Search

Pansharpening is a very debated spatio-spectral fusion problem. It refers to the fusion of a high spatial resolution panchromatic image with a lower spatial but higher spectral resolution multispectral image in order to obtain an image with high resolution in both the domains. In this article, we propose a novel variational optimization-based (VO) approach to address this issue incorporating the outcome of a deep convolutional neural network (DCNN). This solution can take advantages of both the paradigms. On one hand, higher performance can be expected introducing machine learning (ML) methods based on the training by examples philosophy into VO approaches. On other hand, the combination of VO techniques with DCNNs can aid the generalization ability of these latter. In particular, we formulate a $\ell _2$ -based proximal deep injection term to evaluate the distance between the DCNN outcome, and the desired high spatial resolution multispectral image. This represents the regularization term for our VO model. Furthermore, a new data fitting term measuring the spatial fidelity is proposed. Finally, the proposed convex VO problem is efficiently solved by exploiting the framework of the alternating direction method of multipliers (ADMM), thus guaranteeing the convergence of the algorithm. Extensive experiments both on simulated, and real datasets demonstrate that the proposed approach can outperform state-of-the-art spatio-spectral fusion methods, even showing a significant generalization ability. Please find the project page at https://liangjiandeng.github.io/Projects_Res/DMPIF_2020jstars.html .

Recently, many researches propose to employ attention based encoder-decoder models to convert a sequence of trajectory points into a LaTeX string for online handwritten mathematical expression recognition (OHMER), and the recognition performance of these models critically relies on the accuracy of the attention. In this paper, unlike previous methods which basically employ a soft attention model, we propose to employ a posterior attention model, which modifies the attention probabilities after observing the output probabilities generated by the soft attention model. In order to further improve the posterior attention mechanism, we propose a stroke average pooling layer to aggregate point-level features obtained from the encoder into stroke-level features. We argue that posterior attention is better to be implemented on stroke-level features than point-level features as the output probabilities generated by stroke is more convincing than generated by point, and we prove that through experimental analysis. Validated on the CROHME competition task, we demonstrate that stroke based posterior attention achieves expression recognition rates of 54.26% on CROHME 2014 and 51.75% on CROHME 2016. According to attention visualization analysis, we empirically demonstrate that the posterior attention mechanism can achieve better alignment accuracy than the soft attention mechanism.

As the technology continues scaling down and the data rate is increasing dramatically, PCB design has become more challenging than ever. Keeping track of signal and power integrity becomes a requirement for PCB designers in order to meet the product development time and reduce the overall cost. In high speed IO signals, vias play a critical role for the signal integrity quality. It is important to design the via impedance within a certain range. Due to server products have a higher thickness of PCB design than other consumer products, and its data rates are on the cutting edge, the need for the via optimization is acquired. In this work, the investigate on the via impedance control is studied to optimize the via design on the pitch, individual and oval antipad shapes. And the further analyses of via design parameters are examined to understand the effects of dominance.

A 2 × 2 dual-polarized patch antenna array operating at 60 GHz is proposed. Based on the conventional low-cost standard PCB process, for the reliable antenna design a five- laminates stack-up structure is used to reduce the complexity of the manufacturing and cost. To improve the antenna bandwidth and gain performance, a stacked patch is adopted and two probe-fed structures are realized to support both horizontal and vertical polarizations. A 2 × 2 prototype is designed, fabricated, and measured for demonstration. For both polarizations, measurements show that their -10 dB impedance bandwidth can cover the entire 60 GHz band, and their isolation is lower than -22 dB. Moreover, their gain is higher than 10 dBi with a peak gain of about 11 dBi.

We proposed a cascade-based control for the consensus tracking problem of multiple underactuated surface vessels (USVs) with uncertainties and the environment disturbance under sampled-data interaction. The proposed control contains two parts. The frist part is a second-order distributed estimator which is designed to obtain the estimate position of the leader. Thus the leader-follower problem become as an estimate trajectory tracking problem. The second part is the local control law of the USV which is designed based on the estimate position by using sliding mode and backstepping methods to achieve the consensus tracking. In terms of Lyapunov and graph theory, several sufficient conditions are given to ensure the consensus tracking of the multi-USV system. Besides, the tracking error can be arbitrary small if we choose proper parameters in the control input and the distributed estimator.

