Abstract:This article addresses the issue of traditional video encryption systems being unable to balance real-time and security in encrypting large amounts of data, and designs and implements a video encryption scheme based on the SM2 algorithm. This scheme is based on the State Secrets SM2 algorithm and a lightweight stream cipher generated by the "xoshiro256ss" pseudo-random number generator. It uses hybrid encryption and "one-time encryption" to ensure data security. This solution achieves data exchange between devices and streaming media servers through the RTSP protocol, and can decrypt and play video data through QT, providing an effective solution to the problem of video surveillance encryption that cannot simultaneously balance security and real-time performance. After testing, the system has an encryption time of about 3 ms for I-frames.

Abstract:To address the problem of reduced operational efficiency of photovoltaic cleaning robots caused by inconvenient transportation in the desert area of Shago, an aerial transfer system for photovoltaic cleaning robots based on drones was proposed in this study. Initially, a near-infrared light source guide sign was designed, and the OPENMV module was employed for image recognition. The threshold segmentation method and edge recognition method were applied to construct a guide sign recognition algorithm, enhancing the positioning accuracy and anti-interference capability of the guide sign. Subsequently, a conical adsorption docking structure and a parallel symmetrical centering docking structure were designed to solve the docking issues between photovoltaic cleaning robots and drones under inclined conditions. A prototype of the drones and two docking platforms was then developed, and experiments were conducted on the visual positioning accuracy of drones and the transportation performance of the docking platforms. The experimental results indicate that the visual positioning error of drones remains within 10 cm, effectively reducing the external environment′s interference on the recognition effect. Both docking platforms adapt to docking operation requirements at photovoltaic panel inclination angles within 40 degrees, achieving a docking success rate consistently exceeding 90%, thereby meeting the operational requirements for stable docking and simple separation.

Abstract:Addressing the issue of significant phase errors in the secondary side current and reduced system transfer efficiency due to the neglect of higher harmonics when the operating frequency of the bidirectional half-bridge CLLC deviates from the resonant point and the parameters of the front and rear resonant cavities do not match, the traditional fundamental harmonic analysis method is used to calculate the phase of the secondary side current for synchronous rectification. This paper proposes to include the higher harmonics ignored by the traditional fundamental harmonic analysis in the calculation, using the extended harmonic analysis to equivalently represent each harmonic as an independent voltage source, and to correct the equivalent resistance of the load in conjunction with the law of conservation of power, further establishing an extended harmonic approximation model. According to the model, the effects of each harmonic acting individually are calculated, and the results of the actions of each harmonic are accumulated to derive the expression of the secondary side resonant current, from which the zero-crossing angle of the secondary side resonant current can be calculated, achieving accurate calculation of the phase where the secondary side current crosses zero. Finally, a 500 W rated power experimental prototype is built, and the experiment shows that compared with the traditional fundamental harmonic analysis method, the efficiency of the converter is improved across the full load range, with the highest efficiency improvement of 1.8% at the rated power, verifying the correctness and effectiveness of the proposed theory.

Abstract:In order to solve the low power and fast transient response problem of LDO without off-chip capacitance, a transient enhancement LDO circuit based on feedforward compensation was proposed. The error amplifier uses a recycling folded cascode amplifier, which has low power consumption and high gain. The feedforward compensation is implemented by connecting multiple small gain stages, which improve the system stability with low quiescent current. The output voltage is coupled to the transient enhancement circuit through capacitors, providing a charging and discharging path for the power transistor to improve the transient response performance. The circuit is designed based on the SMIC 180 nm CMOS process, the simulation results show that after the LDO transient enhancement within the load current jumping range of 100 μA to 50 mA. The output voltage overshoot is reduced by 343 mV, or about 39%, the output voltage undershoot is reduced by 592 mV, or about 57% . The load regulation is 0.005 7 mV/mA and the line regulation is 0.22 mV/V. The quiescent current of the circuit is about 3 μA, and the LDO current efficiency is 99.99%.

