Abstract:In order to solve the problem that weak signal is difficult to detect in very low frequency and ultra-low frequency communication, a low noise preamplifier for low frequency is designed. Firstly, the noise model of amplifier is established and the noise sources and noise matching are analyzed, then the design points of the low noise amplifier circuit are described in detail, and a multi-layer shielding design method is proposed. Through simulation and experimental measurement, it is concluded that at 1kHz~1MHz, the voltage gain can reach 38dB; at 1kHz, the equivalent input noise is 0.99nV/√HZ, the noise coefficient reaches 0.76dB. The noise performance is greatly improved, it can meet the performance design requirements of low noise amplifier and can realize the effective reception of weak signal in VLF and VLF long-distance communication.

Abstract:In modern industrial production, the quality requirements for product parts are very strict. Often the performance of the product cannot be exerted due to the poor airtightness of a part. Therefore, high requirements are put forward for the airtightness test of the product. . Aiming at this problem, this paper designs a set of bubble detection system based on OpenCV, which can be used to achieve air tightness detection. In this paper, the relatively mature technology of photography is used as the basis to capture bubble images in real time. First, Gaussian filtering is used to remove the noise in the bubble image, and the binarization process is performed, and then the Canny edge detection algorithm based on the big law is used. The edge contour of the bubble is detected, and finally the bubble contour is filled by the image filling algorithm, and the volume of the bubble is calculated by the force analysis method. In order to verify the effectiveness of the system, this paper made a simple experimental device to determine whether the bubble detection system is effective. The experimental results show that the system based on the algorithm has high real-time performance, fast calculation speed, high accuracy and can reach more than 94% effectiveness.

Abstract:Aiming at the problem of balance stability caused by uncertain disturbance and imprecise model, a cascade fuzzy adaptive PID controller is designed. The controller uses the flywheel inertial system, space state equation and motion control strategy for mathematical modeling, relies on the model predictive control technology to obtain the current attitude information of the unmanned bicycle, combines the fuzzy adaptive PID algorithm and linear quadratic control method to design the system controller, and uses the reaction inertia moment generated by the inertia flywheel when changing the acceleration, Realize the static balance control of the unmanned bicycle, input the handlebar angle and rear wheel speed to realize the motion control of the unmanned bicycle. The results of simulation test and prototype experiment prove the rationality and effectiveness of the upright control and motion control of the unmanned bicycle, which makes the system have the ability of anti-interference and strong robustness.

Abstract:Aiming at the problem of phase-locked loop (PLL) which can accurately obtain the voltage phase information of power network, when the frequency changed suddenly, the large phase deviation affected the phase tracking performance, in this paper the second-order generalized integrator (SOGI) was improved by introducing a proportional component, and proposed an improved digital phase detector structure based on the improved SOGI and Park Transform, and explained the working principle of PARK transform in detail, the control structure of the improved PLL was analyzed, then it is modeled and simulated in Simulink and realized digitally with FPGA. The simulation results show that the PLL composed of the improved Phase detector has better phase-locked performance, high phase-locked precision and fast time response compared with the traditional PLL.

Abstract:With the rapid development of computer vision, pedestrian detection technology has become one of the hot topics. SSD algorithm in pedestrian detection algorithm is more stable than other pedestrian detection algorithms. Aiming at the problem of low detection accuracy when SSD algorithm is used in pedestrian detection, this paper proposes a pedestrian detection algorithm which uses RESNET residual network as the feed-forward network of SSD to extract pedestrian features and generate the target box through the eigenvalues. The original SSD algorithm and the improved SSD algorithm are compared through the training set and the test set of INRIA pedestrian detection data set. The experimental results show that, compared with SSD algorithm, the improved pedestrian detection algorithm is more robust, and the average accuracy of the algorithm is improved by nearly 5%.

