Abstract:The new international system of units (SI) had been implemented since 20th May 2019, that has epoch making magnificent meaning. This review includes each developing process to seven constants from the seven elementary units of Second, Meter, Kilogram, Ampere, Kelvin, Candela, Mole and revealed intrinsic relationship between them, and gave the deep analysis of the meaning of the unit’s definitions. This old seven elementary units were still irreplaceable from SI dimension, but they have no more acted as the source of traceability, and their definition has been updated. However, the seven elementary constants have already become the root of traceability. Here also gives a prospect of the meaning that new SI will develop metrology.

Abstract:The estimation accuracy of state of charge (SOC) based on battery model mainly depends on the accuracy of the model. Under dynamic conditions, the input current of the battery changes drastically, and the traditional identification method has poor convergence, which leads to the reduction of model accuracy. In order to improve the accuracy of battery model under dynamic conditions, the traditional least square method with forgetting factor (FFRLS) is improved. By setting the accuracy threshold and introducing gradient correction method, an improved recursive least square method with forgetting factor (IFFRLS) is proposed. Online parameter identification is carried out by using the improved algorithm, and a secondorder RC equivalent circuit model is established. Compared with other models established by traditional parameter identification, the effectiveness of IFFRLS in improving the accuracy of the model is verified, and the average error of the model is 0003 8 V. Finally, the models established by different identification methods are combined with EKF algorithm to estimate SOC, and their errors are compared. The results show that the highprecision model identified by IFFRLS can effectively improve the estimation accuracy of SOC, and the error is within 151% under DST condition.

Abstract:Aiming at the problem of cracks in solar cells during the production process and under the condition of limited database of solar cell defects, the particle swarm optimization (PSO) is applied to optimize the support vector machines (SVM) to detect the surface cracks of solar cells. Firstly, in order to reduce the influence of uneven light distribution caused by electroluminescence (EL) detection in the image acquisition process, Retinex enhancement processing is performed on the image of the solar cell assembly. Secondly, in the frequency domain, the Gabor transform is used to extract the texture features of the image to obtain the crack feature. Finally, the texture features of each solar cell component are reduced by principal component analysis (PCA) and then they are input into the PSO_SVM system for classification and recognition. Using this method to experiment with 600 EL images of solar cells, only one image was detected by mistake, and the classification accuracy is 9933%. Comparing this algorithm with decision tree classification, extreme learning machine (ELM), convolutional neural network (CNN) and SVM algorithm, PSO_SVM achieves the highest recognition rate.

Abstract:Aiming at the problem of network overfitting and model performance under standard caused by the lack of defective data amount of solar cells, In this paper, a true and false data fusion algorithm based on deep convolution confrontation generation network and random image Mosaic is proposed, which improves the training data volume by 800 times. At the same time, the network model is optimized with light weight to reduce model training parameters. The experimental results show that the test accuracy of the trained model after the data fusion and expansion of the data set is nearly 30% and 17% higher than that of the original training set and the traditional data enhancement algorithm. After the lightweight treatment, the model parameters were reduced to about half of the previous ones, and the test time for each image was shortened from 57 ms to 22 ms. The research shows that the fusion algorithm can effectively alleviate the problem of network overfitting caused by insufficient training data. The lightweight optimization model not only ensures the accuracy, but also compresses the size of the model to speed up the test.

Abstract:Aiming at the problem that the particle filter cycle life prediction algorithm has a poor longterm prediction effect on lithium iron phosphate battery, neural network is used to learn the historical data of the battery, and the training learning value is substituted as the observation value into the particle filter algorithm to modify the particle state value; for phosphoric acid In the dynamic equation of the lithium iron battery, there is no problem that the life is directly related to the observation value. The posterior probability relationship between the battery life and the capacity observation value is derived. The posterior probability density relationship under the Monte Carlo method is obtained, and the battery life is given. Forecast uncertainty expression. The experimental results show that the neural network training value is used as the observation value of the improved particle filter dynamic equation algorithm, and the method is effective and reduces the prediction error.