Abstract:In the field of autonomous driving, simulation-based testing is an important means for identifying and addressing long-tail problems. This paper proposes a driving risk field model that considers road curvature radius for cut-in scenarios at curves. By integrating Prescan/Simulink simulation platform and genetic algorithms, an automated risk scenario generation framework is constructed. The framework uses the relative driving safety index (RDSI) as the optimization objective, which overcomes the limitations of traditional risk indicators that fail to promptly recognize cut-in risks by preceding vehicles in complex scenarios. Various cutin scenarios under different road curvature radii are generated through simulation, and typical test cases are selected for analysis. The results show that the RDSI indicator achieves a 65.6% higher warning success rate compared to the time-to-collision (TTC) indicator and can identify potential risks earlier. Additionally, experiments reveal that different road curvature radii significantly impact collision risk.

Abstract:A closed-loop control scheme for the driving mode of a vibrating ring gyroscope based on phase control and automatic gain control is proposed, which is to better control the working state of the gyroscope driving mode, thereby improving the stability of the driving amplitude and the real-time tracking ability of the resonant frequency. First, based on the dynamic characteristics of the vibrating ring gyroscope and the electromechanical interface features, an electromechanical interface with symmetric electrostatic driving and differential capacitive sensing was designed. On this basis, a closed-loop drive control scheme based on phase control and automatic gain control was proposed. Furthermore, an electromechanical coupling simulation model of the closed-loop drive control was constructed. The self-excited oscillation control of the drive mode, amplitude stability, and frequency tracking state were simulated, validating the dynamic performance of the system. Finally, a closed-loop control circuit for the driving mode was designed based on the proposed scheme, and the performance of the circuit was tested. The experimental results show that the driving signal frequency of this circuit can effectively track the variations in the resonant frequency of the gyroscope′s driving mode. When the temperature varies within the range of -40℃ to 60℃, the variation range of the drive detection voltage is less than 5.08%, and the relative change in amplitude with temperature is less than 0.059%, which meets the requirements for stable amplitude drive over the full temperature range.

Abstract:Fluxgate-based direct current transformer (DCCT) is widely used in applications such as instrument calibration and current monitoring in DC power distribution, where extremely high precision is required. To meet the demand, the ripple in the output signal must be minimized, with induced modulation ripple being the dominant component of the output ripple. Therefore, it is essential to model and analyze the transmission characteristics of the induced modulation ripple and investigate corresponding suppression methods. Based on the typical structure of fluxgate-based DCCT, a transmission model for induced modulation ripple was proposed. The coupling and propagation mechanisms of the ripple were analyzed, and a transmission function was derived according to the magnetic characteristics of the system. Key factors affecting the suppression of induced modulation ripple were identified, providing a theoretical basis for the design of parameters in closed-loop feedback structures. The results show that, with the model′s guidance, the output ripple at the fundamental frequency of the prototype is attenuated by 53.4 dB, consistent with the theoretical predictions.

Abstract:Temperature measurements inside closed vessels can prevent equipment damage and safety accidents caused by high or low temperatures. Most of the existing studies on temperature measurement inside closed vessels are intrusive and single-point measurements. A method for visualizing the temperature field of a closed vessel based on acoustic tomography is proposed, in which a positive problem model is established by approximating the acoustic slow function distribution through a radial basis function, a full variational regularization term constraint is introduced into the model, and the acoustic inverse problem is solved by an alternating direction multiplier algorithm. At the same time, a finite element simulation model is established to analyze the propagation characteristics of the acoustic wave in the closed container. The simulation results show that the maximum relative error between the simulated propagation time and the theoretical value is 1.80%, and the error between the reconstructed temperature field and the modeled temperature field is within 3%. Finally, an acoustic temperature field experimental test system is constructed and experiments are used to further verify the feasibility of acoustic temperature measurement, and the experimental results show that the technique can reconstruct the temperature field of closed containers with liquids. This study is important for measuring the internal temperature and preventing overheating faults to ensure the stable operation of the equipment.

