Abstract:Photoplethysmography (PPG) signal, which contain abundant information related to blood pressure, can be used for cuffless blood pressure measurement. However, PPG signal is easily disturbed by noise, and the accuracy of blood pressure measurement depends on high quality characteristics of the PPG signal. Therefore, we propose a method integrating modern signal processing and pulse wave characteristic parameters analysis to improve the accuracy of cuffless blood pressure measurement based on PPG signal. Firstly, the effective PPG signal is reconstructed by combining ensemble empirical mode decomposition and signal quality detection algorithm to suppress noise interference, so as to ensure the validity of waveform and frequency characteristics of PPG signal. Combined these PPG characteristics and individual parameters, the BP neural network blood pressure measurement model is established. The method, called mean impact value, is used to select the parameters to reduce redundancy, then the genetic algorithm is used to optimize the neural network. Finally, we establish the final blood pressure measurement model. The experimental results show that the systolic and diastolic blood pressure measurement errors ≤ 10mmHg are 93.1% and 94.83%, respectively, by using the proposed method. The results meet the blood pressure measurement standards and can effectively realize the cuffless blood pressure measurement.

Abstract:In order to improve the positioning accuracy and instruction transmission efficiency of BeiDou in aviation emergency rescue system, the information optimization processing technology based on BeiDou/5G is studied. The optimization algorithm model of positioning accuracy based on BeiDou satellite-based augmentation is established. The compression coding method of BeiDou spatial and temporal information transmission is studied, and an aviation emergency rescue system platform based on BeiDou/5G fusion is designed. The positioning accuracy and information transmission performance are verified and analyzed by the system. The results show that the positioning accuracy of emergency rescue terminal is about 1 meter after optimization, and the word length required for a single communication content of BeiDou short message is effectively reduced. The communication efficiency is improved, and the information reliable processing between the rescue terminal and the command center is implemented. The results have certain reference value for the information exchange of aviation emergency rescue system.

Abstract:In order to improve the control accuracy of attitude angles and enhance the anti-interference ability of unmanned helicopter, a novel ESO-LQG-PI control method for TREX 700L helicopter is proposed in this paper. By considering the high non-linearity of the helicopter system, a class of small disturbance incremental linearization processing approach is adopted. Meanwhile, an extended state observer (ESO) is used to estimate and compensate the system disturbance in real time. The proportional integral (PI) controller is added to reduce the steady-state error caused by input signals, then linear quadratic Gaussian (LQG) controller is presented to obtain the optimal control gain. Summing up the advantages of PI and LQG, this paper proposes an ESO-LQG-PI control method. The experimental results show that the hovering attitude controller of unmanned helicopter improves the control precision of the control system, the error range of attitude angle is ±0.5°. It can realize the hovering control of the helicopter well, and has anti-interference ability.

Abstract:At present, most electric vehicle charging devices have limited output power and the charging methods used are relatively simple, which causes problems such as long charging time, high energy consumption and high temperature rise in the charging process of electric vehicles. In order to solve these problems, this paper designs a high-power charging equipment and proposes a multi-stage current charging method for electric vehicles based on the optimal charging curve, and this charging method is applied to the high-power charging equipment. Firstly, the electrical structure diagram of the high-power charging equipment is designed; secondly, the implementation process of the multi-stage current charging method is given according to the charging characteristics of the power battery; then, the experimental comparison and analysis with the traditional constant current charging method and constant current and constant voltage charging method are conducted on the high-power charging equipment. The results show that the multi-stage current charging method can shorten the charging time by more than 5% compared with the traditional 1C constant current charging and 1C constant current and constant voltage charging methods, reduce the charging temperature rise by more than 50% compared with the traditional 2C constant current charging method and 2C constant current and constant voltage charging, and make the battery capacity reach 80% in a relatively short time , which can meet the requirements of fast charging of electric vehicles. Therefore, the method proposed in this paper can balance the charging time and charging temperature rise of electric vehicle power battery, prolong the service life of electric vehicle power battery, and have better overall performance.

