Search

Ridge regression is frequently utilized by both supervised learning and semisupervised learning. However, the results cannot obtain the closed-form solution and perform manifold structure when ridge regression is directly applied to semisupervised learning. To address this issue, we propose a novel semisupervised feature selection method under generalized uncorrelated constraint, namely SFS. The generalized uncorrelated constraint equips the framework with the elegant closed-form solution and is introduced to the ridge regression with embedding the manifold structure. The manifold structure and closed-form solution can better save data's topology information compared to the deep network with gradient descent. Furthermore, the full rank constraint of the projection matrix also avoids the occurrence of excessive row sparsity. The scale factor of the constraint that can be adaptively obtained also provides the subspace constraint more flexibility. Experimental results on data sets validate the superiority of our method to the state-of-the-art semisupervised feature selection methods.

In order to improve the operation performance and environmental adaptability of the manipulator, it is essential to control the position and force of the manipulator simultaneously. This paper proposed a hybrid control method of auto-dynamic bit based on bit control to realize flexible grasping of manipulator in unknown environment. The constraint conditions and control types of the hybrid force/bit control are analyzed firstly, and the dynamic model when contacting with the environment is established. Then, an improved position controller based on conventional fuzzy-PID control strategy is introduced to realize the hybrid force/bit control based on the position control by combining constraint estimation method and impedance force control. Finally, the control performance of the force/bit hybrid controller is verified by simulation experiments. The results show that the designed control strategy has relatively ideal force control precision and higher trajectory tracking ability, and the manipulator model has better compliance.

Recently, single image super-resolution (SISR), aiming to preserve the lost structural and textural information from the input low resolution image, has witnessed huge demand from the videos and graphics industries. The exceptional success of convolution neural networks (CNNs), has absolutely revolutionized the field of SISR. However, for most of the CNN-based SISR methods, excessive memory consumption in terms of parameters and flops, hinders their application in low-computing power devices. Moreover, different state-of-the-art SR methods collect different features, by treating all the pixels contributing equally to the performance of the network. In this paper, we take into consideration both the performance and the reconstruction efficiency, and propose a Light-weight multi-scale attention residual network (MSAR-Net) for SISR. The proposed MSAR-Net consists of stack of multi-scale attention residual (MSAR) blocks for feature refinement, and an up and down-sampling projection (UDP) block for edge refinement of the extracted multi-scale features. These blocks are capable of effectively exploiting the multi-scale edge information, without increasing the number of parameters. Specially, we design our network in progressive fashion, for substituting the large scale factors ( × 4) combinations, with small scale factor ( × 2) combinations, and thus gradually exploit the hierarchical information. In parallel, for modulation of multi-scale features in global and local manners, channel and spatial attention in MSAR block is being used. Visual results and quantitative metrics of PSNR and SSIM exhibit the accuracy of the proposed approach on synthetic benchmark super-resolution datasets. The experimental analysis shows that the proposed approach outperforms the other existing methods for SISR in terms of memory footprint, inference time and visual quality.

Recently, deep learning algorithms have demonstrated remarkable performance in object detection tasks and have applied to many fields such as intelligent networked vehicles and drones. Different from other fields like video surveillance, the fields of intelligent networked vehicles and drones only have resource-limited platform and have high requirements in model storage and real-time inference. To satisfy these requirements, model compression approaches are developed to reduce model parameters and computation. As channel pruning is a coarse-grained hardware-friendly weight pruning method, it is widely used in model compression. Most studies have only focused on utilizing the scale factor of Batch Normalization (BN) layers to determine whether to prune a certain channel. However, the shift factor turns to very important when its value is large. The channel with large shift factor has a great influence on subsequent layers. In our paper, we take shift factor into consideration to protect these channels. We first introduce network building blocks to analyze YOLOv3-SPP3 model and demonstrate the importance of shift factor on the channel saliency in the pruning process. Then we present a new pruning method called high-ratio channel pruning (HRCP) to improve the performance of the pruned model based on definition of the channel saliency with scale factor and shift factor of BN layer. Our experimental results show that HRCP performs better than SlimYOLOv3 when model volumes are reduced by 50%, 88%, and 92% accordingly. In high-ratio pruning, our HRCP can be pruned more 9.7% volumes than SlimYOLOv3 under the same performance condition.

The continuously developing Global Navigation Satellite Systems (GNSS), in particular with the latest operational BDS and Galileo, provide good opportunities to enhance the performance of precise positioning. This study evaluates the performance of high-rate precise point positioning (PPP) applied for seismogeodesy with very recent BDS and Galileo. The static experiments with 20-Hz GNSS data show that Galileo and BDS-3/BDS-2 achieve comparable performances, with average 3D RMS values of 0.6–1.3 cm and 0.8–1.2 cm, respectively. In the dynamic experiments, we use a single-axis earthquake simulator of Quanser Shake Table II to simulate seismic waveforms of real earthquakes and collect 20-Hz GNSS data to evaluate the PPP performance. Taking the displacements recorded by the embedded encoder of the shake table as the truth, the accuracies of PPP displacements during the seismic events are 4–8 mm for GPS, 6–9 mm for BDS-2, and 3–6 mm for Galileo and BDS-3/BDS-2. By combining BDS-3 with BDS-2, the positioning accuracy of BDS is improved by 50% with respect to BDS-2 only. Galileo and BDS-3/BDS-2 perform slightly better than GPS, which is likely because the two latest operational systems use the more precise passive hydrogen maser clocks and have lower signal-in-space range error. The multi-GNSS solutions with GPS + Galileo + BDS-3/BDS-2 further improve the accuracy reaching 2–5 mm. We also identify a scale factor in the high-rate PPP displacements, which is probably induced by the dynamic stress error and related to the oscillating frequency. The scale factor reaches the maximum of 1.71 at an oscillating frequency around 2.5 Hz and then descends as the frequency increases. The improved performance of high-rate PPP augmented by the recent BDS-3 and Galileo is helpful to improve the practical level of high-rate GNSS in seismogeodesy and expand its applications to many other high-rate and high-accuracy applications.