Abstract:In recent years, with the rapid development of new energy vehicles, 3D object detection, as a core foundation of autonomous driving technology, has become increasingly important. Strategies that integrate multimodal information, such as radar point clouds and images, can significantly enhance the accuracy and robustness of object detection. Inspired by BEVDet, this paper proposes an improved multimodal fusion 3D object detection method based on the BEV (bird′s eye view) perspective. The method employs a ConvNeXt network combined with an FPN-DCN structure to efficiently extract image features and utilizes a deformable cross-attention mechanism to achieve deep fusion of image and point cloud data, thereby further enhancing the detection accuracy of the model. Experiments on the nuScenes autonomous driving dataset demonstrate the superior performance of our model, with an NDS of 64.9% on the test set, significantly outperforming most existing detection methods.

Abstract:Gait recognition of lower limb exoskeleton is a key technology to realize human-machine cooperative control, however, the existing gait recognition methods face the challenges of insufficient efficiency of local feature extraction, weak generalization ability of small samples, and high computational overhead of the model when dealing with one-dimensional time series data. Aiming at the above problems, this paper proposes a hybrid model based on 1D-CNN-SVM, which automatically extracts local features of 1D time-series data by a 1D convolutional neural network (1D-CNN) and realizes highly robust classification under small-sample conditions by using support vector machine (SVM). The experimental results show that the model achieves an overall recognition rate of 99.00% on the customized gait dataset, which is 5.67% and 7.99% higher than the traditional SVM model and the single 1D-CNN model, respectively. In addition, the number of parameters of this model is only 26 156, and the single-sample inference time is as low as 0.06 ms, which is significantly better than the hybrid 1D-CNN-LSTM model. This study provides a solution for gait recognition of lower limb exoskeleton that still combines generalization ability, recognition ability and light weight under small sample conditions.

Abstract:Aiming at the problem that it is difficult to estimate the ground impedance of multi-electrode injected ground current field information transmission in real time, the real-time estimation method of multi-electrode injected ground current field ground impedance is proposed.Firstly, the geodesic equipotential surface around the electrodes is constructed by combining cylindrical and hemispherical surfaces, and the geodesic conductor geometry model of the current field of the multi-electrode injection ground is established according to the distribution of the equipotential surface, so that the ground impedance of the multi-electrode injection ground can be estimated through the integration of the length of the equivalent conductor.Under the four test conditions of mountain, forest, field and grassland, the distance between electrodes is set as 0.14, 0.57, 1 and 1.4 times of the depth of electrodes into the ground, and the current injection experiments are carried out for 4, 6 and 8 electrodes, and 4, 6, 8, 10 and 12 electrodes, and the average value of the relative error of the proposed method is 6.0%, and the experimental results show that the proposed method can meet the real-time ground impedance requirements under different injection conditions, estimation under different injection conditions.

Abstract:The default Horizontal Pod Autoscaler strategy in Kubernetes has limitations due to its inherent response mechanism, leading to scaling delays. To improve resource response performance and resource utilization, this paper introduces an elastic scaling strategy based on time-series resource load prediction. The proposed prediction model, WTT-iTransformer, is specifically designed to forecast cluster resources. It is known that iTransformer excels in long-term sequence prediction and can capture the correlations between multiple variables by embedding variable sequences as tokens. By adding a Wavelet Transform Convolutional layer and integrating a Multi-Scale Temporal Convolutional Network, the WTT-iTransformer model is constructed, enabling more precise extraction of long-term features and dependencies in resource time-series from both the time and frequency domains, which aligns better with the prediction of container usage characteristics. Based on the load variation prediction of this model, rapid scaling can be implemented at the early stages of high and low traffic occurrences, addressing the issues of delayed responses and low resource utilization. Experimental results show that the WTT-iTransformer demonstrates better stability and lower training error during training, accurately predicting cluster load trends. The improved elastic scaling strategy, compared to the traditional HPA in Kubernetes, is more intelligent and stable, showing significant improvements in scenarios with notable load characteristics and frequent burst traffic, thus holding broad application potential.