Abstract:The complex heterogeneous background in the nearinfrared images of photovoltaic solar cells makes the detection of internal defects become a very challenging problem.Thus, an object detection framework based on deeplearning residual channel attention Faster RCNN (RCAFaster RCNN) is proposed, which employs convolution layer and pooling layer to extract the image features, and sends them to the novel residual channel attention (RCA) module for complex background feature suppression and defect feature highlighting, then the region proposal network recommends a more accurate proposal containing defects, finally the classification and position network is applied to achieve highprecision defect classification and position estimation.The experimental results show that the defect detection accuracy of RCAFaster RCNN has improved to 8329%, which proves the effectiveness of the proposed method.

Abstract:circuit model; Thevenin model; DP model; model evaluation

Abstract:The fatigue performance of copper conductor is a key factor affecting the service life and reliability of flexible electronic equipment. Therefore, a quantitative test method for the fatigue performance of membrane  base structures is proposed. Firstly, the variation rule of viscoelastic parameters of copper wires with deposition thickness of 25 nm to 400 nm on flexible substrate was analyzed by using nanometer dynamic mechanical analysis technology, and the fatigue life of membrane  base structure was determined. Secondly, the microstructure of copper wires was observed and analyzed by scanning electron microscope (SEM) to explore the mechanism of wire thickness affecting the fatigue performance of membrane  base structure. Finally, the test results of copper wires with different thickness are analyzed and summarized to optimize the length and scale process design parameters of flexible electronic devices. The results show that the fatigue life of copper wires with the thickness of more than 100 nm is millions of times, the grain distribution is uniform, the average grain size is 50 nm, the stable distribution of the grain inhibits dislocation movement, and the fatigue performance is reliable. Therefore, copper wires with a thickness of 100 nm deposited on the flexible substrate have great advantages in improving the service life of flexible electronic devices.

Abstract:In wireless networkon chip, congestion around wireless node and load balance in subnet and global network play an important role in communication efficiency. In this paper wepropose a global traffic balance mechanism(GTB) basedon Edgefirst algorithm. Firstly, we optimize wired/wireless data packet partition method to mitigate congestion around wire lessnode. Then we propose Edgefirst routing algorithm to balance load inside subnet according to wireless node running conditions. In the end,a global subnet congestion a wareness (GSCA) judge mechanism is proposed to endow long distance data packet with higher priority to transit through subnet with low congestion level, which balances traffic loading lobalnet work. Evaluationresults show that this mechanism guarantees low latency and high through out, significantly improves network load balance ability with acceptable hardware and power consumption.

Abstract:For the problems of aeroengines such as complicated health monitoring data, insignificant performance degradation characteristics, lack of effective health index construction method and difficulty in remaining useful life (RUL) prediction, an optimal selection and fusion based on grey theory for multimonitoring parameters and a grey forecasting model with full order time power terms (FOTPGM (1,1)) method are proposed. Firstly, the performance degradation state characterization capability of the highdimensional monitoring physical parameters of the aeroengine are parametrically measured by the grey relational analysis (GRA) method, the linear correlation analysis is applied to optimally select of the monitoring parameters as well. Secondly, the grey approximate correlation degree was utilized to make weighted fusion of the selected features, which establishes the uniform health index (HI) of the aeroengine. Taking the HI failure threshold of the training aeroengines as reference, the HI failure threshold of the test aeroengines was determined by the matching method of sliding WindowEuclidean distance. Finally, FOTPGM (1,1) was adopted to predict the RUL of the test aeroengines. Through the experimental analysis, this method can effectively predict the remaining useful life of the aeroengine and obtain a higher prediction accuracy than the traditional method. This method provides a novel idea and means for realizing the fault prediction and health management of the aeroengines.

Abstract:To enhance the accuracy of gesture recognition using electromyogram(EMG) signals, we present an EMG signal feature extraction method based on timefrequence domain analysis. Firstly, a wireless EMG signal acquisition device is designed. Secondly, a gesture recognition method based on multivariate empirical mode decomposition (MEMD) and TeagerKaiser (TK) energy operator is proposed. Multidimensional scaling (MDS) method is used to reduce the dimensionality of multichannel features. then, linear discriminative classifier (LDA) is used to classify and recognize gesture features. The accuracy of this algorithm for UCI database can reach 9896%. The recognition accuracy for selfcollected database can reach 9937%. Meanwhile, F1 score also enhances significantly. The experiments verify that the method we proposed can reach a higher accuracy recognition results than other typical methods.

