Abstract:Taking the dynamic wind speed measurement based on the buoy platform as the background, according to the principle of motion attitude compensation and actual needs, a dual-antenna GNSS/SINS integrated attitude measurement system is designed. Kalman filtering is used to achieve information fusion. The heading angle information output by dual-antenna GNSS is introduced into the measurement, which solves the problem of weak observability of heading angle of the inertial system, and the effectiveness of the system is verified by experiments. The experiments results show that the motion compensation scheme can improve the accuracy and stability of the attitude measurement, accurately measure the attitude information of the platform, and effectively reduce the influence of the platform motion on the wind speed measurement, and the wind speed measurement error is less than 4%. It meets the needs of marine buoy observation.

Abstract:A fixed-tuned D-band third-harmonic mixer is presented with broad bandwidth and low conversion loss (CL) based on a compact transition-match structure (CTMS). The CTMS is composed of a short suspended stripline (SSLIN) on the top side and a triangle-shaped transition on the bottom side, located in the radio frequency (RF) waveguide. The mixing diode with two Schottky junctions in series is located directly on the short SSLIN in the RF waveguide, designed to be at the maximum of the electrical field of the RF signal. The single balanced configuration is then achieved, and the RF signal is only mixed with the odd harmonics of the local oscillator (LO) signal. The triangle-shaped transition helps to minimize the discontinuity between the waveguide and the microstrip circuits. The mixing circuit is miniaturized and the mixer bandwidth is broadened benefiting from the CTMS. Dimension of the circuit substrate is 5.1 mm×6.9 mm×0.127 mm. The measured results indicate that the single-side band (SSB) CL of the mixer is typically 13.0 dB, and ranges from 10.8 dB to 15.7 dB over 135-165 GHz. The instantaneous intermediate frequency (IF) is available up to 15 GHz. Keywords: terahertz (THz); wideband; third-harmonic mixer

Abstract:Conical coil has the comprehensive characteristics of planar spiral coil and cylindrical spiral coil and can be used in magnetic coupling mechanism of wireless power transmission system. In this paper, a model is established based on circuit theory, and the relationship between the magnetic coupling mechanism parameters and system output parameters is deduced. Then, Maxwell software was used to analyze and compare the traditional planar spiral coil and cylindrical spiral coil from the perspective of magnetic induction intensity and mutual inductance, and two kinds of conical coils with different turns spacing were proposed. Ferrite plate and ferrite strip were used for the two kinds of coils respectively. Finally, an experimental prototype of the wireless power transmission system is built. Experimental and simulation results show that compared with the cylindrical spiral coil, when the conical coil is used as the transmitting coil, the power transmission efficiency of the receiving coil is up to 85.5% when the offset distance is 0 ~ 75mm, and the power transmission efficiency is up to 4.9%. When the offset distance is 0 ~ 50mm, the output power is also high, up to 264W, which increases by 15.3%.

Abstract:The video surveillance system collects video data in real time, which can serve as an effective third-party witness and provide favorable clues and information for case detection. However, due to the huge amount of data and the high retrieval workload, it brings inconvenience to the case collection. Aiming at this problem, this paper aims to retrieve the target person in the video evidence, and improves the existing cross-camera person weight recognition method to realize the rapid re-identification of the target person. Firstly, the target person image is segmented to obtain the block feature map; secondly, a local fusion module is introduced to fully retain local feature information and local correlation information; then a global fusion module is introduced to fully characterize the global image features while removing background noise; Finally, the cross-entropy loss and triplet loss function are integrated to accelerate the model convergence and effectively prevent over-fitting. The simulation experiment results show that compared with the existing methods of human weight recognition, the method in this paper has higher accuracy; the application software results show that the method in this paper can quickly locate the target person across cameras and meet the fast retrieval requirements of the target person in the video evidence.