Abstract:Intelligent integrated excation in coal mine tunnels is a crucial research focus in the global coal mining industry, with significant implications for achieving safe and efficient extraction in deep coal mining operations. The boom-type roadheader, which serves as the primary equipment for underground excavation, currently faces a critical technological challenge: the precise positioning of its key nodes. Aiming to address the low positioning accuracy and poor robustness of roadheaders currently, an improved time difference of arrival (TDOA) fusion positioning algorithm based on regularized constrained total least squares (RCTLS) and alternating direction method of multiplier (ADMM) is proposed. Meanwhile, a positioning system of roadheader in tunnel based on ultra-wide band (UWB) is constructed. Considering the analytical TDOA algorithm easliy falls into the local optima due to the ranging error, a closed-form solution of analytics is selected as the initial value of ADMM, and the objective function is iteratied via dual auxiliary variables to achieve the positioning. The experimental result indicates that the RCTLS-ADMM enhances positioning accuracy by reducing the average positioning error of X, Y and Z axes from 0.159 m, 0.154 m, and 0.167 m to 0.139 m, 0.133 m, and 0.141 m, and improves the positioning accuracy by 12.57%, 13.64% and 15.57% respectively in long-narrow environments if the UWB ranging error is non-negligible. The positioning strategy of roadheader provides the significant parameters for achieving self-dominant control of roadheader and has practical applicated value.

Abstract:To address the challenge of reusing hardware logic in the implementation of a multi-axis stepper motor controller, which leads to excessive consumption of logic resources, a time-division multiplexing strategy grounded in a speed curve algorithm has been proposed. Initially, leveraging the kinematic theory of rigid bodies rotating about a fixed axis alongside the control principles of stepper motors, a mapping relationship between the pulse period of stepper motor control and the corresponding kinematic physical quantities is established. Subsequently, the two rotational modes of uniform acceleration and uniform deceleration are integrated with the pulse calculation formula, optimizing the velocity curve calculation method. Building upon single-axis non-time-division multiplexing control, the design of the time-division multiplexing multi-axis velocity curve algorithm is executed by fully utilizing the time intervals of control pulse outputs. Ultimately, the IP core for the time-division multiplexing controller of the two-axis stepper motor is developed, achieving a 33.68% reduction in logic resource usage and 14.04% reduction in thermal power consumption compared to the two-axis non-time-division multiplexing IP core. A hardware experimental platform is constructed to validate the algorithm, with results indicating that the time-division multiplexing IP core enables precise control of the two-axis stepper motor, maintaining an angular displacement following error within ±8 steps (±0.9°).

Abstract:The assessment of state of health (SOH) of batteries is one of the key technologies in battery systems, and its accuracy is crucial for the safe operation of battery systems. The relaxation voltage curve contains rich battery information and has a short relaxation time, making it suitable for evaluating the state of health of batteries under non constant operating conditions. This article uses the relaxation voltage curve to evaluate the state of health of lithium iron phosphate battery modules. Firstly, a relaxation voltage model for lithium iron phosphate (LFP) battery modules based on linear correlation between time constant and relaxation time was established, and particle swarm optimization (PSO) algorithm was used to identify the parameters of the relaxation voltage curve and extract health factors. Secondly, a hybrid model of convolutional neural networks (CNN) and bidirectional long short-term memory (BiLSTM) optimized based on pelican optimization algorithm (POA) was developed to evaluate the SOH of batteries. The research results show that regardless of whether the relaxation voltage curve is obtained by charging and discharging at 1/2 C rate or 1 C rate, the relative error (RE) between the variable time constant voltage value identified by PSO algorithm parameters and the true relaxation voltage value does not exceed ±0.12%, indicating that PSO method has good parameter identification effect on relaxation voltage at different rates. Using the relaxation voltage curve after charging and discharging at a rate of 1/2 C, the relative error of SOH evaluation obtained using the POA-CNN-BiLSTM model in the test set still does not exceed ±1.2% even when the training set is as low as 5%, At a charge discharge rate of 1 C, when the training set was as low as 5%, the relative error of SOH evaluation obtained using the POA-CNN-BiLSTM model in the test set still did not exceed ±1.5%, indicating that the POA-CNN-BiLSTM model has high accuracy in evaluating battery SOH.