Using the magnetic-control sputtering fabrication, WO 3 /Ag/WO 3 thin film was successfully formed as a sandwich structure treated as a transparent conductive thin film. The thicknesses of silver and tungsten oxide were 10 to 20nm and 50 to 60nm, respectively. For the increase of silver thickness from 10 to 20nm, the sheet resistance is reduced. However, this effect in tungsten oxide is entirely reversed. As the electrical measurement was applied to this tested sandwich structure, the conductivity is better than that for the single layer of tungsten oxide due to the conduction contribution from the silver layer.

Recently, an effective segmentation-free approach via deep neural network based hidden Markov model (DNN-HMM) was proposed and successfully applied to offline handwritten Chinese text recognition. In this study, to further improve the modeling capability, we adopt deep convolutional neural networks (DCNN) to calculate the HMM state posteriors. First, on the frame basis, the DCNN-HMM can automatically learn the features from the raw image of the handwritten text line via the convolutional architecture rather than the handcrafted gradient features using in the DNN-HMM. Second, we examine several important factors of DCNN to the recognition performance, namely the kernel size, the number of blocks and convolutional layers. We also improve the language modeling by using more text data and high-order N-gram. Tested on ICDAR 2013 competition task of CASIA-HWDB database, the proposed DCNN-HMM could achieve a character error rate (CER) of 4.07\%, yielding a relative CER reduction of 30.8\% over the DNN-HMM approach. To the best of our knowledge, this is the best published result of the segmentation-free approaches. Furthermore, we explain why DCNN-HMM is more effective than DNN-HMM via the visualization of feature learning and the error pattern analysis.

Due to the recent breakthrough of bioimaging, automated classification of protein subcellular localization based on immunohistochemistry (IHC) images has become an important topic of proteomics research. Inspired by the impressive performance of deep learning in various image classifications, we trained a deep neural network model to classify protein images of eight subcellular localizations, which is able to achieve higher classification accuracies than using traditional models of support vector machine. Intermediate outputs of the neural network were visualized to show that our model can capture subtle texture features from IHC images and lead to better subcellular location classification results. In addition, our results show that data rebalance can significantly improve the classification performance in this multi-class deep classifier

A parameterized design of speed servo controller for brushless DC motors is presented based on a robust composite nonlinear control framework. The servo controller consists of a linear control part for achieving fast response, a nonlinear feedback part for suppressing the overshoot, and a disturbance compensation mechanism for erasing the steady-state error. A reduced-order extended state observer is adopted to estimate the un-measurable state variables and unknown disturbance. Simulations are first carried out in MATLAB/Simulink, followed by real-time experiment using a TMS320F28335 digital signal controller board. The results confirm that the proposed control scheme can achieve fast, smooth and accurate tracking of target speed, and has a good performance robustness against the load disturbance and parameter variations.

This paper proposes a novel segmentation-free approach using deep neural network based hidden Markov model (DNN-HMM) for offline handwritten Chinese text recognition. In the general Bayesian framework, three key issues are comprehensively investigated, namely feature extraction, character modeling, and language modeling. First, as for the feature extraction on the basis of each frame or sliding window, the gradient-based features are extracted for the DNN-based classifier. Second, the text line is sequentially modeled by HMMs with each representing one character class. Meanwhile the DNN-based classifier is adopted to calculate the posterior probability of all HMM states. Finally, the character n-gram language model is integrated with the DNN-HMM character model for the Bayesian decision. The experiments on the ICDAR 2013 competition task of CASIA-HWDB database show that the proposed approach can achieve the best published recognition results to our knowledge, yielding a character error rate (CER) of 6.50%, which significantly outperforms the previously best reported oversegmentation approach (with a CER of 9.25%) and the segmentation-free approach using multidimensional long-short term memory recurrent neural network (MDLSTM-RNN) approach (with a CER of 10.6%).