Recently, deep learning achieves competitive accuracy and robustness and dramatically improves the performance of target scale estimation through pre-trained special network branches. Yet, a fast and robust scale estimation method is still a challenging problem for visual object tracking. Early correlation filter tracking algorithm uses a multiscale search method to estimate the scale with the constant number of scale factors and invariant aspect ratio, which is redundant for the video frames with little or no scale change. Also, an independent network branch for target scale state is proposed, but the training network needs an abundance of datasets, and the effect is not very stable for the unseen target object. Aiming at the problems of existing scale estimation solutions, several variable scale learning methods are proposed to explore the scale change of the target. Firstly, we proposed a variable scale factor learning method, which makes us rid of the commonly used multiscale search with the flaws of fixed scale factors. Secondly, we used a multiscale aspect ratio solution to make up for invariant aspect ratio. Thirdly, the first and second scale methods were combined to propose a variable scale aspect ratio estimation method. Finally, the proposed scale estimation methods were embedded into the state-of-the-art ECO (Efficient Convolution Operators) and ATOM (Accurate Tracking by Overlap Maximization) trackers to replace the original scale methods for verifying the effectiveness of our proposed method. Extensive experiments on OTB100, UAV123, TC128 and LaSOT datasets demonstrate that the tracking performance can be improved effectively by using the proposed scale methods.

Recent research on single image super-resolution (SISR) has achieved great success due to the development of deep convolutional neural networks. However, most existing SISR methods merely focus on super-resolution of a single fixed integer scale factor. This simplified assumption does not meet the complex conditions for real-world images which often suffer from various blur kernels or various levels of noise. More importantly, previous methods lack the ability to cope with arbitrary degradation parameters (scale factors, blur kernels, and noise levels) with a single model. A few methods can handle multiple degradation factors, e.g., noninteger scale factors, blurring, and noise, simultaneously within a single SISR model. In this work, we propose a simple yet powerful method termed meta-USR which is the first unified super-resolution network for arbitrary degradation parameters with meta-learning. In Meta-USR, a meta-restoration module (MRM) is proposed to enhance the traditional upscale module with the capability to adaptively predict the weights of the convolution filters for various combinations of degradation parameters. Thus, the MRM can not only upscale the feature maps with arbitrary scale factors but also restore the SR image with different blur kernels and noise levels. Moreover, the lightweight MRM can be placed at the end of the network, which makes it very efficient for iteratively/repeatedly searching the various degradation factors. We evaluate the proposed method through extensive experiments on several widely used benchmark data sets on SISR. The qualitative and quantitative experimental results show the superiority of our Meta-USR.

In this paper, the authors presented a new fractal microstrip patch antenna for Dedicated Short Range Communication (DSRC) (5.850–5.925 GHz) (IEEE 802.11p) service band. The proposed structure is obtained by etching the Minkowski boxes in all the four corners of a square geometry. Further, a square, obtained by imposing a scale factor of 50% of the existing sides, has been erased from each of the corners of the resultant structure. The final structure is obtained by superimposing this existing one with its $$45^\circ $$ rotated version. In order to get proper circular polarization and satisfactory S $$_{11}$$ characteristics from this structure, three diagonal grooves and four cross-coupled rectangular thin slits are carved from the resultant fractal geometry. A single layered Roger 5880 PCB laminated material is used to design the structure and simulated it using CST Microwave studio. The proposed antenna is a right-handed circularly polarized (RHCP) one with resonating frequency of 5.91 GHz. The measured CP gain of the antenna at resonating frequency is 5.58 dBic, and it gives a 3-dB axial bandwidth of 102 MHz which entirely covers the DSRC band. The measured radiation efficiency of the antenna is 94% and its computed aperture efficiency becomes 72%. The measured result of the prototype makes an excellent agreement with the simulated counterpart.

This study investigated the variation in electronic portal imager pixel scale at the isocenter plane for Elekta Agility linear accelerators. An in‐house MATLAB script was written to process and calculate the pixel scale based on a metal calibration plate supplied by Elekta. Eight pixel plates were compared and found to have manufacturing tolerances within 0.1 mm of nominal dimensions. The impact of these variations on pixel scale factor was negligible, and plates could be used interchangeably. Uncertainties from other parameters such as source‐to‐surface distance and user variability summed to a combined uncertainty of 0.0003 mm/pixel, compared to a pixel scale range of 0.003 mm/pixel measured across 10 machines. Most of the inter‐machine variation was shown to be attributable to differences in source‐to‐panel distance. Other factors such as focal spot size and shape, electronic portal imager manufacturing consistency, panel sag, and setup errors may account for the residual variation. Individual characterization of machine and imaging panel pixel scale factors is important to ensure accurate geometric information is derived from electronic portal images, which is critical where the portal imager is used for multi‐leaf collimator calibration or other clinical tasks.

The rotational inertial navigation system (RINS) could greatly improve navigation performance by rotating the inertial measurement unit with gimbals. And the self-calibration of error parameters could be achieved conveniently. Among the rate-biased Ring laser gyro (RLG), the input angular velocity can reach tens of degrees per second. Therefore, the attitude errors caused by scale factor is dozens of times that of the other RINS. In this paper, gyro scale factor errors are calibrated through optimal estimation with attitude errors in dual-axis laser RINS. During the movement of vehicle, the dynamic gyro scale factor errors calibration can be performed online by adding parking points. Moreover, the parking points can be accurately determined by the statistical characteristics of accelerometers and gyros output information, without extern sensors. The experiment results indicate that the calibration repeatability scale factor parameters are less than 1ppm, and the attitude errors cause by the scale factor error is reduce from tens of arc seconds to 10 arc seconds, which verifies the effectiveness and accuracy of the method proposed in this paper. Through the compensation of calibration results, the attitude error of RINS is further improved in both the static and dynamic experiment.

Process intensification has emerged as a strategy to satisfy the objectives of process efficiency, sustainable development and economic competitiveness. Unfortunately, intensified technologies are too numerous to fit in a time-constrained academic syllabus. For high pedagogic impact, appropriate case studies should be favored to cover a wide range of strategies while demonstrating these opportunities, using accessible tools and concepts. The developed case study demonstrates the gain of sustainability offered by applying multi-scale design to a cyclonic separation, enabling miniaturization at constant efficiency. The analysis leads to well-known multi-cyclonic systems. This example is particularly striking for students since multi-cyclonic vacuum cleaners are a domestic illustration of this strategy. The analysis aims to miniaturize the separator, at constant efficiency and dissipated mechanical power. By applying a scale factor to cyclone dimensions, while maintaining the efficiency, a first strategy demonstrates that miniaturization alone cannot be sustainable. The best strategy considers parallel miniature cyclones, which enables to satisfy all objectives: the volume and dissipated mechanical power can be reduced simultaneously, requiring a multi-objective optimization. Scenarios for integration of the case study in various curricula are proposed, by differentiating the learning outcomes as a function of student level and course objectives.