Abstract:In order to ensure the safe operation of industrial machinery and equipment, avoid serious equipment damage caused by bearing damage, and realize the fault diagnosis system of rolling bearings under different working conditions on mechanical equipment, a fault diagnosis system of rolling bearings under different working conditions was designed based on convolutional neural network. The system gram matrix method is used to convert one dimensional time series data for the characteristics of two-dimensional figure, convolution neural network training to maximize learning characteristic information in the data, the trained model deployed in PC written in LabVIEW real-time fault diagnosis, the method of case western reserve university in the United States bearing experiment data center data sets, experimental verification: On the bearing data set of Case Western Reserve University in the United States, the identification accuracy of the method is 99.85% and the comprehensive identification accuracy is 91.67% under different operating conditions. Experimental results show that the method achieves relatively accurate identification effect and has good generalization ability. It has accumulated application experience for fault diagnosis of rolling bearing under variable working conditions.

Abstract:To tackle problem of insufficient capability of autonomous inspection employing UAV in electricity pipe gallery without prior map, this paper proposes a UAV inspection method which is independent of prior map. In the first place, the UAV is equipped with Lidar and builds the map in real time by SLAM algorithm. The UAV searches for frontiers by RRT algorithm based on global grid map. Secondly, traditional RRT algorithm is difficult to judge integrity of map and guarantee completeness of inspection. To improve coverage, a frontier detection method based on digital image processing is proposed. Canny operator is used to detect frontiers in the map which is built in real time and evaluate completeness of the map. Finally, in order to verify effectiveness of the proposed method, simulation experiments are conducted in the model of a 220kV electricity pipe gallery which is located in Wuxi, Jiangsu province. RRT algorithm and the proposed algorithm in this paper are used to carry out autonomous inspection test respectively. The results show that the proposed algorithm can improve inspection coverage by 21.8% compared with RRT algorithm.

Abstract:Aiming at the quality of wireless communication in EMUs, a dual-frequency monopole antenna is designed based on microstrip line and branch method to meet the vehicle-mounted requirements of high-speed train. The size of the antenna is 23mm×36mm×1.3mm, the maximum gain is 3.8dBi, its working frequency band is 2.28~2.64GHz and 5.47~6.15GHz, the relative bandwidth in the 2.4GHz and 5.8GHz frequency bands is 14.6% and 11.7%, covering the ISM frequency band . Taking the internal environment of the CRH380A EMUs as the research object, using the CST electromagnetic simulation software, the simulation model of the vehicle-mounted antenna layout inside the carriage is established. At the same time, compared the transmission effect in the train with the free space. The antenna transmission coefficient in each BSS is between -20 and -50 dB in the ISM frequency band, and the RSRP belongs to the optimal level I. The effectiveness of the designed antenna and the reliability of its layout are verified. The research results have a good reference value for the design of the on-board antenna of the EMUs and the layout of the in-vehicle antenna.

Abstract:To deal with the insulation fault caused by debris produced by amorphous alloy core, firstly, the amorphous alloy transformer 3-D model was established, then the deformation of the core was calculated ,the results demonstrated that bigger load current caused bigger vibration displacement. The debris is produced when vibration displacement exceeds the ultimate strength of amorphous alloy. Particularly, when the short circuit occurs, lots of debris was produced due to significantly vibration displacement. Then electric field distribution when debris exists was calculated, the impact of different size, quantity and the angle between debris and windings on electric field distribution were also obtained, And it is concluded that the debris generated by the iron core vibration has an influence on the insulation.

Abstract:In response to the slow coding speed and high resource overhead of the Context-based adaptive binary arithmetic coding under the H.266/VVC video coding standard, the reconfigurable oriented architecture optimized the coding architecture based on the intrinsic parallelism of the algorithm, and designs and implements a parallel mapping method for the CABAC encoder in conventional coding mode based on a dynamically reconfigurable array processor. The array structure is able to dynamically reconfigure the optimized algorithm according to the coding input, and the software reconfiguration method is used to implement the entropy coding process without the high resource overhead of dedicated hardware encoders. Simulation results show that the mapped encoding process completes five binary sequences per encoding cycle with an average encoding efficiency of 384.13Mbin/s. FPGA based test results show that the software reconstruction approach reduces the resource overhead and improves the encoding efficiency by 5.47% compared to dedicated hardware implementations, and improves the encoding efficiency by 7.03% compared to similar reconfigurable video encoding structures.