Abstract:To overcome the problem of long convergence time and severe shaking occurred in the path tracking control of a mobile robot adopting traditional sliding mode control approach, a path tracking method is developed for guiding a mobile robot by employing the backstepping sliding model control technique and double power reaching law, a remote control platform for a mobile robot is build based on Kinect. First, pyramid and threshold segmentation algorithms are used to preprocess the original images which containing the desired path are captured by the Kinect. Then, skeleton pixels of the desired path are extracted from segmented images by Zhang fast parallel thinning algorithm and further converted into the physical coordinate frame. Next, the linear regression algorithm is adopted to identify path parameters by fitting transformed physical coordinates. Finally, backstepping sliding mode control and double power reaching law are synthesized to design the proposed path following controller capable of guaranteeing a fast cross-track error minimization and convergence rate. Through simulation experiment, the distance deviation and angle deviation of the proposed can converged in 2.3 seconds and 2.5 seconds respectively, and have better steady-state performance. Based on the real scene, the iRobot Create mobile robot path tracking remote control experiment results show that the path tracking error is only 0.0127 meters. Simulation and experimental results verify that the proposed control method and remote control platform can weaken the shaking and increase the speed of convergence.

Abstract:In order to meet the requirements of clinical delivery and solve the shortcomings of traditional forceps and manual aspirator, a closed-loop PID controlled electronically-controlled vacuum extractor was developed and designed. The ATmega128L microcomputer processes the negative pressure information fed back by the negative pressure sensor through the PID control algorithm processing negative pressure sensor feedback negative pressure information to output signals to control the frequency converter, and then controls the vacuum extractor to carry out accurate control of the negative pressure. At the same time, the system also has real-time screen negative pressure information display, emergency one-button release of negative pressure, negative pressure abnormal sound-light alarm and other functions, which enables medical staff to operate more confidently and efficiently. The test compared start-stop control that the commonly used in the market with the PID control algorithm which used in this system. Result showed that the PID control algorithm was more accurate and efficient; In the repeatability experiment, the target values of -40kPa, -50kPa and -60kPa were respectively tested, and the mean values of the test were -40.21±0.66kPa, -49.67±0.87kPa and -60.57±1.24kPa. Results showed that the system had a good stability and could meet the needs of clinical midwifes.

Abstract:In order to solve the problem that the traditional fixed time to collision (TTC) threshold early warning model is prone to early warning under higher speed conditions, this paper proposes a forward collision warning system that considers the speed of the vehicle. On the basis of the traditional collision time warning model, the speed factor is introduced to optimize the model to form a dynamic TTC warning model; use PreScan/Simulink co-simulation data to curve the speed factor; finally, for the emergency braking condition of the preceding vehicle, use The PreScan/Simulink software performs comparative simulation experiments on the traditional early warning model and the dynamic TTC early warning model. The results show that the early warning success rate of the fixed TTC threshold early warning model is 80.28%, and the success rate of the dynamic TTC early warning model is 96.94%, which verifies the effectiveness of the method.

Abstract:The vibration signal of the main shaft gearbox is greatly affected by external factors, which leads to a low fault diagnosis accuracy rate. This paper proposes to diagnose the fault of the spindle box based on the speed signal collected by the virtual prototype. Firstly, the virtual prototype of main gearbox is built, and the contact parameters of gear pairs were set based on Hertz contact theory, and the accuracy of the model simulation is verified by comparing the theoretical speed. Secondly, by simulating the common fault types of a variety of gearboxes, the trend of the output speed of each shaft under different working conditions is analyzed. Finally, a KNN fault diagnosis model is established based on simulation samples, and the Z-Score method is improved. The results show that the diagnosis accuracy of the improved fault model is significantly improved, which realizes the correct diagnosis of common faults of the gearbox.

Abstract:Due to the limitation of pixel accuracy, there are a large number of sawtooth contour segments in the spectacle lens contour data extracted from the image, which seriously affects the smoothness and accuracy of the contour data. In order to obtain a smooth contour curve, this paper performs segmental fitting processing on the original contour data: firstly, the curvature of each point on the contour is estimated according to the curve direction before and after the target point, and then the contour segments with roughly the same curvature on the contour are calculated according to the least squares. The principle is to fit a circle, and finally the remaining contour segments are conformally fitted based on the Bézier curve. The experimental results show that compared with the existing smoothing algorithm, the segmented fitting algorithm described in this paper has higher accuracy while obtaining a smooth contour curve. It is proved that the segmented fitting algorithm can be used to smooth the contour curve of the spectacle lens.