Abstract:For offshore oil pollution cleaning, manual cleaning methods are costly and harmful, especially unable to adapt to the complex sea conditions under severe weather and the toxic oil pollution environment prone to fire. For this reason, the introduction of multiple unmanned surface vehicles loaded with oil pollution cleaning equipment has become a necessary way. In order to solve the efficient cooperation between multiple unmanned surface vehicles, a custom right-angle path planning algorithm is proposed. First, perform edge contour extraction, convex hull point search and clustering on a given oily image. Secondly, determine the target convex hull point and its assignment with multiple unmanned surface vehicles; then, through the custom right-angle method path planning, a collision-free path from the current position to the target convex hull point is planned. Finally, a comparative simulation experiment was carried out to verify that the custom right-angle path planning algorithm has fewer turning points in the path, easy to track by an unmanned surface vehicle, and more feasible and superior.

Abstract:In view of the increasingly complex topology structure of power grid, it is difficult to quickly mine effective fault information from massive data after a fault and has high computational complexity, and the fault data is incomplete and uncertain, leading to the failure to get correct diagnosis results. To solve this problem, this paper introduces the self-coding association rule mining algorithm (ACT-Apriori) into power grid fault diagnosis. The initial fault decision table was established by taking the protection and circuit breaker action data as the condition attribute and the fault line as the decision attribute. Then a self-coding association rule mining algorithm is used to extract kernel attributes and the optimal threshold is determined by dynamic threshold interactive mining technology. Finally, the simplest fault decision table is formed, and the fault information of each case is diagnosed and reasoned. In this paper, the four-bus distribution system is used as the simulation object, and compared with the traditional Apriori algorithm, FP-growth algorithm and the latest FP-Network algorithm, the calculation results show that: Compared with the traditional association rule algorithm, the running time of the improved algorithm is reduced by 90.69% and 83.55%, and the memory footprint is reduced by 21.43% and 15.38%, respectively. Compared with the FP-Network algorithm, the time complexity and space complexity are optimized to a certain extent. In addition, the proposed algorithm has high fault tolerance for single, double and rare faults with incomplete fault data, and the accuracy is 95.24%, which can effectively achieve rapid fault diagnosis.

Abstract:AbstractA Semi-Global Stereo matching algorithm with reordered Census transform and unidirectional dynamic programming optimization is proposed for improving match accuracy and weak immunity in the computation of matching cost. Firstly, the pixels in the Census Transform window in different scales are reordered and the median values are taken to calculate the Hamming distance, which solves the problem of over-reliance on the center pixel of the Census Transform window in traditional algorithm. Then, to improve the matching accuracy, the path aggregation algorithm based on unidirectional dynamic programming is applied to optimize the initial generation value, which can reduce the abnormal matching points and perfect the parallax reconstruction of the weak texture parts. Finally, an winner-take-all strategy is adopted to select the parallax corresponding to the minimum cost aggregation value for pixel selection, and the wrong parallax is eliminated by using left-right consistency detection in the parallax optimization stage. The experiment shows that this improved semi-global stereo matching algorithm generates an 8.22% reduction in the average mis-match rate of the initial parallax map, which is relatively higher in quality, and the flat mismatch rate under different noises is below 8%, which effectively enhances the robustness against noise and improves the matching accuracy.

Abstract:When interferometer algorithm works, it uses the phase differences of signal and the phase differences in local data to calculate the DOA. The interferometer direction-finding method has been widely used in engineering because of its flexible and the small amount of calculation. However, because of the array fabrication errors and the receivers’ inconsistent of phase, the accuracy of interferometer could be strongly affected as the calibration effect of local phase samples will fail when time-changing phase errors in present. Aiming at improving the algorithm stability in phase interference, we propose a refinement method of interferometer algorithm based on the response anisotropy of array and its optimized version, both of which improve the algorithm accuracy and weaken the dependence on the system’s phase consistency. Furthermore, computer simulation verifies that under the condition of different phase errors, the average directional error of the optimized interferometer algorithm is reduced to 38% compared with traditional interferometer algorithm, which shows that the accuracy and stability of the interferometer algorithm are significantly improved under interference conditions.