Abstract:Cross-linked drape control system is a nonlinear, time-varying, strong coupling and multi disturbance complex control system. In order to improve the anti-interference ability and robustness of the suspension control system, this paper puts forward a suspension control strategy based on ADRC. The ADRC is designed for the speed loop, current loop and flux linkage of the caterpillar AC asynchronous motor, the designed ADRC effectively improves the suspension control accuracy of the system. Considering the existence of time delay, the control performance of the system is reduced to a certain extent, the Smith predictor technology is introduced into the design of ADRC. An output predictive ADRC based on smith predictor is designed, which effectively reduces the influence of time delay and improves the robustness and anti-interference performance of the system. Finally, the simulation analysis is given, and the results verify the effectiveness of the designed composite controller.

Abstract:To solve the problem of high output noise and low output precision of gyroscope in micro-electromechanical systems(MEMS). Based on the adaptive filtering algorithm and the wavelet threshold algorithm, the wavelet threshold algorithm and the fuzzy theory are combined，this paper proposed to apply the Sage-Husa adaptive filtering algorithm combined with the wavelet fuzzy threshold denoising algorithm to MEMS gyroscope denoising. Firstly, the improved Sage-Husa adaptive filtering algorithm is used to preprocess, and the influence of interference data on filtering is suppressed by modifying the predicted value of the state. Then, the wavelet fuzzy threshold denoising algorithm is used to post-process the signal, so as to achieve the effect of suppressing random noise. The experimental results show that the denoising effect of the algorithm is better than the Sage-Husa adaptive filtering algorithm and the wavelet threshold algorithm, the noise variance is reduced by 78.4% and 14.6%, and the signal-to-noise ratio is increased by 43.7% and 16.3% respectively in the static experiment. The algorithm can adaptively reduce the adverse effects of outliers and maintain the original signal waveform. Compared with Sage-Husa adaptive filtering algorithm and wavelet fuzzy threshold algorithm, the noise variance is reduced by 62.7% and 31.6%, and the signal-to-noise ratio is increased by 47.8% and 10.0% respectively in dynamic experiments.

Abstract:An improved YOLOv5 weld defect detection method based on deep learning was proposed in response to improve the efficiency of automatic detection and processing of weld defects in industry. Aiming at the insufficient weld sample data, a mosaic+mixup data augmentation strategy was proposed. At the same time, in order to reduce the amount of calculation and parameters of network, a lightweight GhostNet network was introduced to replace the residual module in CSP1 module in YOLOv5 backbone network, and CIOU_Loss was used as the coordinate position loss to improve the convergence rate and accuracy of the algorithm. Finally, the testing set was used for weld defect detection. The improved YOLOv5 has a mean average precision (mAP) of 96.88%, and the detection time of a single image is no more than 50 milliseconds, which is better than the traditional machine learning algorithms, could meet the real-time detection requirements of weld defects in practical engineering.

Abstract:It is a simple and effective way to determine the key factors and trace the causes of pollution by analyzing and predicting the influencing factors of air quality. Aiming at the prediction accuracy of current air quality prediction methods is not high, and it is easy to fall into the local optimal value problem, a novel model is proposed, which is based on Cross Recurrence Quantification Analysis (CRQA) and Deep Belief Network-Extreme Learning Machine (DBN-ELM) air quality data prediction method. Firstly, CRQA is used to analyze the correlation degree among various factors affecting air quality and screen out the critical factors affecting air quality. Then, the main influencing factors of air quality obtained are input into the DBN-ELM model for prediction. Concretely, DBN is used to extract key features of main air quality factors, and ELM is used for nonlinear approximation of final air quality time series data. The experimental results show that in the air quality data set of the Beijing Olympic Sports Center, the RMSE value and R2 value of this model are 1.7759 and 0.9833 respectively, which are better than other models. Furthermore, the effectiveness of the proposed model is verified by scatterplot and quantile-quantile plot.

Abstract:Aiming at the problem of local minimum and unreachable target when using artificial potential field method for mobile robot path planning in the presence of complex obstacles. In this paper, an improved artificial potential field method based on concave obstacle patching is proposed for local path planning. Firstly, the concave obstacles are filled to prevent the robot from entering the local minimum area. Then, by adding the distance influence factor, the repulsion field function is improved to make the target point the smallest point in the global situation field, and prevent the robot from falling into the target unreachable area. Finally, the simulation results show that the improved artificial potential field method proposed in this paper can solve the local minimum problem and the target unreachable problem in the environment with complex obstacles. Compared with other algorithms, it can effectively reduce the path compensation and improve the planning efficiency.