Abstract:Aiming at the problem that the single parameter representation in the transformer fault diagnosis method is not comprehensive enough, a dry-type transformer fault diagnosis model based on the mixed characteristics of MGCC is proposed. First, the preprocessed dry-type transformer noise signal passes through the Mel filter and the Gammatone filter to obtain the MFCC with general anti-noise performance and the more robust GFCC characteristic parameters; then, the two parameters are linearly superimposed and the Fisher is used to compare with discarding the components with lower contribution rate, the mixed parameter MGCC is obtained; finally, it is sent to the LSTM classification model for pattern recognition. Calculation results show that the fault diagnosis rate of the proposed mixed feature MGCC is as high as 96.11%, which has better accuracy and noise immunity than a single cepstrum feature parameter of the acoustic signal.

Abstract:In development and application of wind energy, for improving the prediction accuracy of wind power output, a prediction model based on improved optimal foraging algorithm for support vector machine（IOFA-SVM） is proposed. Cauchy variation and differential mutation strategy are added into the traditional optimal foraging algorithm to improve the global optimization ability to obtain the optimal parameters of SVM. Using the improved IOFA-SVM model to predict wind power output and comparing the results with BP, GWO-SVM and OFA-SVM models, the three evaluation indexes MAE, NMAE and NRMSE of the model decreased by 0.59%, 0.53% and 0.50% respectively, which shows that the IOFA-SVM model does improve the precision and accuracy of wind power output prediction, and is important to dispatch electric energy and maintain power grid stability.

Abstract:Mine hoist is the throat of mine production, and its reliability plays an important role in safe and efficient production of coal mine. Hoist fault prediction is an important measure to improve the reliability of hoist. Aiming at the problem of large relative error of traditional hoist fault prediction model, this paper establishes a hoist fault prediction model based on least squares support vector machine. After training the model with the collected data of the hoist, the real-time data of the hoist are predicted to find out whether it is possible to be abnormal. The experimental results show that the average absolute percentage error and mean square error of the model are 1.1301% and 1.1663 respectively, which are lower than the wavelet neural network prediction model, and it has shorter running time and faster convergence speed. The prediction method can accurately find the fault of mine hoist in advance and provide an important basis for the predictive maintenance of mine hoist.

Abstract:In order to reduce the reliability of data transmission through long cable in aerospace test, a high reliability transmission system based on FPGA is designed. Transmission system hardware adopts Chengdu vibration core domestic chips GM8223 and GM8224, with the use of drivers and equalizers to increase and enhance the driving ability of the circuit; Preweight design is added to enhance the high frequency component of the signal and compensate attenuation. The software of the transmission system explains the high reliability transmission in terms of the initialization of the sender, LVDS transceiver logic flow, and CRC+ 8B / 10B dual coding. Initialization can ensure that the sender sends a valid number; CRC+ 8B / 10B dual encoding reduces bit error rate by sacrificing bandwidth. It has been verified by practice that zero bit error rate transmission can be achieved on the 240Mbit/s coaxial cable with the length of 120m.

Abstract:At present, most digital cameras take color images with red, green and blue components using 8-bit color depth bits. When these images are processed, analyzed and processed to extract information from the image or to satisfy the visual and psychological requirements, the processed images are prone to color slicing, especially when the image is enlarged,which is not conducive to the subjective visual effect of the human eyes.The larger the color depth, the more detailed the color change. Images with more than 8 bits of color depth have better performance in later processing. This paper expands the color depth of the image by averaging multiple images, and enhances the low illumination image after color depth expansion. The experimental results show that the color change of the high color depth image after processing is smoother and delicate, and the resulting image quality is higher. The validity of this method is verified and it has some practical value.