In recent years, the electronic medical records can be easily stored with the rapid development of healthcare technology and cloud computing. However, the patient's medical information security is currently concern issue. Despite of many literatures discussed about it, these literatures still face many security challenges. Therefore, we propose an authentication scheme based on cloud environment. In our scheme, we allow people to use the digital develop mechanism to achieve the sharing medical data. By the way, the biometric fingerprint feature and digital signature are used to ensure the security of medical information in our scheme.

Currently, the power GIS system mainly adopts two-dimensional electronic map to show the power network and related information. But, it's difficult to meet the panoramic display requirement due to lack of vertical viewing angle. This paper introduces a visualization method based on Cesium and 3DCityDB to achieve panorama visualization of power data in 3D virtual environment. The map data of study area is extracted from OpenStreetMap, and used to model city buildings. Power objects and features are designed, and managed by 3DCityDB. Power web 3D GIS communicates with server side by web feature service. Finally, the 3D system is deployed, and its functions are tested. Through this study, we extend the power 3D-GIS to B/S architecture. It is beneficial for the power 3D-GIS to cross platform. The tools used in this paper are open source software and convenient for system upgrade.

This paper proposes an alternative design of expanded proximate time-optimal servomechanism (PTOS) controller for typical servo systems characterized by an integrator cascaded with a damping block. The control law is parameterized by the damping ratio and natural frequency of the closed-loop poles within a speed-dependent linear region. The control scheme aims to accomplish maximum acceleration or braking whenever appropriate and then smoothly switches into a linear control law to achieve a smooth settling. An extended state observer (ESO) is also included to estimate the speed and disturbance simultaneously for the purpose of feedback and compensation. The control scheme is applied to a permanent magnet DC motor servo system for set-point position regulation. MATLAB simulation is carried out, and then experimental verification is conducted using a TMS320F28335 digital signal controller board. The results confirm that the servo system can track a wide range of target references with superior transient performance and steady-state accuracy.

Shunt Active Power Filter (SAPF) with LCL output filter is a typical current tracking control system. This article explores a new control method for fast tracking of compensation current and suppressing its resonance. The state-space model is first established in the synchronous rotating d-q coordinate, which can make the active and reactive power components decoupled, and at the same time the harmonic current can be directly extracted, thus the harmonic detection and compensation control in the SAPF system are unified in the same realization. With the observer-based feedback, the dynamic response and stability of the system can be improved through pole placement in state feedback, while effectively suppressing resonance and reducing the damping loss. Simulation results are given to verify the effectiveness of the system.

This paper presents a novel approach to writer adaptation based on convolutional neural network (CNN) as a feature extractor and improved discriminative linear regression for online handwritten Chinese character recognition. First, the proposed recognizer consisting of CNN-based feature extractor and prototype-based classifier can achieve comparable performance with the state-of-the-art CNN-based classifier while it could be designed more compact and efficient as a practical solution. Second, the writer adaption is performed via a linear transformation of the extracted feature from CNN. The transformation parameters are optimized with a so-called sample separation margin based minimum classification error criterion, which can be further improved by using more synthesized adaptation data and a simple regularization method. The experiments on the data collected from user inputs of Smartphones with a vocabulary of 20,936 characters demonstrate that our writer adaptation approach can yield significant improvements of recognition accuracy over a high-performance baseline system and also outperform a state-of-the-art approach based on style transfer mapping especially with increased adaptation data.

In this paper, a modified version of composite nonlinear feedback (CNF) control method is proposed to achieve fast and accurate trajectory tracking for linear systems subject to input saturation and unknown disturbances. The basic idea is to incorporate a reference generator which constructs the target state vector corresponding with the trajectory reference signal, and an extended state observer which estimates both the system states and the disturbance, into the framework of robust CNF control. The control scheme can retain the fast transient performance of CNF control, and at the same time eliminate the steady-state bias due to disturbances. Simulation studies are conducted to verify that the control method can achieve good transient performance and steady-state accuracy in various trajectory tracking tasks and is more robust against the amplitude variations of disturbance/target reference.

In this paper, a novel one-class classification approach, namely, robust smooth one-class support vector machine (RSOCSVM) is proposed. The proposed method can efficiently enhance the anti-noise ability of the traditional one-class support vector machine (OCSVM). Utilizing the smooth technique, RSOCSVM reformulates the quadratic programming problem of OCSVM as an unstrained optimization format. Moreover, half-quadratic minimization is used to solve the obtained unstrained optimization problem. Experimental results on two synthetic data sets and nine benchmark data sets demonstrate that the proposed method is superior to the traditional OCSVM.