As the indoor dynamic target localization does not detect and repair the circumferential jump value in time, which leads to large position and attitude errors and low-velocity stability, a combined Global Navigation Satellite System/Simultaneous Localization and Mapping- (GNSS/SLAM-) based high-precision indoor dynamic target localization method is proposed. The method uses Empirical Mode Decomposition (EMD) to decompose the noisy signal, obtains the noise energy as the dominant mode from the decomposed components, extracts the useful signal energy as the main dividing point, removes the high-frequency signal, constructs the low-frequency component to realize low-pass filtering and denoising, selects a suitable threshold processing function to make the high-frequency signal component retain the detailed signal effectively to realize high-frequency component denoising, detects and fixes the circumferential jump of the observed data, and detects and fixes the circumferential jump of each frequency. The indoor dynamic target positioning method is established by combining GNSS/SLAM to achieve high accuracy target positioning. The experimental results show that the position and attitude errors are small, and the velocity is stable, which indicates that the position information is closer to the dynamic target, i.e., the target positioning is more accurate. To address the problems of scale drift and frequent initialization due to environmental factors in monocular vision SLAM, an Ultra Wideband (UWB)/vision fusion localization model that takes into account the scale factor is proposed, which can give full play to the complementary characteristics between UWB and vision; solve the problems of visual initialization, scale ambiguity, and absolute spatial reference; and improve UWB localization accuracy and localization frequency as well as reducing the number of base stations. The model can reliably achieve 0.2 m accuracy in indoor environments with sparse texture or frequent light changes.

The development of wave solid-state gyroscopes (VTG) is one of the promising areas of development of gyroscopic angular velocity sensors. VTG from the standpoint of manufacturing technology, tuning and control systems, as well as accuracy characteristics, has a number of advantages compared to other types of gyroscopes. When developing VTG, they strive to reduce the gyroscope's own care, zero signal bias, and the non-linearity of the scale factor in the operating temperature range However, when creating the device, due attention is often not paid to the existing opportunities to improve the dynamic accuracy of the gyroscope by developing promising structural solutions for building control circuits and information processing. The solution to this problem was the goal of the work.Using the methods of the theory of automatic control, the dynamics of a wave solid-state gyroscope with a metal resonator and piezoelectric elements in the closed-loop mode of Сoriolis acceleration compensation are studied. Piezoelectric elements perform the functions of displacement and force sensors.Two promising structural solutions for constructing VTG control and information processing circuits are proposed and considered. Relations are established for selecting the parameters of the links of these contours, which provide an increase in the dynamic accuracy of the gyroscope. In the first case, the proposed structure for constructing the VTG allows us to significantly reduce the dynamic errors caused by the difference in the scale coefficient of the VTG at different frequencies of the measured angular velocity in the bandwidth. Such a structure for constructing a VTG can be recommended when solving a measurement problem in which it is necessary to accurately measure the angular velocity, and the phase lag of the output signal in relation to the measured angular velocity is of secondary importance. In the second case, the proposed structure of the VTG construction corresponds to the transfer function of the relative measurement error with secondorder astatism, and the absolute measurement error in the frequency band of 10 Hz does not exceed 0.1 %.

Current works on super-resolution have obtained satisfactory results since the advance of the convolution neural network. Nevertheless, most previous works use one network for one integer scale factor so ignore the super-resolution of the arbitrary scale factor. In this work, we propose a novel approach called Global Enhanced Upscale Network (GEUN) to tackle super-resolution with a single model adapting the arbitrary scale factor. In our GEUN, we propose the Global Enhanced Upscale module to replace the conventional upscale module. Our GEUN can upscale low-resolution images with an arbitrary scale factor through only one model. Extensive experimental results demonstrate the superiority of our GEUN.

Purpose Laboratory calibration methods are time-consuming and require accurate devices to find the error coefficients of the low-cost microelectromechanical system (MEMS) accelerometer. Besides, low-cost MEMS sensors highly depend on temperature because of their silicon property and the effect of temperature on error coefficients should also be considered for compensation. This paper aims to present a field calibration method in which the accelerometer is placed in different positions without any accurate equipment in a few minutes and its temperature is changed by a simple device like a hairdryer. Design/methodology/approach In this paper, a non-linear cost function is defined based on this rule that the magnitude of the acceleration measured by the accelerometer in static mode is equal to the gravity plus error factors. Also, the dependency of error coefficients of the accelerometer is presented as a second-order polynomial in this cost function. By minimizing the cost function, the accelerometer error coefficients include bias, scale factor and non-orthogonality and their temperature dependency are obtained simultaneously. Findings Simulation results in MATLAB and empirical results of a MPU6050 accelerometer verify the good performance of the proposed calibration method. Originality/value Finding a fast and simple field calibration method to calibrate a low-cost MEMS accelerometer and compensate for the temperature dependency without using accurate laboratory equipment can help a wide range of industries that use advanced and expensive sensors or use expensive laboratory equipment to calibrate their sensors, to decrease their costs.

In this work, we proposed a novel Coordinate Rotation DIgital Computer (CORDIC) rotator algorithm that converges faster by performing radix-2,4 and 16 CORDIC iterations while maintaining the scale factor implicitly constant. A mixed-radix is used to achieve convergence faster to reduce the computational latency of the CORDIC algorithm. The main concern of the higher radix CORDIC algorithm is the compensation of a variable scale factor. To solve this problem, the Taylor series approximation of sine and cosine is proposed for a higher radix CORDIC algorithm to achieve the scaling-free rotation of the two-dimensional vector. The scaling-free rotation of the proposed CORDIC algorithm removes the read-only memory (ROM) needed to store scale factor of higher radix CORDIC algorithm. Further, the proposed CORDIC algorithm is designed in rotation mode and optimized by removing the Z datapath for the digital signal processing (DSP) applications for which the angle of rotation is known in advance. Finally, the multipath delay commutator (MDC) fast Fourier transform (FFT) algorithm is implemented with the proposed CORDIC algorithm based rotator on FPGA. The proposed design is compared with existing designs. In a comparison between the radix-16 CORDIC rotator based FFT implementation and our proposed implementation, it has been found out that implementation proposed in this article has used 17% fewer resources.