Abstract:In order to accurately measure the radiated emission caused by the pantograph arcing of EMU in the presence of environment electromagnetic noise, a transient electromagnetic signal extraction method based on sparse decomposition is proposed. Based on the Gabor atom library, an asymmetric Gaussian model is established by introducing an asymmetric parameter and an attenuation parameter, which makes the improved atom library more consistent with the asymmetric envelope and different attenuation rate characteristics of the pantograph arcing radiation pulses in the time domain. Based on the improved atomic library, the pantograph arcing signal is reconstructed by the sparse decomposition, thus realize the accurate measurement of pantograph arcing radiation. Simulation and laboratory experiments show that the method can effectively suppress the preecho phenomenon in the sparse decomposition based on traditional Gabor atomic library, and the reconstruction accuracy is higher; the denoising capability of the method under narrowband interference is about 17 dB higher than the denoising method based on EMD theory, and the error between the recovered transient signal and the original signal is within 3 dB.

Abstract:Aiming at the problems of long detection time, inconvenient operation and difficult online detection of traditional measurement methods of chemical oxygen demand (COD), a COD detector based on sequential injection analysis (SIA) and image processing technology is designed. The core of the system is the principle design of sequential injection image analysis platform. Image processing process design, digestion and detection process design, digestion tank and structure design of the detection pool. Experimental research on the new method of SIA combined with image processing and analysis technology to determine COD is carried out. Based on the national standard water quality detection standard, the sequential injection image analysis and detection process is improved, and after the water sample is taken, the potassium dichromate and silver sulfate are added directly to digestion, and then the image is collected after cooling. Based on the partial least squares regression of the single variable, the regression model of COD concentration on the characteristic values of image color Red(R), Green(G), Blue(B), Hue（H), Saturation(S) and Intensity(I) is established, and the regression equation is obtained. The multiple coefficient of determination R2=0995 1, and the root mean square error RMSE=4937 3. The experimental results showed that the minimum detection limit was 20 mg·L-1, and there was no significant difference in repeatability (RSD≤438%) compared with the national standard method. The system can analyze COD content in different water samples stably and efficiently. The method of COD determination based on SIA combined with image processing is helpful to improve the technical performance of water quality testing instrument and is suitable for many kinds of water quality testing parameters.

Abstract:In order to solve the problem of poor adaptability of the existing fuzzy PID control algorithm when the train is running at high speed in complex environment, the tracking error of the train will be large when it is disturbed by external factors. This paper proposes a train speed control algorithm based on RBF neural network PID control. Firstly, when building the optimization model of the train, the characteristics that the train must be in the inert condition when passing through the phase separation area are fully considered. According to the characteristics of the electric phase separation and speed limit conditions, the sections in the process of the train running are divided to simplify the solving process. Then, the RBF neural network PID controller is used to track and simulate the target speed curve. At the same time, the existing fuzzy PID control comparison was made. The results show that the speed tracking algorithm based on RBF neural network is effective.

Abstract:Aiming at the problem that the poor recognition ability of mobile robot to AprilTag image is caused by uneven illumination intensity and overly fast movement, an improved method of AprilTag image preprocessing was proposed to improve the success rate of image recognition. Firstly, a method of tail shifting is used to change the colored image into gray image. On the basis of the above, a bilinear interpolation method is adopted to carry out downsampling on the gray image to improve the overall processing speed. The gray image is processed by histogram equalization to eliminate the influence of uneven illumination and enhance the contrast of gray image. Then, bilateral filtering was carried out for the gray image to remove the image noise, and Canny operator was used to detect the edge of gray image to improve the success rate of AprilTag image recognition and the subsequent positioning accuracy. The experiment verifies the effectiveness of the improved algorithm. The success rate of the proposed method for AprilTag image recognition is increased by more than 3% compared with the traditional method under different light conditions, and the realtime positioning error of the mobile robot is controlled within 1～2 cm, the validity and feasibility of this method are verified.

Abstract:In order to solve the problem that it is difficult to identify the illegal use of mobile phones accurately and efficiently in some places with special regulations on mobile phone use, this paper proposes an improved SSD model to detect the illegal use of mobile phones. The SSD model is used to obtain the location and area classification of the primary target, and the improved DenseNet model is used to determine the primary target frame, so as to obtain the accurate mobile phone detection boundary box. In order to improve the data preprocessing process, the strategy of combining data amplification with image quality improvement is adopted. The experimental results on the mobile phone set built by ourselves verifying the effectiveness of these improved strategies. The location accuracy of the improved SSD model can reach 911%, and the recognition accuracy can reach 98%, which is 35% higher than the original SSD model. The improved SSD model has the characteristics of high recognition accuracy and high positioning accuracy, which can provide theoretical basis and technical support for intelligent recognition of illegal use of mobile phones.