Abstract:Due to the long-term operation of centrifugal pump in harsh environment, affected by field working conditions, medium corrosion and other factors, many fault signals represent obvious nonlinearity and time-varying nonstationarity, large amount of data, and it is difficult to predict the operation state in real time and accurately. In this paper, a centrifugal pump state prediction method based on PSO（Particle Swarm Optimization） optimized LS-SVM（Least Squares Support Vector Machines） was proposed. Firstly, the kernel parameter g and penalty factor C of least squares support vector machine are quickly and automatically optimized by using the global search characteristics of particle swarm optimization algorithm. Secondly, the average absolute error, average relative error and root mean square error are determined as the prediction accuracy evaluation indexes. Finally, the prediction method in this paper was verified by the real-time collected data. The results show that compared with LS-SVM prediction model, PSO optimized LS-SVM model reduces the computational complexity, has the advantages of strong generalization ability and high prediction accuracy, and the average absolute error, average relative error and root mean square error are reduced by 52%, 56% and 44% respectively. This method can provide a theoretical basis for predictive maintenance and has a good application prospect in engineering practice.

Abstract:Aiming at the problems of the traditional adaptive filtering algorithm, such as poor noise reduction performance, slow convergence speed and insufficient ability to deal with mutation, this paper proposes an improved equation error algorithm and mirror image optimization algorithm. Among them, the improved equation error algorithm based on FURLMS algorithm carries out offline quadratic path modeling, which solves the problems of noise reduction performance and convergence speed. In order to improve the ability of the system to deal with mutation, the algorithm is optimized based on the FURLMS algorithm. The results show that when the system frequency is about 250 Hz, the mean square error of the proposed two algorithms can be stabilized at -20 db, and the improved equation error algorithm and the image modification algorithm have the noise attenuation effect of 28dBA and 30dBA respectively.

Abstract:in view of the large-scale and complex field measurement environment, it is difficult and time-consuming to set up field measurement and control points at multiple points. In this paper, the optimization of POS data accuracy by differential method is the research base point, and two-level innovation is carried out based on the post difference principle. Firstly, based on the influence of external orientation elements on the accuracy of POS data, the real POS data is obtained by using camera delay measurement and control technology, and the double difference equation of code pseudo range and carrier phase is fused. The post differential GNSS module based on ubloxn-m8t chip is constructed to eliminate the error of POS data. At the same time, the bundle method area network adjustment model of simultaneous GPS and IMU observation equation is used Through the evaluation of imaging accuracy, it is concluded that the post difference can significantly improve the accuracy of POS data, and the utility is more obvious under few control points and no structure routes.

Abstract:Electrocardiography (ECG) feature parameter extraction technology is one of the research hotspots in the field of human body signals intelligent detection. This paper systematically reviews the common automatic detection and extraction methods of ECG feature parameters, including differential threshold methods, template matching methods, wavelet transform methods, and neural network methods, and explains the mechanisms, characteristics and main application research directions of various methods, and analyzes the advantages and disadvantages of each method in different application scenarios. The neural network feature extraction method has high accuracy and good robustness, and it is the research trend and hot spot of ECG feature parameter extraction. In subsequent stages, deep learning and self-learning of neural network can be combined with differential threshold, template matching, wavelet transformation and other feature extraction methods to achieve higher requirements for complex ECG feature parameter detection.

Abstract:With the rapid development of the 5th generation wireless communication technology, the broadcast features of wireless communication bring convenience to life, but also make the communication process has the risk of being intercepted and eavesdropped by illegal users. In order to protect the information transmission process, low probability of detection communication gradually attracts the attention of academia and industry. Different from the traditional fixed jammer assisted low probability of detection communication, mobile jammer assisted can greatly improve the scene adaptability of low probability of detection communication. This paper establishes a low probability of detection communication model based on vehicle-mounted mobile jammer, and uses beamforming technology to analyze the performance of low probability of detection communication in Rayleigh fading channel. Firstly, the error probability of test side is analyzed by hypothesis test and the optimal detection threshold is obtained. Then the average concealment probability of the model is solved according to the optimal detection threshold and the expression of interrupt probability is deduced. Finally, the optimization algorithm is designed with covert throughput as the optimization objective. Simulation results show that compared with fixed jammer, the performance of low probability of detection communication is improved by using vehicle-mounted mobile jammer.