Abstract:Aiming at the shortest optimal route planning, firstly, this paper comprehensively analyzes the constraints and simulation environment of UAV route planning, constructs the route planning algorithm simulation environment, defines the performance constraints of UAV, and then proposes a route evaluation function that can integrate multiple constraints; Then, aiming at the problems of local optimum and slow convergence of genetic algorithm, considering the coupling relationship between the problems, a fusion improvement scheme of fitness value calibration, population diversification and elite retention strategy is proposed. The experimental results show that the improved genetic algorithm can save about 11.8% fuel loss, and the UAV has relatively fewer maneuvers, which improves the flight safety and efficiency of UAV.

Abstract:Aiming at the problem that asymmetric delay will seriously affect the clock synchronization accuracy under high network load, a linear estimation method of clock model is proposed to improve the clock synchronization accuracy under high network load. Firstly, a linear model is established for the master and slave clocks; and then the slave clock uses the master clock as the reference clock to process the four timestamps obtained in the current synchronization cycle, and combine them into two endpoints, through at least two synchronous cycle timestamp information and characteristic of the endpoint, find linear upper bound function and linear lower bound function from the model of the slave clock, the linear function of the slave cl-ock in the current synchronization cycle is determined by the mean value of the two, and the timestamp value is estimated a-ccording to the linear function, so as to estimate the master/slave clock offset in the current synchronization cycle. In order to verify the effectiveness of the proposed algorithm, a clock synchronization module based on the open source software Linux PTP is used to perform experimental verification and synchronization accuracy test on the DAC model and the proposed algor-ithm. Experimental results show that the clock model linear estimation algorithm avoids the continuous compensation in the sa-me direction for the local clock frequency, and achieves 23.48ns clock synchronization accuracy while making up for the defi-ciency of the DAC model.

Abstract:In order to effectively suppress background noise in the complex environment of the factory and obtain the useful information contained in the audio signal, an audio noise reduction method based on the combination of multi-window spectral subtraction and least mean square filtering algorithm is proposed. First use the improved multi-window spectrum subtraction, the gain factor in the modification of the spectrum reduction relationship is used to suppress the noise-free frequency, effectively avoid the production of music noise and increase the audio perception of non-smooth noise interference. Then use the changeable step-by-side adaptive LMS filtering algorithm based on the double-type normal cleaning function to adjust the step-up audio signal that has initially denoising the audio signal, thereby achieving the purpose of eliminating the noise component in the audio. The simulation experiment results show that this method is less than about 7dB of the signal-to-noise ratio of 7dB compared to the traditional multi-window spectrum reduction, and the fixed step LMS algorithm is increased by 3 to 4 dB, and the LMS algorithm of the traditional multi-window calibration step is increased by 1 to 2dB, and this method is simple and easy to have a good practical application value.

Abstract:To improve the adaptability of the models in the power grid with changing topology frequently, based on the Fuzzy Reasoning Spiking Neural P System (FRSNPS), this method takes lines, buses, and transformers as the candidate faulty elements and three universal diagnosis models are established. Even with the topology change, the three universal diagnosis models have invariable structures. Firstly, fuzzy initial values are used to represent the possible incomplete and uncertain alarm data. Simultaneously, according to topology around the candidate faulty element and the operation of the protective relays and circuit breakers, the input neurons are normalized to reduce the modeling complexity and enhance the universality. And considering the fault characteristics of different elements, different rule neurons are introduced in the matrix reasoning to improve the tolerance rate of fault diagnosis. Finally, the three models are used to diagnose the failure cases in IEEE 30-node system. And the model is compared with the traditional FRSNPS and Petri Net methods. The three diagnosis models have simple structure. In the case of abnormal operation of the protection system, they can still diagnose the faulty elements with 100% efficiency, and the average fault confidence is 0.8161, which is higher than the other two methods, and can effectively adapt to the power grid with changing topology frequently.