In the milling process, chatter is easy to occur and has a very adverse impact on the quality of the workpiece and the production efficiency. A chatter feature extraction method based on optimized variational mode decomposition (OVMD) and multi-scale permutation entropy (MPE) was proposed to solve the problem that it is difficult to detect the machining chatter state during milling. The methodology presented in this article allows the occurrence of machining chatter to be effectively identified through real-time digital signal processing and analysis. First, in order to solve the problem of variational mode decomposition (VMD) parameter selection, an automatic selection method based on particle swarm optimization (PSO) and the maximum crest factor of the envelope spectrum (CE) was proposed. Then, the decomposed signal was reconstructed based on the energy ratio. In order to solve the problem that the single-scale permutation entropy (PE) cannot detect milling chatter well, the MPE was introduced to detect milling chatter. Finally, experimental verification was carried out, and the MPE of the reconstructed signals at different scales was extracted and analyzed. The results show that using the OVMD algorithm to process the signals can significantly improve the discrimination of MPE. With the increase of the scale factor, the MPE of the milling signals tends to decrease. At the same time, MPE is better than single-scale PE in chatter detection, and the MPE at scale factor of 4 is more conducive to chatter detection.

The Fiedler vector of a graph is the eigenvector corresponding to the algebraic connectivity, which is the second-smallest eigenvalue (counting multiple eigenvalues separately) of the corresponding Laplacian matrix. We propose a continuous-time distributed control protocol to drive the value of the state variables of a network toward the Fiedler vector, up to a scale factor. Our protocol is unbiased and robust with respect to the initial network state, but the knowledge of the algebraic connectivity is required. By means of the proposed control law, we design a local state feedback that achieves desynchronization on arbitrary undirected connected networks of diffusively coupled harmonic oscillators. We provide numerical simulations to corroborate the theoretical results.

The random vector functional link (RVFL) network is suitable for solving nonlinear problems from transformer fault symptoms and different fault types due to its simple structure and strong generalization ability. However, the RVFL network has a disadvantage in that the network structure, and parameters are basically determined by experiences. In this paper, we proposed a method to improve the RVFL neural network algorithm by introducing the concept of hidden node sensitivity, classify each hidden layer node, and remove nodes with low sensitivity. The simplified network structure could avoid interfering nodes and improve the global search capability. The five characteristic gases produced by transformer faults are divided into two groups. A fault diagnosis model of three layers with four classifiers was built. We also investigated the effects of the number of hidden nodes and scale factors on RVFL network learning ability. Simulation results show that the number of implicit layer nodes has a large impact on the network model when the number of input dimensions is small. The network requires a higher number of implicit layer neurons and a smaller threshold range. The size of the scale factor has significant influence on the network model with larger input dimension. This paper describes the theoretical basis for parameter selection in RVFL neural networks. The theoretical basis for the selection of the number of hidden nodes, and the scale factor is derived. The importance of parameter selection for the improvement of diagnostic accuracy is verified through simulation experiments in transformer fault diagnosis.

A maximum likelihood estimator is presented for self-calibrating both accelerometers and gyroscopes in an inertial sensor array, including scale factors, misalignments, biases, and sensor positions. By simultaneous estimation of the calibration parameters and the motion dynamics of the array, external equipment is not required for the method. A computational efficient iterative optimization method is proposed where the calibration problem is divided into smaller subproblems. Further, an identifiability analysis of the calibration problem is presented. The analysis shows that it is sufficient to know the magnitude of the local gravity vector and the average scale factor gain of the gyroscopes, and that the array is exposed to two types of motions for the calibration problem to be well defined. The proposed estimator is evaluated by real-world experiments and by Monte Carlo simulations. The results show that the parameters can be consistently estimated and that the calibration significantly improves the accuracy of the motion estimation. This enables on-the-fly calibration of small inertial sensors arrays by simply twisting them by hand.

In this paper, we consider a compensation-type quartz pendulum accelerometer, a feature of which is the use of a digital feedback amplifier in the control loop (for the implementation of a discrete controller and algorithmic compensation for errors from the pendulum unbalance and temperature dependences of the zero offset and scale factor). The problem is related to the construction of a digital control system for the considered accelerometer, which provides a high-precision measurement of the projection onto the sensitivity axis of the actual apparent acceleration of the object on which the accelerometer is installed. A solution to this problem, which allows finding digital control algorithms using various methods of synthesis of discrete controllers of the control loop, is provided. The obtained solutions in the time and frequency domains are analyzed, confirming the high quality of regulation. The results of experimental studies are presented.

Doppler Velocity Logs (DVL) is an important information source for underwater vehicle navigation because it can provide accurate velocity information. As the main system, SINS/DVL integrated navigation is widely used in the underwater vehicle. However, DVL scale factor error, and misalignment with inertial measurement unit (IMU) are the two sources of integration inaccuracy. A novel calibration method aided GNSS is proposed in this paper, which is designed based on velocity observation and position observation respectively. The proposed method combines the Quasi-Newton method and Quaternions method, and we name it Quasi-Newton Quaternion (QNQ) calibration method. In order to verify the performance of the proposed method, another calibration method based on the Kalman filter is introduced in simulation, vehicle tests, and Sea trial. The test results indicate that the proposed calibration method can perform a good estimation of DVL error.

To assess the safety of structures after the occurrence of an earthquake or decide about demolition or repair of structures, maximum residual interstory drift ratio (MRD) is one of the most important engineering demand parameters. For evaluating the probability of exceedance of MRD in structure from a specified limit, nonlinear dynamic analyses are performed by employing an intensity measure (IM). IMs link the structural seismic response analyses with the seismic hazard analysis. Employing an optimal IM in seismic evaluation of structures makes the results more realistic and reliable. The power of an IM to predict the seismic response of structures with low dispersion is called efficiency. Moreover, the ability of an IM for rendering the seismic response of structures independent from the other features of the ground motion is named sufficiency. In this study, the efficiency and sufficiency of some traditional and advanced IMs (including scalar and vector ones) for predicting MRD capacity of steel buckling restrained braced frames are evaluated. It is shown that all of the IMs considered do not have sufficiency with regard to scale factor (SF). Then, (Sa(T1),SaRatioOp‒D) as an optimal vector IM (optimized by multi-objective particle swarm optimization algorithm) having improved sufficiency with regard to SF is proposed.

Research on super-resolution has achieved great success on synthetic data with deep convolutional neural networks. Some recent works tend to apply super-resolution to practical scenarios. Learning an accurate and flexible model for super-resolution of arbitrary scale factor is important for realistic applications, while most existing works only focus on integer scale factor. In this work, we present a residual scale attention network for super-resolution of arbitrary scale factor. Specifically, we design a scale attention module to learn discriminative features of low-resolution images by introducing the scale factor as prior knowledge. Then, we utilize quadratic polynomial of the coordinate information and scale factor to predict pixel-wise reconstruction kernels and achieve super-resolution of arbitrary scale factor. Besides, we use the predicted reconstruction kernels in image domain to interpolate low-resolution image and obtain coarse high-resolution image first, then make our main network learn high-frequency residual image from feature domain. Extensive experiments on both synthetic and real data show that the proposed method outperforms state-of-the-art super-resolution methods of arbitrary scale factor in terms of both objective metrics and subjective visual quality.