Abstract:Traditional network traffic classification methods are difficult to distinguish network traffic encrypted using VPN. In order to achieve classification of encrypted traffic, a network traffic image classification method based on multiple composition methods is proposed. Five special composition methods are studied to convert encrypted network traffic into traffic images, and finally convolutional neural networks are used to classify them. The experimental results on self-collected VPN encrypted traffic dataset and ISCX VPN-nonVPN public dataset show that the classification accuracy of this encrypted traffic classification method reaches more than 90% and 95% on the two datasets, respectively. The classification accuracy of diagonal or waterfall composition method has about 1% improvement over the traditional line composition method. The special composition method achieves the improvement of encrypted traffic classification accuracy by enhancing the correlation of pixel points in the traffic image and increasing the image features in the traffic image.

Abstract:When defect detection is carried out on the surface of liquid crystal screen, the horizontal and horizontal texture of the surface will interfere with the detection accuracy. In this paper, a defect detection method for liquid crystal screen is proposed. Firstly, the texture of liquid crystal screen is suppressed by the image structure extraction method based on the relative total variation model. This model is applicable to both non-uniform and anisotropic textures, and can effectively decompose the structural information and texture in liquid crystal image. The extracted structural information is enhanced by high-boost filter while keeping the characteristics of low-frequency component, and the saliency detection of defects in the image is carried out. The experimental results show that this method can effectively suppress the background texture information and retain the defect information completely, which can realize the saliency detection of LCD screen defects and improve the accuracy of LCD image defect detection.

Abstract:Image based denoising algorithm is the basis of building intelligent video surveillance system, which is of great significance to capture things accurately. Considering that the hard threshold of BM3D denoising algorithm cannot adapt to the noise intensity and lacks the protection of image edge texture information, an improved BM3D image denoising algorithm based on adaptive filtering is proposed. Firstly, the adaptive filter is used to replace the hard threshold filter to match the similar blocks in the basic estimation stage. More accurately, soft threshold is applied to the high-noise region and total variation is applied to the low-noise region. Then in the final estimation stage, the K-means clustering method is used to find the matching blocks to obtain the final denoising image. The experimental results show that the new algorithm improves the PSNR of images by an average of 0.89dB, and the SSIM of images by an average of 1.05 times. At the same time, it avoids the edge ringing effect caused by the traditional algorithm, which is beneficial to practical application.

Abstract:In view of a series of problems such as large error and slow detection speed of traffic signs under natural scenes in China.An improved YOLOV4 algorithm is proposed.Firstly, image enhancement, image denoising and other processing are added to the input end of the algorithm. Then, the detection layer of the algorithm is modified by deleting the 19*19 detection layer and adding the 152*152 detection layer.Finally, the K-meansⅡ clustering algorithm is used to carry out clustering analysis on the reconstructed traffic sign data set, and the initial candidate box of the network is redefined.The experimental results show that the improved algorithm can detect small traffic signs accurately and in real time in the natural scene.96mAP is obtained on the traffic sign data set based on the CCTSDB data set, and the detection speed is 26FPS, which is 1.7% and 1.4 higher than YOLOV4 algorithm respectively.

Abstract:It is of great significance to study the attitude parameters of projectile target intersection for weapon damage effectiveness evaluation. The attitude parameters of projectile are usually extracted based on field experimental images. Therefore, in view of the problems such as high noise interference intensity, low imaging contrast and difficult control of the accuracy of central axis extraction of projectile in field experimental images, etc. The edge preserving and denoising algorithm based on Gaussian and adaptive median filtering, the filling and segmentation algorithm based on Canny edge detection and the central axis extraction method based on the target edge contour are designed. The results show that the influence of the external field noise and the uneven brightness of the projectile surface are significantly reduced. The edge preserving and denoising of the projectile image, target segmentation and extraction are realized, and the central axis of the projectile is detected automatically. The result is that the linear equation of the central axis on the image plane is extracted effectively.

Abstract:In urban traffic, safety accidents caused by electric bike riders often occur. Wearing a safety helmet can effectively avoid or reduce the damage caused by a safety accident. Therefore, many cities have promulgated relevant regulations on wearing a safety helmet. Aiming at the existing problem of low detection accuracy of helmet wearing, this paper proposes an algorithm for detecting the safety helmet wearing based on improved YOLOv3. The improved algorithm in this paper adopts the weighted feature fusion of channel and spatial attention modules, and combines densely connected networks to improve the effect of feature extraction, and adds a spatial pyramid pooling structure to enhance features. In this paper, the improved algorithm is tested and compared with original YOLOv3 at the self-built electric bike helmet wearing detection data set. The obtained results show that the mean average precision of the improved algorithm proposed in this paper reaches 93.29%, which is much higher than the original YOLOv3 algorithm. Experiments confirm that the proposed model can effectively enhance detection accuracy of electric bike helmet wearing detection.