Cellulose nanocrystals (CNCs) fibers have been applied in numerous areas with characteristics of low cost, light weight, nanoscale and strong mechanical strengths. It can be used as strengthening construction materials with a high stiffness which is comparable to steel. Many of the cellulose derivatives are anisotropic which can display birefringence patterns. It is important to detect the orientation and the birefringence for characterizing the structure of CNC fibers. In this paper, we describe the numerical implementation for obtaining fiber orientation map using their scattered light distribution and detecting birefringence by calculating their mean intensity at various rotation angle.

This paper presents a study of designing compact classifiers using deep neural networks for recognition of online handwritten Chinese characters. Two schemes are investigated based on practical considerations. First, deep neural networks are adopted purely as a classifier with a state-of-the-art feature extractor of online handwritten Chinese characters. Second, the so-called bottleneck features extracted from a bottleneck layer of deep neural networks are fed to the prototype-based classifier. The experiments on an in-house developed online Chinese handwriting corpus with a vocabulary of 15,167 characters show that compared with prototype-based classifier widely developed on the mobile device, deep neural network based classifier can yield significant improvements of recognition accuracy with acceptably increased footprint and latency while the bottleneck-feature approach can bring a more compact classifier with an observable performance gain.

Shunt active power filter with LCL output filter involves the tracking of transient current with high frequency harmonic components. The current control brings some complexity. This paper presents a kind of current control algorithm, which combines the triangle carrier comparison method with hysteresis control method. At the same time, the automatic self-correction of the grid-side disturbance while controlling current is discussed. The proposed algorithm is characterized by fast tracking of harmonic current, small computation capacity and easy implementation for engineering applications. Simulation and experiment verify the effectiveness of the proposed method.

In this paper, a robust version of composite nonlinear feedback (CNF) control method is proposed to achieve fast and accurate set-point tracking for linear systems subject to input saturation and disturbances. The basic idea is to treat the unknown disturbance as an extended state variable to be augmented with the plant model, and then design an extended state observer to estimate both the states and the disturbance, and a disturbance compensation mechanism is incorporated into the framework of the original CNF control, so as to alleviate the steady state bias due to disturbances, while retaining the fast transient performance of the CNF control. There is no explicit integration action introduced in this method, thus the headache of integrator windup will not occur. The new control method is then applied to design a two-inertia servo drive system. Simulation studies are conducted to verify that this control method can achieve good transient performance and steady-state accuracy in set-point tracking and is more robust to amplitude variations of disturbance/set-point.

A mode switching control scheme is proposed to achieve fast set-point tracking in motor servo systems. The control scheme incorporates a composite nonlinear feedback (CNF) control law into the framework of proximate time-optimal servomechanisms (PTOS). The CNF control law consists of a linear feedback part for achieving fast response and a nonlinear feedback part for suppressing the overshoot, so as to improve the transient performance in short span positioning, while the PTOS control law is responsible for fast acceleration and deceleration when the tracking error is large. A parameterized MSC controller is designed and the closed-loop stability is analyzed theoretically. The control scheme was then applied to the position-velocity control loop in a permanent magnet synchronous motor servo system. Experimental test has been conducted using the TMS320F2812 DSP. The results verify that the servo system is capable of fast and smooth positioning on a wide range of target locations.

The operation of active power filter requires that the DC-link capacitor voltage is pumped up reliably without current inrush. In this paper, a smooth and reliable, or soft start-up scheme for DC-link capacitor voltage pumping is proposed, by inserting a resistor in series with the inductor of LCL-Filter on the grid side. The resistor serves to limit the electric current, and at the same time improve the control system characteristics. The system model with the resistor in circuit is analyzed, and simulation and experimental results are provided to demonstrate the feasibility of the proposed scheme.