Effects of a fully natural aging of MEMS accelerometers are evaluated with regard to changes in their performance. Two models of commercial dual-axis accelerometers (two pieces of ADXL 202E and 203 by Analog Devices Inc.) with analog outputs were tested over a period of about 10 and 4 years, respectively. A custom computer controlled test rig was used for performing relevant experimental studies, employing the gravitational acceleration as the reference source. A methodology of determining the proposed indicators of aging phenomena is presented and discussed. Changes of the offset voltage and the scale factor were observed and a way of evaluating the overall error due to combined influence of these two parameters is proposed. It was found out that the changes of the output voltage generated by the tested accelerometers were considerable, resulting in respective maximal errors of about 52 mg (2.6%) (ADXL 202E) or 20 mg (1%) (ADXL 203). Simple ways of reducing the effects of aging are proposed.

The security of multimedia information is a prime concern in the present digital world. As a remedy, a robust color image watermarking in the transform domain using artificial intelligence is reported in this article. A color host image is secured by embedding a color watermark by utilizing the singular value decomposition and discrete wavelet transform. The color watermark information is scrambled with a chaotic map to give an additional stage of protection to overcome the problem of illegal copying and alteration of digital data by unauthorized users. All three RGB color channels of host singular values are modified with principal components of respective RGB information of scrambled watermark image. Ownership of data is provided by extracting the watermark in the presence of a secret key used while embedding process. To overcome the trade-off between the imperceptibility and robustness of the proposed algorithm artificial bee colony is used to optimize the scaling factor. The implementation of the proposed algorithm is performed on the MATLAB tool to compute the performance against the intentional and non-intentional image processing attacks. Comparative analysis with other related watermarking algorithms proves the robust, secure, and invisible nature of the proposed watermarking scheme.

In some applications, such as surveillance and biometrics, image enlargement is required to inspect small details on the image. One of the image enlargement approaches is by using convolutional neural network (CNN)-based super-resolution construction from a single image. The first CNN-based image super-resolution algorithm is the super-resolution CNN (SRCNN) developed in 2014. Since then, many researchers have proposed several versions of CNN-based algorithms for image super-resolution to improve the accuracy or reduce the model’s running time. Currently, some algorithms still suffered from the vanishing-gradient problem and relied on a large number of layers. Thus, the motivation of this work is to reduce the vanishing-gradient problem that can improve the accuracy, and at the same time, reduce the running time of the model. In this paper, an enhanced dense space attention network (EDSAN) model is proposed to overcome the problems. The EDSAN model adopted a dense connection and residual network to utilize all the features to correlate the low-level feature and high-level feature as much as possible. Besides, implementing the convolution block attention module (CBAM) layer and multiscale block (MSB) helped reduce the number of layers required to achieve comparable results. The model is evaluated through peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) metrics. EDSAN achieved the most significant improvement, about 1.42% when compared to the CRN model using the Set5 dataset at a scale factor of 3. Compared to the ERN model, EDSAN performed the best, with a 1.22% improvement made when using the Set5 dataset at a scale factor of 4. In terms of overall performance, EDSAN performed very well in all datasets at a scale factor of 2 and 3. In conclusion, EDSAN successfully solves the problems above, and it can be used in different applications such as biometric identification applications and real-time video applications.

In recent years, navigation technology has rapidly developed. System-level calibration technology for inertial navigation system (INS) has been widely used because it does not rely on high-precision equipment. However, INS introduces a bad coupling error in high-precision ring laser gyroscope (RLG) scale-factor error estimation, accompanied by a large accelerometer-slope bias error. To solve this problem, an improved system-level fitting calibration method is proposed, which changes the 90° rotation increment to 450°, while maintaining the original calibration sequence. Compared with the traditional system-level calibration method, the calibration error of the gyroscope scale-factor error is reduced to 0.2 times, under simulation conditions, without affecting the calibration accuracy of the other parameters. The extreme difference of the gyroscope scale-factor calibration error on multiple experiments reduces from 4.75 to 1.40 ppm under experimental-verification conditions. The mean values of the gyroscope scale factor were closer to the real reference values (no more than 2 ppm). The integral calibration results meet the requirements of a high-precision RLG INS. The proposed method maintains the original calibration sequence and exhibits small changes with high practicability. The proposed method is suitable for applications with high requirements for gyroscope scale factor accuracy, such as highly dynamic aircraft.

Deep neural networks with single upsampling have achieved the improvement of performance for single image super-resolution. However, these networks lose a lot of details of low-resolution image in the reconstruction process. In this paper, we propose a gradual deep residual network for super-resolution (GDSR), which consists of multiple reconstruction network with 2 scale factor (2X reconstruction network). In 2X reconstruction network, a residual block connected by residual (RBR) is proposed to form a deep residual network, which is used to extract the depth features of low-resolution images; then the extracted features are upsampled into the features of high-resolution image by sub-pixel convolutional layer. GDSR gradually reconstructs high-quality high-resoluiton images from low-resolution images by multiple 2X reconstruction networks. Extensive experiments on benchmark datasets demonstrate that the proposed GDSR outperforms the state-of-the-art methods in terms of quantitative evaluation, visual evaluation, and execution time evaluation.

Under the influence of deep learning, many trackers have emerged recently. Among them, Siamese network reaches a pleasant balance between accuracy and speed, but its tracking performance still lags behind other trackers. In this paper, we have proposed a Hierarchical Correlation Siamese Network (HC-Siam) for object tracking. The tracker uses convolutional features of each layer to compare the correlation and identifies the position of the tracking object depending on the greatest correlation. Meanwhile, we have designed a Correlation Attention Module (CA-Module). For various objects, this module can assign different weights to the hierarchical correlation and help the network choose the distinct correlation from the hierarchical correlation. Besides, objects’ size and scale constantly varied during tracking, we claimed to use the separate scale factor in the wide and high directions to decrease the deformation of bounding boxes and increase the accuracy of our tracker. On the OTB dataset, the accuracy of HC-Siam is 6.5% higher than the baseline, and the speed of our tracker can reach 85 fps. On the VOT dataset, HC-Siam also has better performance in speed and accuracy.