Abstract:It is difficult for the existing edge detection technology to eliminate the influence of noise and scratch on the edge detection, and maintain the clarity and continuity of image edge. An improved edge detection technique based on second order differential operator and mathematical morphology is proposed. Firstly, considering the morphological characteristics, the image processing method of open and closed morphological operation is improved to remove the scratches on the workpiece surface in advance. Then the edges of the workpiece are detected by the secondorder differential Laplace operator. Finally, in order to achieve better image denoising effect, a kind of enhanced denoising combining Gaussian and bilateral filtering is improved, and the final algorithm is verified by experiments. Experimental results show that the improved algorithm is effective in removing the surface scratches of the workpiece, and the edge clarity and peak signaltonoise ratio (PSNR) are greatly improved, laying a foundation for improving the workpiece identification accuracy.

Abstract:For the problem of complex texture and difficulty in defect detection of mesh fabric. An algorithm based on local binary pattern (LBP) and low rank sparse matrix decomposition for defect detection of mesh fabric is proposed. Firstly, the local binary pattern with equivalent invariant rotation is used to extract the features of the mesh fabric image, and the texture feature matrix is obtained. Then, an appropriate lowrank sparse decomposition model is constructed based on the texture feature matrix. Finally, the significant graph generated by sparse matrix was segmented by OTSU optimal threshold segmentation algorithm, so that the defects of mesh fabric could be detected. Compared with KSVD algorithm, the experimental results show that the average accuracy of the method in this paper is 8994%, the average recall rate is over 9388%, and the total accuracy of classification is over 92%.

Abstract:In order to solve the problem that the local binary pattern (LBP) is easily affected by random noise and edge points on the image, and the threshold cannot be automatically selected when the local binary mode describes the texture features of the image, resulting in poor robustness, a palmprint recognition method based on weighted adaptive multiple uniform local binary pattern (WAMULBP) and twodimensional principal component analysis (2DPCA) is proposed. Firstly, the histogram equalization (HE) is used to perform pretreatment of the palmprint region of interest (ROI) image to reduce the impact of the illumination change during imaging on the final palmprint recognition success rate. Secondly, the preprocessed image is divided into sizes Uniform subblocks and use adaptive multiple uniform local binary pattern (AMULBP) algorithm to obtain texture feature histograms and weights of each subblock. Finally, the texture feature histogram of each subblock is multiplied and concatenated to obtain the final texture feature histogram, after the 2DPCA dimension reduction, the Euclidean distance discriminant method is used for palmprint recognition. Comparing experiments on Hong Kong Polytechnic University PolyU database, Tongji and IITD noncontact database, selfbuilt noncontact database and their noise database. The lowest equivalent error rates are 1879 0%, 2019 2%, 2184 9%, 2663 2%, 4380 3%, 4730 1%, 5005 0% and 5223 7% and the recognition time is within 1 s. Compared with other algorithms, the recognition accuracy and robustness are effectively improved while ensuring realtime performance.

Abstract:Aiming at the problems of high station deployment cost,equipment maintenance difficulties and insufficient flexibility of the radar system caused by the wired transmission of data in the existing high frequency surface wave radar(HFSWR) system, a radar data wireless transmission scheme is proposed and a rapid evaluation method of the wireless communication quality is designed in this paper. In this scheme,the wireless transmission link network is established through the wireless bridges, and TCP protocol is used to achieve reliable data transmission. Communication test results in various scenarios show that in the case that there is no obvious obstruction between the transmitter and the receiver and the transmission distance is less than 400 m, an effective transmission rate of 11 Mbps per link can be guaranteed. On this basis, the wireless transmission scheme has been applied to the surface wave radar system. The feasibility of the wireless transmission of radar data has been verified by field test and the reliability of the evaluation methodhas been verified.