Aiming to the defects in the traditional speed-control of rotary screen rack, such as poor dynamic performance, poor noise immunity, untimely regulation and strong dependence on personnel and experience and so on, this paper proposes a linear quadratic regulator with status feedback, which was used in the constant-speed control of LHG type rotary screen rack. It can be easily and quickly solve Riccati equation by using Matlab software to get the optimal state feedback gain matrix. The simulation result shows that the proposed method is obviously superior to the unit negative feedback control method in dynamic performance, noise immunity and steady precision.

This study develops a real-time traffic information-based emergency medical service (RTIEMS) system by employing sensor devices, webcam, 2.4 GHz ISM band RF module, ZigBee, GPS, Google Maps, and WiMAX mobile network. In the ambulance, patient biosignals consisting of electrocardiogram (ECG), temperature, oxygen, and pulse can be wirelessly transmitted to the in-car gateway. Then, together with patient real-time video, those biosignals are transmitted via WiMAX mobile network to the server located in the hospital emergency room for immediate first-aid preparation. Furthermore, to avoid jam-packed areas, a traffic guiding subsystem is presented based on the WebGIS that consists of Google Maps and GPS to help patient be delivered to the hospital with shortest time. Experimental results verify the effectiveness of the proposed RTIEMS system in lengthening the golden rescue time; thus significantly enhancing service quality of emergency medical system.

In the process of short distance data collection, there have some problems, such as the duplication of work, the complexity of cable connection, the requirement of real time data processing and the electromagnetic interference and so on, so we design the system of data acquisition based on STM32 and virtual instrument. The system uses Cortex-M3 core ARM processor to accomplish the data acquisition of the closed environment, use the Bluetooth serial port module to implement the wireless data transmission, and use the virtual instrument to handle the received data in the host computer. Finally, the test shows that the system with the transmission rate less than 119200bps can ensure the stable and reliable data transmission. The system can be used widely in the wireless data collection, the wireless handheld meter reading and the industrial real time information collection.

This paper describes a novel hierarchy algorithm for extracting bridges over water in optical images. To reduce the omission of bridges by searching the edge, we extract the river regions which the bridges are included in. Firstly, we segment the optical image to get the coarse water bodies using iterative threshold, eliminate the noise regions and add the missing regions based on k-means clustering with texture information and spatial coherence. Then, the blanks are connected based on shape features and candidate bridge regions are segmented from river regions. Finally, the bridges are verified by geometric information and the ubiety between bridges and river. The results show that this approach is efficient and effective for extracting bridges in satellite image from Google Earth and in aerial optical images acquired by unmanned aerial vehicle.

Speckle filtering of polarimetric synthetic aperture radar (POLSAR) is more complicated than a single polarization SAR, because it needs preserve the characteristics of amplitude of each polarization channel and of the relative phase differences among the channels. Fortunately, J.S.Lee et al. have published a scattering-model-based speckle filter (SMBF) in 2006, that filter has good performance for speckle reduction and preserving the characteristics of POLSAR data. However, the Freeman-Durden decomposition, an important procedure of SMBF, is based on the reflection symmetry assumption and not rotation-invariance. In this paper, an improvement of SMBF is proposed for polarimetric SAR data with orientation angle compensation.

Diluted magnetic semiconductors (DMSs) such as transition metal TM-doped ZnO and TiO2 are attracting more and more attention due to their promising applications in spintronic devices. Ferromagnetism (FM) in these materials is often associated with ether free carriers or structural defects. Recently, it has been reported that codoping with hydrogen could be an effective way to tune the concentration of mobile carriers or defects and to enhance the FM in ZnO-based DMSs. However, the mechanisms about the effect of hydrogen treatment induced enhancement of FM in TM-doped ZnO system remains unclear. Some authors claim that ferromagnetic metal clusters can be easily reduced with hydrogen treatment and then contribute to the FM, while other authors suggest that the hydrogen can diffuses into the grain interior and causes more defects or defect complex in TM-doped ZnO which result in a larger saturation magnetization.

The sentence structure similarity measure is an important contents in the study of clause relevance. The accuracy of Chinese Sentence Similarity Measure will directly affect the final results of the relevance of clause. This paper proposes a method for Chinese sentence structure similarity measure. The method compares the parts of speech string corresponding to two small sentences of an Chinese complex sentence, and find the most similar structure and the most effective parts of speech sequence, then calculate the structure similarity of two clauses combining with word shapes. This method can also excludes the effect on computing the similarity of the structure from the differences of the length between two clauses. Finally, the preliminary experiment show that the method is feasible and effective.