Abstract:Circular array targets are widely used in 3D measurement and camera calibration. Under the perspective projection of the camera, the circular marker on the target will degenerate into an ellipse. However, the geometric center of the ellipse is not the central projection of the circular marker on the image plane. Therefore, a fast image point extraction algorithm based on projective geometry is proposed by using the geometric invariance of collinear and common points under projective transformation. Firstly, the general equation is obtained by fitting the ellipse after edge detection, shape filtering, and sub-pixel edge extraction. Secondly, the coordinates of two vanishing points along the coordinate axis of the plane target are got by projective transformation matrix. Finally, using the principle of perspective invariance, the vanishing points and the ellipse general equation are used to solve the two sets of common tangent point coordinates of the ellipse image. The intersection of two sets of tangent lines is the center projection of the circular marker on the image plane. The experimental results show that compared with the traditional method of searching common tangent points, the reconstructed average distance of the extracted center image points is reduced by 32.34%, and the precise location of the circular markers is realized, and the solving process is greatly simplified, which has strong practicability.

Abstract:In order to address the leakage detection problem of traditional water supply pipeline, in this paper, we propose a pipeline leakage detection algorithm based on the sparse and lightweight convolutional neural network technology. First, the sound signal leaked from the pipeline is collected by the sound sensors. After preprocessing operations such as stereo conversion, resampling, and length alignment, it is converted to a mel spectrogram. Then, a sparse and lightweight convolutional neural network model is proposed to perform feature extraction and leak detection on the mel spectrogram. Due to the sparse and time-delayed characteristics of sound feature images, we introduce the Inception structure to improve the feature extraction ability. In addition, to deploy the proposed model to the edge side, a lightweight convolutional neural network based on SqueezeNet is designed to reduce model parameters and thus reduce the model complexity. Massive experimental results show that the proposed pipeline leakage detection algorithm has less computation complexity and better recognition accuracy.

Abstract:Aiming at the problem that pharmacists make mistakes due to fatigue in the process of pill sorting, a pill detection algorithm based on improved EfficientDet is proposed in this paper. Firstly, mosaic data enhancement technology is introduced to improve the complexity of sampling data; Then, the backbone network EfficientNet is improved and optimized, and the feature fusion layer of CBMA module is embedded to improve the extraction ability of key features of pills by enhancing learning features; Finally, a cross level data stream from the lower layer to the upper layer is added to the feature fusion part of BiFPN. By making full use of multi-level data, the efficiency of multi-scale feature fusion at different levels is improved. Experiments show that the improved EfficientDet algorithm has a map value of 99.84% in the test, which is 0.65% higher than the original EfficientDet algorithm. At the same time, it also has higher accuracy and better practical application than the target detection networks with better performance such as YOLOv3, YOLOv4 and YOLOv4-Tiny.

Abstract:The edge of electroluminescence image is fuzzy and the texture is not clear, which makes it difficult to quantitatively evaluate the defects of crystalline silicon photovoltaic cell. In order to solve this problem, a defect detection method based on electroluminescence polarization image fusion is proposed. First, based on the analysis of the crystalline silicon photovoltaic cell structure, the electroluminescence polarization imaging mechanism was introduced. Then, the Laplacian pyramid was used to decompose the obtained infrared intensity images and polarization images, and guide filter was used to enhance the high-frequency components. The high and low frequency parts were fused by the rule of regional energy maximum and regional energy weighted average. Finally, a short wave infrared polarization detection platform was established for photovoltaic cell inspection. The results show that polarization imaging can highlight the contour edges and texture detail of the photovoltaic cell defect image. The photovoltaic cell defect features in the fusion image are more prominent. The objective evaluation index such as information entropy and standard deviation are significantly improved, which verifies the effectiveness of the proposed method.

