and Normal Control (NC) Classification of the population. The framework first uses deep learning network to perform feature extraction on all pre-processed MRI slices, and then classifies them through 1-D CNN to achieve automatic classification of AD. A large amount of experimental data shows that the classification accuracy of the 1-D CNN framework for AD vs. MCI vs. NC, AD vs. MCI, AD vs. NC, MCI vs. NC has reached the average accuracy of 0.9550, 0.9933, 0.9945 and 0.9597, respectively, which show the proposed framework is fast and simple. It effectively improves the classification results over all the categories and achieves a superior classification performance for AD diagnosis.

Passive (handheld) THz radiometer scanner provides a new method of security inspection, which can not only protect the privacy of scanned person but also prevent potential damage to human body from high radiation, comparing with traditional instruments. This paper investigates the theoretical bases of object detection under human body background using THz Radiometer, and presents the corresponding object detecting algorithm based on DSP methods.

Nowadays, the mainstream video coding standard H.265/HEVC has achieved excellent compression performance, but it still can not match the increase of huge amount of data. In order to further optimize intra prediction of HEVC, we proposed an iterative filtering algorithm which combines the characteristics of HEVC intra prediction modes. Proposed algorithm classifies PU into different types by analyzing the characteristics of different intra prediction modes. Through discrete training methods, the coefficients of each type of filters can be obtained. And multiple iterative filtering can make the prediction more accuracy. We further optimize the complexity of the algorithm to make a balance between performance and complexity. Experimental results show that proposed algorithm can achieve better coding performance compared with HEVC baseline algorithm and some latest algorithm.

An asymmetric coplanar waveguide-fed (CPW) ultra-wideband (UWB) antenna based on L-strip technology with dual-band rejection characteristics is presented in this paper. The proposed antenna consists of a CPW-fed slot strip rectangular patch with L-strip open stub in the front side and two inverted L-shaped strips conductor with L-strip slot on the backed ground plane. The size of the proposed antenna is 25x25x1.6 mm 3, and the simulated and measured results show that the proposed antenna has good impedance matching and radiation patterns, and it operates in 2.93 GHz -12.94 GHz with dual-band rejection performance in the frequency bands of 7.61 GHz - 8.58 GHz and 9.13 GHz-10.03 GHz for satellite and X- band communication.

In this paper, we introduce a plasma electron density measurement sensor based on a Low Temperature Co-fired Ceramics (LTCC) antenna, which has three layers. The variation range of plasma frequency spans from 0.6 to 6GHz, and the resonant frequency of the corresponding LTCC antenna varies from 4 to 5GHz. The measurement accuracy reaches 0.1 GHz, and the measurement accuracy is over 90%. With the increase of plasma density, the resonance frequency of the presented LTCC antenna moves to higher frequencies gradually. The measurement principle is that when the effective permittivity of plasma is different (different densities), the resonant frequencies of antennas is also different. The presented plasma electron density measurement sensor is analyzed theoretically, and numerical analysis are given.

We present a novel type of tunable gravity tailored metamaterial absorber which can obtain absorption in GHz region and can become a frequency selective surface by rotating it 180°. The original sandwich structure was modified in this structure. By digging grooves in dielectric substrate which owns surface metal resonance units, EGaIn is placed in the grooves for replacement of the metal plane. Due to the fluidity of liquid metal (eutectic gallium-indium), EGaIn would flow into different part of this unique designed structure when it was rotated, this instrument can achieve a 99.17% absorption at 6.87 GHz. This instrument will change into a FSS which can manage a stopband from 7.95 to 9.5 GHz. Importantly, compared with the most of current tailoring approaches of the absobers, the proposed method is more convenient.