Battery power modules (BPMs) with associated power converters are connected in series to comply with a high load voltage. With bidirectional power converters, BPMs can perform buck conversion for the charging operation. The continuous conduction mode (CCM) and the discontinuous conduction mode (DCM) of buck converters in series are analyzed. According to the analyzed results, an equalization strategy is proposed to solve the charge imbalance problem by scheduling the individual battery current. Experiments are carried out to demonstrate the effectiveness of the balance charging method.

This paper presents the operation of battery power modules (BPMs) connected in series for the high voltage output applications. As compared to the conventional applications of the battery banks with a single converter, the configuration has lower component stresses leading to a higher operating efficiency. The batteries can be substantially controlled independently, so that the advantages with parallel operation more efficient utilization of battery power and extensive battery management as well as charge equalization can be realized. Circuit operations and theoretical analyses are discussed on the boost-typed BPMs with series output configuration. Practical measurements are made to demonstrate the effectiveness of charge equalization.

This paper explores a method for establishing diagnostic criterion of TCM (traditional Chinese medicine) syndrome differentiation by studying the set-up of diagnostic standard model of syndrome differentiation of Type 2 diabetes. We present, based on dynamic Kohonen network, the artificial neural net (ANN), fuzzy system (FS) that were used to build up adaptive fuzzy inference system model. The system was tested for its reliability via fisher-iris determination. Then with the help of this model the clinical data was studied, simultaneously, according to TCM basic theories. We obtained the diagnostic criterion of commonly-seen syndrome of Type 2 diabetes, whose rationality was verified by data check-up.

Analysis of elastic laminae provides vital information on both the gross and fine structures of the arterial wall, which is vital to assess issues such as mechanical damage caused by cardiovascular disease or clinical treatment. For example, the distribution of thickness, and distance between them, of the elastic laminae within the aorta can provide important clues on how hypertension affects aortic structure temporally. The requisite microscopic (i.e., histologic or immunohistologic) images have numerous segments, making their measurement and characterization a laborious and inaccurate task. In this paper, we propose a novel scheme for automated analysis of lamina thickness. Results show that our scheme can accurately orient image samples, and detect and calculate thickness distribution in cross sections of the aorta. This new technique not only drastically reduces the human labor, it also eliminates inevitable errors and inconsistencies of manual measurements.

In this paper, a self-calibrated speaker tracking system applied to both image tracking and sound source detection is proposed. The sound source estimated by microphone system is used to supply video tracking system. On the other hand, the direction of speaker detected by image tracking system is fed back to calibrate the phase distortion between microphones. Some experiments using actual devices are also done to test the performance of the proposed method. It is shown that significant improvement in source localization accuracy is possible when the camera and the microphone system are well calibrated.

In a mobile communication system, the base station antennas are often required to radiate shaped-beam patterns, which have as low a level as possible toward the interference zone where the same frequency is used and, conversely, have as high a level as possible toward the service zone. So, the synthesis of the shaped-beam pattern is one of great importance to designing the base station antennas. We present a new synthesis method of the shaped-beam pattern by controlling only the excitation phases of the array elements. We first apply array theory and optimization theory to construct an objective function. Then, we employ the quasi-Newton method to minimize this nonlinear problem. Two synthesis examples based on the 16-element linear arrays are presented to demonstrate the efficiency of our method.

Time-domain integral equation (TDIE) methods, for the solution of transient electromagnetic problems have steadily increased in popularity. A novel construction method is presented to design the new temporal basis function. We first specify the principal properties of the desirable temporal basis function. According to these properties and choosing a suitable function with some parameters, we construct a nonlinear programming problem with respect to these parameters, from which the new temporal basis function can be derived. The obtained temporal function has the any-order continuous derivative and is not a truncated version of any function. By using the construction method, a new temporal function, namely, the rational fractional temporal function, is also presented. The numerical result shows that the new temporal function pushes the occurrence of the late-time instability of the transient simulation further down in time when compared to the previous temporal functions.