Abstract:In order to reduce the difficulty of adjusting the flight control algorithm of unmanned aerial vehicles, ground test equipment is needed to lift and limit its movement. Therefore, a 4-DOF motion platform driven by ropes and a leading screw is proposed. The vertical linear motion of the platform is provided by the motor-driven lead screw, and the motions in three rotation directions are driven by four motors pulling the ropes. In order to realize the position control of the platform, the dynamic model of the platform is established. Due to the different dynamic characteristics of the platform in each direction, the attitude compensation method realized by the fuzzy controller is used to improve the tracking performance. The simulation results show that in each movement direction, the settling time is less than 0.2 seconds and the position tracking error is zero. So, the system has the advantages of fast response, strong anti-interference ability, and high positioning accuracy. It can assist in the adjustment of a copter’s control system and has great application significance.

Abstract:In order to improve the real-time performance of 3-channel polarization images processing, a parallel processing method for 3-channel polarization images registration is designed and implemented on the multi-core DSP hardware platform TMS320C6670 based on Fourier-Mellin transform. According to the different data space and correlation degree in images registration, operations of FMT algorithm are optimized as multiple operation tasks with high cohesion and low coupling, and a parallel computing structure based on data stream is designed, which overcomes the problems of low efficiency and high data coupling in the original FMT algorithm. According to the processing characteristics of multiple cores and multiple coprocessors of DSP, the parallel task allocation methods such as grouping isomerism and intra-group isomerism are designed. While load balance of the system is ensured, the multi-parallel computing tasks are allocated accurately and efficiently. The experimental results show that the average registration time is about 16.2 ms under the condition that the registration accuracy is less than 1 Pixel and the rotation error is less than 0.2 degree, which can meet the real-time application requirements of polarization imaging detection.

Abstract:In order to meet the demand of high precision image measurement of large object, a method of measuring large size target with wide field of vision by using fish eye camera image distortion is proposed. According to the polynomial approximation function of nonlinear distortion, the camera imaging model is established, and the camera parameters are determined through the stereo calibration plate; Move the camera along the optical axis to collect two images, and establish a measurement model to measure the size of the target to be measured. Through the measurement experiments of checkerboard and the geometric dimensions of a large building, it is shown that the method of directly using the distorted image proposed in this paper is superior to the traditional method of correcting the distorted image into a linear image, and has higher measurement accuracy and stability.

Abstract:For UAV aerial photography, the background is complex, the detection target is small and dense. A lightweight UAV aerial photography target detection algorithm SDS-YOLO based on YOLOv5 is proposed. Firstly, SDS-YOLO algorithm reconstructs the lightweight network structure, the feature extraction network and feature fusion network are reconstructed. Adjusts the detection layer and receptive field architecture, establishes the multi-scale detection information dependence between deep semantics and shallow semantics, increases the weight of shallow network feature layer, and improves the detection ability of small targets; Secondly, the pre selection box is adjusted by clustering and genetic learning algorithm to realize the optimal pre selection box selection mechanism of reconstructed network and accelerate the convergence speed of the model. Finally, SDS-YOLO was trained with varifocal loss to make IACS regression to improve the detection ability of the model to dense objects. The results show that the accuracy of the model is improved by 7.64%; The volume of the model is 4.25MB, which is significantly lower than that of the original model; The speed and amount of reasoning are improved. Compared with the current mainstream algorithms, SDS-YOLO has made good improvements in all aspects to meet the requirements of real-time target detection in UAV aerial photography.

Abstract:No magnetization confirmation is required before the magnetic resonance imaging. In view of the complex background magnetic field and large interference in the environment where the magnetic resonance imaging equipment is located, the system has problems such as high false alarm, detection failure, a ferromagnetic detection system based on fluxgate sensor is designed. The system uses fluxgate sensor and second harmonic method to measure magnetic field signal, and the analog circuit is designed for phase sensitive detection. The closed-loop system improves the linearity and anti-interference through integral feedback to make the probe work in a stable state. Kalman filter is used to process the background magnetic field and magnetic anomaly signal to improve the signal-to-noise ratio, and the energy detector is designed to detect the magnetic anomaly signal. The test results show that the system can reduce the false alarm when the magnetic resonance imaging equipment is working in the complex magnetic field environment outside the magnetic resonance imaging room, and effectively detect ferromagnetic substances.