Abstract:Aiming at the application of large amount of data interaction between processing system and programmable logic in system of chip, a data interaction method between PS and PL based on double data rate synchronous dynamic random access memory is proposed. PS and PL interact with each other according to the custom protocol by accessing the common DDR. The space for data interaction in DDR is divided into instruction space and data space. PS and PL control their reading and writing processes by reading and writing instruction data in instruction space and analyzing the information they transmit according to the protocol. PL accesses DDR through high-speed on-chip bus, and PS uses memory read-write tools to read and write DDR. The test results show that the interaction method has the advantages of high speed, less logic resources and convenient use. The data interaction speed can reach 88MB/s. It is suitable for the application scenario where PS and PL need to interact a large amount of data in real time. It has been successfully applied in the vehicle real-time high-precision positioning system based on three-dimensional lidar.

Abstract:Aiming at the problem of water supply pipeline leakage signal detection and leakage point location, a leakage signal detection system for water supply pipeline is designed. The system is composed of two vibration signal acquisition terminals and upper computer software based on MATLAB. Based on the characteristics of leakage signal, the pre signal conversion circuit, band-pass filter circuit and ADC data acquisition circuit of vibration signal acquisition terminal are analyzed and designed. At the same time, the design principle and function of MCU program and upper computer software are analyzed and explained. VC-02 vibration calibrator and simulated water supply network are used to test the system functions of weak vibration signal acquisition and leakage point location. The test results show that the system can accurately collect the weak vibration signal with acceleration not less than 0.01g (1g = 9.80m / s ^ 2). The maximum relative error of locating the leakage point of 10m UPVC pipeline is 3.97%, and the average relative error is 3.51%.

Abstract:Owing to the continuous changes in vehicles under different road monitoring perspectives, vehicle re-identification is still a challenging task in intelligent traffic system. Most of the existing vehicle re-identification methods are based on the appearance attributes of the vehicle, but the recognition is affected by factors such as illumination and angle, which leads to poor recognition results. Therefore, this paper designs a vehicle posture perception attention enhancement network to improve the re-identification effect of vehicles under the influence of factors such as illumination and angle. First, input the image to the convolutional pose machine to generate 12 keypoints to reconstruct the vehicle frame, and then compare the input image vehicle with the target image vehicle to extract the features of the intersecting area between two images; Finally, the global distance and local loss of vehicle features are calculated, and the recognition results are sorted according to the final results. This paper verifies on vehicle ID and VeRi776 data sets. The experimental results prove that the Top10 detection accuracy of the proposed network is increased by about 10% than other models.

Abstract:Aiming at the problems of low efficiency, low accuracy and inconsistent measurement in the manual measurement of the groove size of the inner wire joint in the actual industrial inspection, a groove size measurement system of the inner wire joint based on machine vision is developed. Firstly, the groove position is located by morphology, and then the edge of the groove is accurately found by bilateral filtering combined with Roberts operator, Finally, the edge is fitted into straight line segments by the least square method, the minimum distance between straight line segments is calculated by Halcon image processing software, and the measurement results are displayed by c# combining Halcon library. The experimental results show that the measurement accuracy of the system is ± 0.01mm, the detection accuracy of the measurement system is 98.73%, the missed detection rate is 0, and the over detection rate is 1.27%; In terms of detection time, the average time for detecting a workpiece is 0.37s. The system has high measurement accuracy, stable measurement and fast operation speed, so it can effectively replace the manual measurement in industrial detection.

Abstract:Aiming at the problems that the traditional detection methods of motorcycle helmet detection have low accuracy, poor generalization ability and large number of target detection network parameters, which are difficult to run on embedded devices, an improved MNXt-ECA-D-YOLOv2 target detection algorithm model of YOLOv2 is proposed. First, MobileNeXt network is introduced to replace original YOLOv2 backbone network, and a densely connected network structure is introduced into the sandglass block of MobileNeXt. At the same time, the effective channel attention mechanism is introduced into the network. And, different activation functions are applied at different depth network layers. Finally, DropBlock module is added before the network output convolutional layer. K-means clustering algorithm is adopted to redesign the anchor box size of self-made dataset. The experimental results show that compared with the original YOLOv2 under the same experimental conditions, the proposed method improves the AP50 metric by 3.53% and the model size reduced by 77.44%, and the detection speed increased by nearly 4 times. Comparison experiments demonstrate that the improved YOLOv2 has a higher average accuracy rate, a smaller model, and faster inference speed in CPU. Therefore, the proposed improved YOLOv2 model is valuable in practical applications.