Distributed position and orientation systems (DPOSs) can provide abundant time-spatial information for interferometric synthetic aperture radar (InSAR) in airborne earth observation systems. However, some key error terms have not been taken into consideration in the traditional low-order error model, which suppresses the performance of the slave POS and further cannot meet the compensation precision of InSAR. To improve the compensation precision, a precise high-order error model with 45 dimensions was derived. Not only does it take into account the influence of scale factor errors and installation errors of the gyro and accelerometer, but it also makes use of random constants and a first-order Markov process model to describe the gyro drift and accelerometer bias. In addition, the flexure angle and its angular rate were added to the state variables of the transfer alignment model. Based on the model, a measurement equation for attitude errors that considers flexure was deduced. Then, a transfer alignment model based on the matching algorithm including position-velocity-attitude was designed. Finally, the proposed model was validated by simulated and real tests, and the experimental results show that its performance is obviously better than that of the traditional model.

The problem of extending the field of effective use of strapdown inertial navigation systems to the class of highly dynamic objects is considered. To ensure effective motion control, high requirements are put forward for the accuracy of determining the orientation angles and coordinates of an object, which can only be satisfied if high-tech precision gyroscopes and accelerometers are used in SINS. Also, the use of SINS is limited by the possible dynamics of the object. The solution to this problem is based on the possibility of taking into account the conditions of the future operation of the SINS due to the rational choice of the placement or configuration of the gyroscopes' sensitivity axes. This choice makes it possible to reduce the requirements for the measuring range of the meter unit while maintaining functionality as part of the strapdown inertial navigation system as a whole. For this purpose, it is proposed to place gyroscopes in a special way relative coordinate system of the instrument. This arrangement allows to use gyroscopes with a relatively small operating range to measure a significant object rotation speed. The choice of such a configuration is realized through the solution of an certain optimization problem with constraints. In the general case, the problem is solved numerically, and in some cases, an analytical solution is possible. An additional advantage of the optimal configuration is the increase in the accuracy of the SINS by reducing the effect of the error of the scale factor of gyroscopes. The resulting solution is applied in practice during the design and creation of a specialized SINS.

To address the problems of insufficient detail extraction and long training time in the super-resolution reconstruction of chest X-ray images, a method of chest X-ray images super-resolution reconstruction using recursive neural network is proposed in this paper. Firstly, this paper designs a lightweight recursive network as the main branch, which solves the problem of training difficulty and time-consuming. Then, to overcome the lack of detail extraction in chest X-ray image, a detail complementary model is designed as another branch of the network to solve the problem of shallow information loss. Finally, the optimized activation function is used to reduce the loss of texture details and make the reconstructed image more complete and richer. When the scale factor is 2, the experimental results show that compared with other methods based on deep learning, such as the deep recursive neural network (DRCN), the details of chest X-ray images reconstructed by our method are more abundant. Specifically, the average value of PSNR and SSIM were improved by 0.17 dB and 0.0013 respectively. Moreover, the reconstruction speed of the images was increased by about 16% compared with DRCN.

This paper proposes an enhanced tightly-coupled sensor fusion scheme using a monocular camera and ultra-wideband (UWB) ranging sensors for the task of simultaneous localization and mapping. By leveraging UWB data, the method can achieve metric-scale, drift-reduced odometry and a map consisting of visual landmarks and UWB anchors without knowing the anchor positions. Firstly, the UWB configuration accommodates any degenerate cases with an insufficient number of anchors for 3D triangulation ( $$N\le 3$$ and no height data). Secondly, a practical model for UWB measurement is used, ensuring more accurate estimates for all the states. Thirdly, selected prior range measurements including the anchor-world origin and anchor–anchor ranges are utilized to alleviate the requirement of good initial guesses for anchor position. Lastly, a monitoring scheme is introduced to appropriately fix the scale factor to maintain a smooth trajectory as well as the UWB anchor position to fuse camera and UWB measurement in the bundle adjustment. Extensive experiments are carried out to showcase the effectiveness of the proposed system.

Advances in deep learning make monocular vision approaches attractive for the autonomous driving domain. This work investigates a method for estimating the speed of the ego-vehicle using state-of-the-art deep neural network based optical flow and single-view depth prediction models. Adopting a straightforward intuitive approach and approximating a single scale factor, several application schemes of the deep networks are evaluated and meaningful conclusions are formulated, such as: combining depth information with optical flow improves speed estimation accuracy as opposed to using optical flow alone; the quality of the deep neural network results influences speed estimation performance; using the depth and optical flow data from smaller crops of wide images degrades performance. With these observations in mind, a RMSE of less than 1 m/s for ego-speed estimation was achieved on the KITTI benchmark using monocular images as input. Limitations and possible future directions are discussed as well.

Recent advanced deep learning studies have shown the positive role of feedback mechanism in image super-resolution task. However, current feedback mechanism only calculates residual errors of images with the same resolution without considering the useful features that may be carried by different resolution features. In this paper, to explore the potential of feedback mechanism, we design a new network structure (progressive up- and downsampling back-projection units) to construct a generative adversarial network for single image super-resolution and use progressive growing methodologies to train it. Unlike previous feedback structure, we use progressively increasing scale factor to build up- and down-projection units, which aims to learn fruitful features across scales. This method allows us to get more meaningful information from early feature maps. Additionally, we train our network progressively; in the process of training, we start from single layer network structure and add new layers as the training goes on. By this mean, the training process can be greatly accelerated and stabilized. Experiments on benchmark dataset with the state-of-the-art methods show that our network achieves 0.01 dB, 0.11 dB, 0.13 dB and 0.4 dB better PSNR results than that of RDN+, MDSR, D-DBPN and EDSR on 8 $$\times $$ enlargement, respectively, and also achieves favorable performance against the state-of-the-art methods on 2 $$\times $$ and 4 $$\times $$ enlargement.

Frequency drop due to loss of massive generation is a threat to power system frequency stability. Under-frequency load shedding (UFLS) is the principal measure to prevent successive frequency declination and blackouts. Based on traditional stage-by-stage UFLS scheme, a new continuous UFLS scheme is proposed in this paper to shed loads proportional to frequency deviation. The characteristic of the proposed scheme is analyzed with a closed-form solution of frequency dynamics. Frequency threshold and time delay are added to make the proposed scheme practical. A line-by-line scheme based on precise load control is introduced to implement the continuous scheme for systems without enough continuously controllable loads. The load shedding scale factor of the proposed scheme is tuned with an analytical method to achieve adaptability to different operating conditions. The adaptability of the proposed scheme is validated with 39-bus New England model and simplified Shandong Power Grid of China.