Abstract:In vehicle platoon control, traditional control strategies struggle to simultaneously meet the requirements of system robustness and high-precision tracking, especially when facing external disturbances and model uncertainties, which exacerbates the issue. This paper proposes a fusion algorithm (LFC-MSE) that combines linear feedback control with vehicle motion state estimation to enhance the accuracy of the following vehicle′s speed and angular velocity, thereby mitigating the adverse effects of external disturbances and communication delays. By utilizing feedback linearization, the nonlinear system of vehicle platooning is transformed into a linear system for solution, and a controller for the vehicle platoon system is designed. In terms of communication delay, the motion state of the vehicle is estimated to improve the response speed and control accuracy of the entire system. Finally, in the CarSim-Simulink co-simulation environment, the dynamic model, parameter model, and control model of the vehicle platoon are established to simulate and validate the proposed LFC-MSE algorithm. The simulation results show that under the control of this algorithm, the lateral error of the following vehicle relative to the leading vehicle is within 0.5 m, and the longitudinal trajectory error is within 1.5 m. Moreover, the LFC-MSE control scheme performs better in maintaining platoon stability, response speed, and reducing energy consumption.

Abstract:During the multi-pin docking and assembly process of connectors, if the precise position and orientation of parts cannot be accurately detected, it is very likely to lead to the failure of the connector docking task, and even cause pin bending or damage to the part structure. However, due to the interference of factors such as image noise, machining marks inside holes, defocus caused by hole depth, and hole wall shadows, it is difficult to accurately obtain a complete and precise circular contour when detecting holes, which has a relatively significant negative impact on the accuracy of connector pose detection. In response to the above problems,this paper proposes an improved circle fitting algorithm based on RANSAC. Firstly, coarse positioning is carried out through the Hough gradient method, and then the inlier set of the circle to be fitted is obtained through RANSAC. Finally, the WLS method, which determines weights according to the relative number of inliers in the neighborhood, is proposed to refit the circle for the inlier set, and then the precise pose of the connector is calculated. This paper also conducts a precision test experiment. The measured average attitude error is 0.051°, and the average position error is 0.567 pixels. Compared with other common algorithms, it is finally verified that the algorithm in this paper has high accuracy and stability, and the detection effect is better than that of other algorithms.

Abstract:Due to the phenomenon of light absorption and scattering, underwater imaging often has problems such as detail loss, color deviation, illuminance loss, and overexposure. To solve these problems, an enhancement algorithm based on illumination adaptation and wavelet fusion is proposed in this paper. The overall brightness of the image is improved by using the optimized logarithm transformation, and the enhanced image adapted to the background illuminance is generated by the convolution operation of Gaussian kernel function, and then the underwater image is enhanced by wavelet multi-scale fusion to enhance the low illuminance area of the underwater image and suppress the overexposed area. Secondly, by calculating the mean of the color channels, the contrast and color saturation of the fused image are adjusted. Finally, the images after Gamma correction and sharpening are fused by wavelet iteration to obtain the final underwater enhanced image. Experimental results show that the proposed algorithm can effectively enhance image detail and restore image color difference. The mean values of IE, UCIQE and UIQM of the image are improved by 7.5%, 36.6% and 199.8%, respectively, compared with the original image.

Abstract:Small target detection is an extremely challenging task in computer vision, where existing detection algorithms suffer from high complexity, large computational overhead, and low detection accuracy, leading to issues such as missed detections and false alarms. In this paper, the LDF-YOLO algorithm is proposed to enhance detection accuracy and decrease missed detection rates for small objects. Firstly, improvements are made to the Head section by introducing a feature transformation module in the feature fusion network and designing the LP-Detect detection head tailored for small objects. Secondly, drawing inspiration from residual gated mechanisms and local feature enhancement strategies, the LR-C2f module is devised to bolster the model′s capability in extracting local features. Finally, the local feature enhancement module is integrated to enhance backbone′s ability to extract information from small objects. On the publicly available Tiny Person dataset, LDF-YOLO outperforms the original YOLOv8 by achieving a 4.5% improvement in mAP0.5 and a 5.5% increase in recall. Experimental results validate the effectiveness of our proposed improvements. Furthermore, generalization comparison experiments on the NWPU VHR-10 and VisDrone2019 datasets demonstrate improvements across all metrics.