Abstract:Aiming at the difficulty of filtering the clustered and isolated data far away from the main point cloud caused by occlusion in the linear measurement data of LiDAR, a point cloud filtering method combined with digital signal processing technology was proposed. The height difference sequence is constructed for the adjacent points in the data section, the discrete Fourier transform was used to solve the amplitude frequency response of the height difference sequence. After low-pass filtering and inverse Fourier transform of frequency domain signal, the comparison signal with greatly modified isolated data was restored. By comparison signal ,we cloud find the location of isolated points in the point data section , so as to filter the point information of isolated area. The performance of the algorithm is verified by experiments. The results show that the filtering method can filter all the isolated data. At the same time, the class I error and class III error are less than 1%, and the filtering has little impact on the characteristics of the main point cloud.It meets the application requirements of Point Cloud Filtering in medium and large measurement fields, and has good engineering application value.

Abstract:Aiming at the problem of low average precision of small target detection in industrial large-size images, an improved Faster R-CNN-Tiny model is proposed. Firstly, the feature pyramid structure is used to improve the second-order detector Faster R-CNN to enhance the feature expression capability and increase the resolution of small target feature mapping to improve the prediction accuracy; secondly, the last piece of the original ResNet structure is changed to deformable convolution to automatically calculate the offset of each point and take features from the most suitable place for convolution, which is used to enhance the small target region Finally, when extracting the features of the region of interest, the contextual information of the content is introduced to improve the accuracy of small target detection. The comparison tests are conducted on the representative satellite remote sensing UCAS-AOD dataset in industry and the quality inspection dataset of surface defects of tiles in Tianchi. The results show that the improved FRC-Tiny model improves the mean average precision of detection by 5.57% and 14.25%, respectively, compared with the original model.

Abstract:Good pantograph net relationship is one of the important factors for the normal run of electrified railway. Using machine vision technology to measure the thickness of pantograph slide can reduce the pantograph net accidents caused by human error. The pantograph slide pictures taken in the laboratory environment were processed by image grayscale processing, perspective correction, pantograph slide image filtering and pantograph slide image enhancement. This paper proposes a method of image tracking after edge detection and morphology open and close operation on pantograph slide image, realize the pantograph slide edge detection and positioning, The minimum value of pantograph slide thickness is calculated. Carry out measurement experiment to the pantograph slide, and the results obtained by this method were compared with the measured values, and the measure error was basically ±0.5mm. The results show that the image processing method proposed in this paper can realize the pantograph slide abrasion detection and has great value to the practical application research.

Abstract:Aiming at the low accuracy of respiratory wave extraction, an improved method for respiratory wave extraction from photoplethysmography (PPG) signal is proposed. Ten groups of pulse and respiratory signals were obtained from the mimic database. The variation mode decomposition (VMD) algorithm optimized by genetic mutation particle swarm optimization is used to decompose the pulse signal of photo capacitance product in the same period, and the intrinsic mode function (IMF) is obtained. The IMF component with correlation coefficient greater than 0.3 is selected to reconstruct the respiratory signal, and the reconstructed respiratory signal is compared with the original respiratory signal. The experimental results show that the average accuracy of respiratory rate is 0.95, the average value of waveform correlation coefficient (RCC) is 0.9451, and the average value of root mean square error (RMSE) is 2.0110. Compared with EMD and EEMD, the algorithm improves respiratory rate by 5% and 3%, and RCC by 19.96% and 13.17%, with higher accuracy. At the same time, the algorithm overcomes the uncertainty of penalty factor and decomposition level selection in VMD algorithm. This is of great significance to clinical practice.