Abstract:With the development of rotorcraft and the increasing complexity of the environment in many missions, single sensor has shown obvious deficiencies. Therefore, better autonomous navigation is needed to complete all kinds of urgent and dangerous tasks. Starting from the basic concepts of vision and inertia, this paper comprehensively analyzes the advantages of this kind of combination. Then, the basic situation of rotorcraft is introduced, and the key technologies of vision/inertial navigation based on rotorcraft are analyzed, including front-end for vision, initialization and fusion methods. Particularly, more outstanding algorithms in recent years are introduced, which help us understand their advantages and disadvantages better. Finally, according to the existing literature, problems that need to be further studied of vision/inertial integrated navigation for rotorcraft are summarized and analyzed.

Abstract:In order to improve the encryption performance of the image and achieve higher-performance image secure transmission and storage, based on the traditional image scrambling and diffusion algorithm, a hyper-chaotic image encryption algorithm with image bit permutation is proposed. The newly constructed four-dimensional hyperchaotic system is used to scramble the bit-decomposed image to generate an intermediate ciphertext image, and then perform diffusion to obtain the final ciphertext image. The use of bit decomposition and hyperchaotic encryption enhances the pseudo-randomness of the key stream and encrypted image. Simulation experiments show that the image encrypted by this algorithm has better distribution characteristics, the algorithm can not only resist statistical analysis and differential attacks, but also has a large key space and better initial value sensitivity.

Abstract:Aiming at the problems that the discrete components used in the traditional missile-borne telemetry communication system occupy a large proportion of space and have poor portability, a software radio zero-IF telemetry transmitter system based on ZYNQ and AD9364 is designed. The system can perform DQPSK modulation on the data collected and framed by the editor, and use the PS terminal to configure the AD9364 registers. The modulated data is transmitted through the LVDS high-speed data interface between the PL and AD9364, and the up-conversion is completed inside the AD9364 chip. After processing, filtering, etc., the data is sent to the ground receiving station using the designed flexible antenna through the power amplifier module. Finally, a telemetering transmitter test system was built, and the telemetering transmitter was tested for modulation and communication, and the system's field data transmission and reception test was realized. Tests show that the system effectively reduces the size, power consumption and weight of the missile-borne telemetry system, and relying on powerful software reconfigurability, not only simplifies the hardware design, but also improves the versatility of the equipment, and the transmission distance is also greatly increased. It meets the miniaturization and versatility requirements of the missile-borne telemetry system.

Abstract:Detection of shallow concealed targets requires that the detection system can effectively penetrate all kinds of media and image the targets at different depths. Millimeter wave is widely used in the field of concealed target detection because of its high penetrability, little damage to organisms, strong adaptability to the environment and good echo contrast of different materials in this band. The broadband millimeter wave detection system collects the reflected signal of the hidden target through FMCW millimeter wave radar array to reconstruct the electric field distribution information and obtain the target image. However, the range resolution of millimeter wave detection system is insufficient, and the reconstructed image will be affected by holographic speckle to introduce artifacts and interfere with the real target image. In order to solve this problem, this paper proposes a real target segmentation method based on fuzzy c-means thresholding algorithm (FLICM) combined with local information to reduce the interference of false targets, and compares the proposed algorithm with the traditional algorithm. The experimental results show that the proposed algorithm effectively realizes the image segmentation of hidden objects, suppresses the influence of speckle, and improves the detection accuracy of hidden objects.