The recent advances in deep convolutional neural networks (DCNNs) have convincingly demonstrated high-capability reconstruction for single image super-resolution (SR). However, it is a big challenge for most DCNNs-based SR models when the scaling factor increases. In this paper, we propose a novel Progressive Residual Network (PRNet) to integrate hierarchical and scale features for single image SR, which works well for both small and large scaling factors. Specifically, we introduce a Progressive Residual Module (PRM) to extract local multi-scale features through dense connected up-sampling convolution layers. Meanwhile, by embedding residual learning into each module, the relative information between high-resolution and low-resolution multi-scale features is fully exploited to boost reconstruction performance. Finally, the scale-specific features are fused to the reconstruction module for restoring the high-quality image. Extensive quantitative and qualitative evaluations on benchmark datasets illustrate that our PRNet achieves superior performance and in particular obtains new state-of-the-art results for large scaling factors such as 4 × and 8 ×.

In current adaptive steganography study, the distortion function is used to describe pixel modification distortion cost. Distortion cost can describe the influence of pixel modification to the cover. And it plays an important actor in adaptive stegaography. Based on previous researches, in this paper, an Anisotropic Distortion cost Update strategy (ADU strategy) is proposed to adaptively update the initial distortion cost (obtained by HILL .etc) to describe the interaction between multiple embedding modifications more accurately. Based on steanographic security theorty, it is proved that clustering modification strategy can improve steganographic security through theoretical derivation. And through study on steganalysis features, it is analyzed and proved that the determination of optimal modification method of the central pixel is anisotropic and much more complicated. Following these two proofs, the image gradient and thermal conductivity are used to quantify the anisotropy of 4 neighboring pixels. Then the optimal modification method of the central pixel is calculated. Finally, the distortion cost of the central pixel can be updated based on the optimal modification method. Experiments with the optimal embedding simulator show that the ADU strategy has better performance when the scaling factor is 2; the proposed ADU strategy can effectively improve the steganography schemes, especially for steganalytic performance with maxSRMd2 features.

Accurate extraction of structural edge information of aluminum foam is an important method to study the complex structural properties of aluminum foam, but the conventional single-scale edge detection method is difficult to achieve complete extraction of structural edge information of aluminum foam. However, the multi-scale fusion edge detection method based on Gaussian smoothing also has some problems, such as strong edge diffusion, weak edge degradation, edge pixel movement and so on. In order to solve the shortcomings of the above methods, this paper proposes a multi-scale edge fusion algorithm based on texture suppression, which can extract the edge information of aluminum foam structure more accurately and completely. Firstly, preprocess the image. The illumination component of the image is extracted by the multi-scale fusion method, and the luminance of the image is corrected by the adaptive luminance correction method based on the two-dimensional gamma function. Secondly, construct multi-scale space. It is proposed to construct the guiding image of the guiding filtering by using the bilateral texture filtering and construct the multi-scale space by changing the scale factor of the guiding filtering. Both bilateral filtering and guided filtering have the function of suppressing the texture information of the image while maintaining the structural edge features of the image. Finally, extract edge seeds and fuse multi-scale edges. A new multi-scale image edge fusion algorithm is proposed, which uses seed edges as a medium to gradually merge multi-scale image edges. In order to extract the edge information of the foam aluminum cross-section structure more accurately and completely, the algorithm further optimizes the edge using gradient direction consistency and non-maximum suppression. In order to verify the feasibility of the proposed algorithm, this paper uses the dataset to test the proposed algorithm and a variety of existing algorithms, and compare the results of various algorithms by quantitative analysis. The experimental results show that the proposed algorithm is feasible and effective, and its performance is better than the comparison algorithm.

The Attitude Heading Reference System (AHRS) has been widely used to provide the position and orientation of a rigid body. A low cost MEMS based inertial sensor measurement unit (IMU) is a core device in AHRS. To improve the AHRS system performance, there is a need to develop (i) stochastic IMU error models and (ii) random noise minimization techniques. In this paper, we modify the Sage-Husa Adaptive Kalman Filter (SHAKF) to incorporate time-varying noise estimator and robustifier, termed as Modified Sage-Husa Adaptive Robust Kalman Filter (MSHARKF). In the proposed algorithm, a three segment approach is used to evaluate the adaptive scale factor followed by the learning statistics. The scale factor is iteratively updated in the MSHARKF equations. In addition, angle random walk (ARW) and bias instability (BI) errors are represented by state-space models. The proposed algorithm is applied to restrain the drift error and random noise in the MEMS IMUs signals. The performance of this algorithm is analyzed using Allan variance (AV) analysis for static signals whereas the Root Mean Square Error (RMSE) values are evaluated for dynamic signals. Experimental results demonstrate the effectiveness of MSHARKF in reducing the drift and random noise in static and dynamic conditions as compared with other existing algorithms. Finally, we present sufficient conditions for convergence proof of the MSHARKF algorithm.

The majority of processes in chemical industry is nonlinear. However we often take advantage of linear approximation and analysis as the useful simplification. Nonetheless, one has to remember that the reality is often complex, nonlinear and full of unknown unknowns. One of the forgotten aspects in control engineering is connected with the symmetricity. Asymmetric properties appear, when the process or instrumentation introduces nonlinearities. Control systems are then exposed to the asymmetrical behavior and should properly react, while their performance measures have to take them into account. This paper proposes robust control performance indexes in form of the M-estimator using logistic ψ function denoted σH and α-stable distribution scale factor y. Additionally, their application procedure in industrial chemical engineering environment is proposed. The approach is illustrated with an example of the pH neutralization process.

To improve the quality of ghost image, we propose an efficient computational ghost imaging method in this article. The primary idea is to estimate the noise value of the ghost image by analyzing and eliminating the source of the noise. The innovativeness of this work lies in analyzing a new means of noise by dissecting the qualitative relationship between transmittance in different objects and speckle patterns. While using a scale factor to describe the change of transmittance at different points of the object. The simulation and experimental results prove the effectiveness and feasibility of the proposed method through two parallel experiments. Compared to other methods, the peak signal-to-noise ratio and contrasts both have significantly increased.

We consider challenges in modeling and estimating high-dimensional functional connectivity in brain networks with a large number of nodes arranged in a hierarchical and modular structure. We develop a multi-scale factor analysis (MSFA) model which partitions the massive neuroimaging time series data defined over the brain networks into a finite set of regional clusters. To achieve further dimension reduction, signals in each cluster are represented by a small number of latent factors. The correlation matrix for all nodes in the network can be approximated by lower-dimensional sub-structures derived from the cluster-specific factors. This enables a reliable and computationally efficient multi-scale analysis of both regional and global network connectivity. To estimate regional connectivity between numerous nodes (within each cluster), we apply principal components analysis (PCA) to extract the factors. Then, we estimate global connectivity (between clusters or sub-networks) based on the factors across regions using the RV-coefficient as the cross-dependence measure. Simulation results show that the proposed MSFA estimator improves accuracy of connectivity estimation in high-dimensional settings. When applied to resting-state functional magnetic resonance imaging (fMRI) data, our method identifies modular structure of resting-state networks at multiple scales, revealing interesting connectivity patterns between voxels, regions of interest (ROIs), and functional networks.