Abstract:Mainstream word-level lipreading models, based on three-dimensional convolutional neural networks and residual networks, struggle to capture the geometric dynamics of lip movements. Their reliance on pixel-level texture details makes them highly sensitive to noise and facial variations. To address these limitations, this paper proposes an end-to-end word-level lipreading model that integrates pixel-level texture detail features, geometry-level contour shape features, and word boundary features, achieving comprehensive multi-feature fusion across temporal, spatial, pixel-level, and geometric-level dimensions. The proposed model incorporates the spatial and channel squeeze-and-excitation mechanism into 3D CNNs and ResNet-18 to enhance texture feature extraction, while an improved spatial-temporal graph convolutional network integrates a global context network to strengthen global geometric relationships. Additionally, word boundary features further guide the model to focus on relevant temporal frames, reducing noise sensitivity. These features are fused and processed by a back-end temporal module to complete the recognition task. Experiments show that when the input is grayscale video, the accuracy of this paper′s model on the publicly available large-scale word-level lip recognition dataset LRW reaches 89.3%, which is improved by 1.3%~3.9% compared with single or partial feature models under the same conditions, and higher than most existing models, which verifies the validity of the proposed model; at the same time, experiments find that, when colorful video is used as the input, the accuracy of the model further improves to 89.7%, verifying the effect of color information on lip recognition.

Abstract:Flexible capacitive strain sensors, with the characteristics of high flexibility, light weight, low power consumption, and easy integration, have important application values in fields such as health monitoring, human-computer interaction, robotics, and smart textiles. At the present stage, it is difficult for flexible capacitive strain sensors to achieve a linear response of the capacitance signal under wide strain, and the application characteristics in specific scenarios have not received due attention. In response to this, this study proposes a capacitive elastic strain sensor using liquid metal as the electrode layer and ecological flexible rubber as the dielectric layer for respiratory monitoring. Through experimental tests, it is found that the gauge factor of this sensor can reach 1.4 within the strain range of 0%~120%, and the minimum resolution is 0.01 mm. After 1 million stretching cycles of 3%, it can still output the capacitance signal. It can be used for long-term monitoring of respiratory signals with tiny deformations, and it has good washability and ethanol disinfection resistance. It can judge the human body′s movement state and respiratory state according to the changes in respiration, providing a new development direction for intelligent healthcare.

Abstract:A dynamic prediction model of sulfur dioxide concentration is proposed to address the challenge of accurate measurement of sulfur dioxide emission concentration at the exit of a limestone-gypsum wet flue gas desulfurization system under deep peaking.The model integrates a convolutional neural network (CNN), a bidirectional gated recurrent unit (BiGRU), and an attention mechanism to predict the sulfur dioxide concentration. The model developes utilizes kernel principal component analysis to determine seven characteristic variables, which are then used as inputs to the model.The attentional mechanism is combined with CNN and BiGRU to construct a model for predicting SO2concentration at the outlet of the desulfurization system. A simulation experiment is conducted with the FDG system of an operating 600 MW supercritical unit as the research object. The simulation results demonstrate that the average absolute error MSE of the dynamic model established in this paper is 0.706 4 mg/m3, the root mean square error RMSE is 0.912 5 mg/m3, and the average relative error is 6.27%, which is 25.07%, 23.45%, and 17.28% lower compared with CNN-BiGRU, and even lower than CNN and BiGRU; The coefficient of determination of the dynamic model was 96.74%, which was 3.91%, 5.26%, and 9.66% higher than CNN-BiGRU, BiGRU, and CNN models respectively. This outcome indicates that the dynamic model based on CNN-BiGRU-Attention exhibits high prediction accuracy and learning ability, and can accurately predict the trend of SO2 concentration at the outlet of the desulfurization system.