Abstract:To solve the problem of model switching slowness in interactive multiple model (IMM) algorithm when the target is maneuvering, an adaptive IMM (AIMM) algorithm with Markov probability transfer matrix online correction is presented. The Markov probability transfer matrix is adjusted by the probability difference between two consecutive moments in the IMM submodel to improve the switching speed and the rationality of the assignment of the submodel, and the tracking accuracy is improved. Secondly, squareroot volumetric Kalman filter (SRCKF) is introduced into AIMM algorithm to solve the nonpositive definite problem of covariance matrix and improve the numerical stability during iterative filtering. AIMMSRCKF algorithm for maneuvering target tracking is proposed. The simulation results show that the algorithm can improve the probability of matching models and shorten the model switching time.

Abstract:Aiming at the problem that the power line foreign object recognition model can use fewer data sets, and the traditional SinGAN model generated data does not match the foreign object recognition model, the quality is poor, and it takes too long, the improved SinGAN model is proposed. Based on the improved SinGAN model, an affine transformation unit and a size transformation unit are added to further enhance the data set, and an image filtering unit is added to improve the data quality required by the power line foreign object recognition model. By improving the SinGAN back propagation training process and SinGAN’s singleprecision generator structure, the quality of model generation is improved and the time spent is reduced. Experimental results show that after 50 experiments, the average Frechet starting distance score (FID) of the improved SinGAN is 91375, and the average training time is 121 h. Compared with traditional SinGAN, it is reduced by 27247% and 8731% respectively. Compared with other mainstream generative adversarial networks, improved SinGAN has better foreign object data generation capability,which can enhance the data required by the power line foreign object recognition model, and has superiority.

Abstract:In order to solve the problems of information loss and artifact in the fusion of visible light and infrared image, the image fusion algorithm of VI and IR based on saliency detection coupled weight mapping is proposed. The VI and IR images were decomposed into two scales to obtain the base layer and detail layer. In order to improve the fusion effect of VI and IR, a salient feature detection method was defined, and the mean filter was applied to each source image to reduce the intensity change between the pixel and its adjacent pixels for finishing the smoothing process. Then the median filter was applied to each source image to eliminate noise or artifacts for preserving the edges. By taking the difference between the mean value and the median filter output to calculate the significance characteristics, the salient information such as edges and lines is highlighted. Then, the significance detection results are normalized to construct the weight mapping for assigning the appropriate weight to the detail layer. And the basic layer and detail layer are fused by different rules, in which the weight map and detail layer are combined to get the final detail information, and the basic layer is fused by the average fusion rules. Finally, the linear combination of the final base layer and detail layer were used to construct the new image. Experimental results show that compared with the current multiscale image fusion technology, the proposed algorithm only uses twoscale decomposition, which significantly improves the fusion efficiency, and the fusion image has better fusion quality, which effectively eliminates the artifacts and less information loss in the fusion process.

Abstract:In the inertial navigation system, in order to improve the attitude measurement accuracy of the gyroscope and suppress the influence of lowfrequency noise, the wavelet transform method is used to fuse the gyroscope and accelerometer data to solve the attitude angle. In this paper, the data collected by the gyroscope is firstly decomposed into two layers of wavelets to remove lowfrequency components and unstable signals, and reconstructed with highfrequency components to obtain filtered gyroscope data. Then use the data collected by the accelerometer to calculate the attitude angle, which is used to continuously iterate the initial quaternion, find the gravity vector from the initial quaternion, and then find the error from the cross product of the gravity vector, and perform PID control to correct the gyroscope’s angle. Finally, the corrected and filtered gyro data are used to calculate the new quaternion using the RungeKutta method, and the quaternion is used to adjust the negative gain, and finally the accurate attitude angle is calculated. The simulation results show that the accuracy of calculating the attitude angle is increased by 80%,which can effectively suppress lowfrequency noise and calculate the attitude angle more accurately, thereby further improving the positioning accuracy of the navigation system.

Abstract:At present, the 5×7 dot matrix spraycode character recognition on the surface of tire rubber has many problems, such as high labor intensity, low efficiency and low intelligence level. Based on machine vision, a novel character recognition method for tire rubber surface is proposed. Firstly, the character occurrence area is stable and the character is white on black background for fast positioning, background segmentation and using morphological manipulation to remove interference. Then, the simplified vertical projection method is innovatively used to fit the threshold of character width, and the special cases of “1“and “I” which are different from other character widths are processed to achieve character segmentation. Finally, the template matching method based on standard correlation matching is used to realize character recognition. Experimental results show that the accuracy of the proposed method is 9951%, which achieves the expected recognition effect.