Convolutional neural networks are built upon simple but useful convolution modules. The traditional convolution has a limitation on feature extraction and object localization due to its fixed scale and geometric structure. Besides, the loss of spatial information also restricts the networks' performance and depth. To overcome these limitations, this paper proposes a novel anisotropic convolution by adding a scale factor and a shape factor into the traditional convolution. The anisotropic convolution augments the receptive fields flexibly and dynamically depending on the valid sizes of objects. In addition, the anisotropic convolution is a generalized convolution. The traditional convolution, dilated convolution and deformable convolution can be viewed as its special cases. Furthermore, in order to improve the training efficiency and avoid falling into a local optimum, this paper introduces a simplified implementation of the anisotropic convolution. The anisotropic convolution can be applied to arbitrary convolutional networks and the enhanced networks are called ACNs (anisotropic convolutional networks). Experimental results show that ACNs achieve better performance than many state-of-the-art methods and the baseline networks in tasks of image classification and object localization, especially in classification task of tiny images.

In this paper, we constructed a new rational fractal interpolation model with scale factors and shape parameters. The model obtains different expressions by changing the scale factor, according with the diversity of image features. On the basis of the constructed model, this paper presents a novel adaptive rational fractal magnification (ARFM) algorithm based on local fractal feature analysis. Firstly, the image is divided into different regions according to an adaptive threshold determined by a calculated fractal dimension, and the scaling factor is calculated based on the relationship between global fractal dimension (GFD) and local fractal dimension (LFD). Secondly, for different regions, different interpolation forms are selected according to regional characteristics. Thirdly, parameter optimization and sub-block selection are studied to further enhance the quality of the magnified image. The experimental results illustrate that the performance of the proposed ARFM algorithm is very competitive compared with the latest magnification algorithms.

In general, surface positioning using a global satellite navigation system (GNSS). Many satellites transmit radio signals to the surface of the earth and it was detected by receiver sensors into a function of position and time. Radio waves really bad when spreading in water. So, the underwater positioning uses acoustic wave. One type of underwater positioning is USBL. USBL is a positioning system based on measuring the distance and angle. Based on distance and angle, the position of the target in cartesian coordinates can be calculated. In practice, the effect of ship movement is one of the factors that determine the accuracy of the USBL system. Ship movements like a pitch, roll, and orientation that are not defined by the receiver could changes the position of the target in X, Y and Z coordinates. USBL calibration is performed to detect an error angle. USBL calibration is done by two methods. In USBL calibration Single Position obtained orientation correction value is 1.13 ̊ and a scale factor is 0.99025. For USBL Quadrant calibration, pitch correction values is -1.05, Roll -0.02 ̊, Orientation 6.82 ̊ and scale factor 0.9934 are obtained. The quadrant calibration results deccrease the level of error position to 0.276 - 0.289m at a depth of 89m and 0.432m - 0.644m at a depth of 76m

Significant research has been conducted into image edge detection, usually focusing on gradient and higher order derivative approaches. Recent development in the field suggests the incorporation of scale factor in filter design; this is to increase robustness against noise and to facilitate the process of generating multilevel edge maps. While multifractal analysis technique uses scale factors, few researchers examine the effectiveness of this technique in image edge map generation due to its poor efficiency. Herein, we propose F-MED, a fast multifractal edge detector for grayscale images based on generating an edge map from different low frequency versions of the original image by means of frequency domain Gaussian low pass filters. The quality of results is evaluated subjectively via visual assessment and quantitatively using a set of full-reference and non-reference measures. Results show that the proposed F-MED can be leveraged to support not only high efficiency, but also quality improvement and high robustness against blurring effect and Rayleigh noise.

To solve the problem that the ship’s strapdown inertial navigation system (SINS) alignment accuracy decreases with the increase of the nonlinear filtering state dimension under mooring conditions, a method based on Kalman filter (KF) and Adaptive scale mini-skewness single line sampling Unscented Kalman Filter (ASMUKF) hybrid filtering algorithm is proposed in this paper. Three improvements are made as the following: (1) adopt a new sampling strategy. To obtain the ASMUKF filtering algorithm, scale mini-skewness single line sampling is used to replaced the traditional symmetrical sampling method and an adaptive scale factor is adapted into the Unscented Kalman Filter (UKF) to correct the real-time transformation sampling process, (2) the improved ASMUKF algorithm is combined with KF to form KF-ASMUKF hybrid filtering model, (3) the hybrid filtering model is divided into linear and nonlinear parts. The linear filtering part adopts the KF filtering model and the nonlinear filtering part adopts the ASMUKF model. Then, the calculation steps of the hybrid filtering algorithm is designed in this paper. The simulation and experimental results show that the hybrid filtering algorithm proposed has certain advantages over the traditional algorithm, and it can reduce the ship’s SINS calculation amount and improve alignment accuracy under mooring conditions.

Recent years deep convolutional neural networks(CNNs) have got great success in the single image superresolution(SISR). However, existing CNN-based SISR methods are hard to achieve ideal performance due to the limited information contained in a single low resolution (LR) image. Moreover, when the scale factor is large, SISR methods become difficult to learn and reconstruct unknown information, giving rise to poor performance. To address these issues, we propose a deep residual learning super-resolution framework MFSRResNet using multi-frame LR images as input. Our method MFSRResNet is based on the SRResNet architecture. The main modification is the number of input frames and number of convolutional layer feature maps. We use five-frame LR images as input rather than a single-frame LR image. We create multi-frame LR images by randomly downsampling a HR image and make sure sub-pixel shifts among them. The multi-frame input method increases the amount of information obtained at the input end, thus substantially improves the reconstruction results. Experiments show that MFSRResNet can well integrate the information between different LR images, and get better reconstruction results. MFSRResNet demonstrate the state-of-the-art performances on all benchmark datasets in terms of Peak signal-to-noise ratio (PSNR) and Structural similarity (SSIM). The significant performance improvement in PSNR/SSIM of MFSRResNet is 2.67dB/0.0495(×3),2.27dB/0.05498(×4) and 1.56dB/0.0504(×8) in average on two benchmark datasets Set5 and Set14 respectively compared with current state-of-the-art SISR methods RCAN.