Abstract:To better use pulse diagnosis methods to study the changes of the individual's physiological state. Given the current pulse wave parameter research that does not consider the pulse wave changes under different pressures, this paper starts from the process of taking the pulse of traditional Chinese doctors and designs a three-channel pulse acquisition system that can apply pressure. The pulse acquisition under step pressure is used. Correlation test and wavelet multi-scale analysis of the pulse waves of 51 individuals of different ages under the optimal pulse pressure, found that the Pearson correlation coefficient r=+0.64 between the rise time of the main pulse wave and the age. The Pearson correlation coefficients of the pulse wave in the high-frequency range (7.7-15.8 Hz and 3.85-7.92 Hz) and age are r1 = -0.69 and r2 = -0.75, respectively. With the low-frequency (0-3.9Hz) energy ratio and the Pearson correlation coefficient of individual age r3 = +0.77, the difference is highly significant (P<0.0001). The results show that when the pulse information is collected under pressure, the pulse wave rise time ratio and pulse wave energy distribution under the optimal pulse pressure may be an effective parameter for evaluating arteriosclerosis, which can be further studied.

Abstract:The unbalance of the wind turbine rotor is unavoidable during its operation, and it will affect its reliability and reduce its lifetime for a long time. Aiming at the problem that the non-linearity and instability of unbalance of the wind turbine rotor, the traditional frequency domain transform spectrum analysis method has certain limitations. Therefore, this paper studies a method for detecting unbalance of the rotor based on variation mode decomposition. Combined with the actual operating data of the wind farm for comparative analysis, this method can decompose the complex multi-signal into several signals, and can effectively extract the fault characteristics. Compared with the traditional frequency domain variation method, it has a great advantage. The results indicate that the rotor aerodynamic unbalance will cause the 1-P vibration significantly increase, and the greater of the deviation of the pitch angle, the greater the axial vibration changes.

Abstract:Since the existing data-driven remaining useful life (RUL) prediction methods of rolling bearings still need a lot of prior knowledge to extract features, construct health indicators, and set fault thresholds, a direct RUL prediction method based on time convolution network (TCN) with multi-head attention mechanism is proposed. In this method, the short-time Fourier transform (STFT) of the original vibration signal is used as the input of the stack noise reduction automatic encoder (SDAE) to get the depth feature representation, and then input it to the attention TCN for RUL prediction. Finally, an example is verified in the rolling bearing data set of PRONOSTIA. The results show that the prediction error-index MAE and MAPE of this method are 53.92% and 46.13% lower than those of the other four methods, respectively, and the score index is 52.98% higher than that of these methods.

Abstract:Rolling bearing is an important part of high-tech mechanical equipment and also one of the important fault sources. At present, few bearing fault samples, uneven data distribution and unstable effect of traditional bearing fault identification methods bring great difficulties to fault identification technology. A generalized S-transform method was proposed by combining deep learning correlation technology with bearing fault diagnosis technology, and taking advantage of deep learning model to recognize two-dimensional images. Generalized S-transform is the inheritance and development of wavelet transform and short-time Fourier transform. By transforming one-dimensional bearing fault signal data into two-dimensional time-frequency diagram, the model of Xception network is fine-tuned and the hyperparameters are optimized, and then the two-dimensional time-frequency diagram is input into the improved Xception network to carry out transfer learning. The above experiments were carried out based on rolling bearing data published by Case Western Reserve University, and the recognition rate of fault signals under diff ____________________________________________________________________________________________ *基金项目：国家自然科学基金项目( 61372128)，教育部协同育人项目(202002179030)，南京信息工程大学滨江学院科研与教研项目（2020yng001,JGZDI201902） erent working conditions reached 99.95%. The experimental results prove that the recognition method based on generalized S-transform and transfer learning is real and effective.

Abstract:The detection method of longitudinal tear of single signal conveyor belt based on image or sound is often affected by light and noise. In order to overcome this limitation, a damage detection method based on multi scale feature cross fusion is proposed. Firstly, Log-Mel feature extraction algorithm is used to upgrade one-dimensional sound signal to two-dimensional spectrum; Secondly, a dual input neural network is built to extract the features of the image and the sound spectrogram at the same time, and they are cross fused at different scales; Finally, according to the fused features, the damage classification is determined by the loss function. Through experiments, the detection accuracy, sensitivity of scratch and longitudinal tear of this method can reach 97.37%, 96.53% and 98.67% respectively, which is 7.04%, 6.96% and 6.4% higher than the existing audio-visual decision level fusion method. Therefore, this method can better meet the reliability requirements of conveyor belt damage detection.