Abstract:A communication and simulation method with an intermediate conversion board as a bridge is proposed to meet the real-time performance and accuracy of the combination weigher simulation in Digital Twin. At first, the control signal of the machinery was sent out by the control board and captured by the intermediate conversion board and then the signal in it was converted and sent to the upper computer . Finally, the model of combination weigher built on the Unity3D in the host computer recognized the data from the conversion board, and thus performed motion simulation of the real machine. To obtain the control signal accurately, the signal was captured by the interrupt on the conversion board, and converted as a fixed format to the upper computer through the Ethernet. To simulate the movement of the combination weigher, the Physics component in Unity3D is used to ensure accurate movement. Finally, the transmission speed in the simulation process of the combination weigher was compared and analyzed. In addition, this method can also be used to simulate high-frequency pulse-driven robots, modular machine tools and other machinery.

Abstract:According to the power demand of portable Raman spectrometer, a small volume, digital power management system is designed.This power management system takes STM32 as the main control chip, adopts TI company's battery management chip BQ3060 and BQ24005 to realize the management of lithium battery, designs the power supply switching module and switch control module to control the double power supply switching and the system switching machine, and uses the DC/DC conversion module to ensure the constant voltage output of the system.Finally, through software programming, the system can inquire the power supply status in real time and adjust the power supply strategy. After debugging the power management system with small Raman spectroscopy system as the load power supply test, the results show that the system’s power supply , small temperature change range, feedback can real-time management of power supply and power supply condition meet the design requirements.This system provides a good solution for portable Raman spectrometer power management system.

Abstract:In order to ensure the accuracy of the temperature measurement, it is necessary to perform cold junction compensation on the thermocouple circuit. At present, when using AD590 to perform cold junction compensation for thermocouple temperature measurement, there are problems such as circuit asymmetry and low safety. This design is optimized for the cold junction compensation of the S-type thermocouple. The thermocouple input circuit is separated from the AD590 cold junction compensation circuit. The output voltage of the cold junction compensation circuit is used as the bias voltage to superimpose the thermocouple signal, and the superimposed signal Perform amplification, filtering and other conditioning. The test results show that the common mode rejection ratio of the optimized circuit is increased by about 50dB, and the cold junction compensation accuracy is controlled within ±0.20FS, which can improve the symmetry and safety of the thermocouple input circuit.

Abstract:In order to solve the problems of zero position, linearity drift and field calibration difficulty caused by the environmental impact of the depth sensing signal of a small underwater vehicle, an improved sensing signal processing method integrating software and hardware with pure software is proposed. This method is mainly based on embedded processor and digital potentiometer, and uses high-order low resistance digital potentiometer to adjust the gain and zero position of operational amplifier. Combined with the corresponding algorithm, it can not only automatically realize the adaptive adjustment of the output signal of the depth sensor with large individual differences, making its linear amplification output range stable, but also has the functions of scalable digital output, real-time internal recording and convenient on-site calibration. The output signals of different batches of depth sensors can be adjusted to linear output in the range of 1V ~ 5V, so as to solve the problem of inconsistent linear interval of post-processing output caused by individual differences of different depth sensors. At the same time, the hardware zeroing method also avoids the limitation of the adjustment range caused by the conventional software zeroing method, and even greater error caused by exceeding the linear region of the sensor. The actual navigation test shows that the accuracy of the method can be improved by about 2.2m under the water depth of about 100m in an inland lake.

Abstract:Total organic carbon content (Total Organic Carbon,TOC) in water is an important index for the classification of ultrapure water grades in experiments. Whether the measurement is accurate or not is related to the accuracy of the relevant experimental results. This paper designs a high-efficiency and high-precision ultra-pure water TOC detection equipment in view of the large volume of the oxidation tank in the existing TOC detection equipment based on ultraviolet oxidation method, which leads to the low oxidation efficiency of ultra-pure water, and most of them combine with NDIR sensors to detect carbon dioxide content and the introduction of complex process, which leads to the complex working process of the equipment and reduces the measurement accuracy and detection efficiency. In this paper, the ultraviolet oxidation method and the resistance method are combined, and an ultra-small volume and precise integrated oxidation-conductance tank is designed based on the structure of the oxidation tank. Based on high performance and low power embedded micro-controller, the hardware design and software development of electronic control system are carried out. Through a series of experiments, it is proved that the design has the characteristics of no pollution, low energy consumption, high efficiency, high precision and low cost while improving the detection rate.