Computer-aided diagnosis (CAD) can detect tuberculosis (TB) cases, providing radiologists with more accurate and efficient diagnostic solutions. Various noise information in TB chest X-ray (CXR) images is a major challenge in this classification task. This study aims to propose a model with high performance in TB CXR image detection named multi-scale input mirror network(MIM-Net) based on CXR image symmetry, which consists of a multi-scale input feature extraction network and mirror loss. The multi-scale image input can enhance feature extraction, while the mirror loss can improve the network performance through self-supervision. We used a publicly available TB CXR image classification dataset to evaluate our proposed method via 5-fold cross-validation, with accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and area under curve (AUC) of 99.67%, 100%, 99.60%, 99.80%, 100%, and 0.999 9, respectively. Compared to other models, MIM-Net performed best in all metrics. Therefore, the proposed MIM-Net can effectively help the network learn more features and can be used to detect TB in CXR images, thus assisting doctors in diagnosing.

Brain tumor segmentation is critical in clinical diagnosis and treatment planning. Existing methods for brain tumor segmentation with missing modalities often struggle when dealing with multiple missing modalities, a common scenario in real-world clinical settings. These methods primarily focus on handling a single missing modality at a time, making them insufficiently robust for the additional complexity encountered with incomplete data containing various missing modality combinations. Additionally, most existing methods rely on single models, which may limit their performance and increase the risk of overfitting the training data. This work proposes a novel method called the ensemble adversarial co-training neural network (EACNet) for accurate brain tumor segmentation from multi-modal magnetic resonance imaging (MRI) scans with multiple missing modalities. The proposed method consists of three key modules: the ensemble of pre-trained models, which captures diverse feature representations from the MRI data by employing an ensemble of pre-trained models; adversarial learning, which leverages a competitive training approach involving two models; a generator model, which creates realistic missing data, while sub-networks acting as discriminators learn to distinguish real data from the generated “fake” data. Co-training framework utilizes the information extracted by the multimodal path (trained on complete scans) to guide the learning process in the path handling missing modalities. The model potentially compensates for missing information through co-training interactions by exploiting the relationships between available modalities and the tumor segmentation task. EACNet was evaluated on the BraTS2018 and BraTS2020 challenge datasets and achieved state-of-the-art and competitive performance respectively. Notably, the segmentation results for the whole tumor (WT) dice similarity coefficient (DSC) reached 89.27%, surpassing the performance of existing methods. The analysis suggests that the ensemble approach offers potential benefits, and the adversarial co-training contributes to the increased robustness and accuracy of EACNet for brain tumor segmentation of MRI scans with missing modalities. The experimental results show that EACNet has promising results for the task of brain tumor segmentation of MRI scans with missing modalities and is a better candidate for real-world clinical applications.

Medical image fusion technology is crucial for improving the detection accuracy and treatment efficiency of diseases, but existing fusion methods have problems such as blurred texture details, low contrast, and inability to fully extract fused image information. Therefore, a multimodal medical image fusion method based on mask optimization and parallel attention mechanism was proposed to address the aforementioned issues. Firstly, it converted the entire image into a binary mask, and constructed a contour feature map to maximize the contour feature information of the image and a triple path network for image texture detail feature extraction and optimization. Secondly, a contrast enhancement module and a detail preservation module were proposed to enhance the overall brightness and texture details of the image. Afterwards, a parallel attention mechanism was constructed using channel features and spatial feature changes to fuse images and enhance the salient information of the fused images. Finally, a decoupling network composed of residual networks was set up to optimize the information between the fused image and the source image so as to reduce information loss in the fused image. Compared with nine high-level methods proposed in recent years, the seven objective evaluation indicators of our method have improved by 6%-31%, indicating that this method can obtain fusion results with clearer texture details, higher contrast, and smaller pixel differences between the fused image and the source image. It is superior to other comparison algorithms in both subjective and objective indicators.

Lactate, as a metabolite, plays a significant role in a number of fields, including medical diagnostics, exercise physiology and food science. Traditional methods for lactate measurement often involve expensive and cumbersome instrumentation. This study developed a portable and low-cost lactate measurement system, including independently detectable hardware circuits and user-friendly embedded software, computer, and smartphone applications. The experiment verified that the relative error of the detection current in the device circuit was less than 1%. The electrochemical performance was measured by comparing the [Fe(CN)₆]^3-/[Fe(CN)₆]^4- solution with the desktop electrochemical workstation CHI660E, and a nearly consistent chronoamperometry(CA) curve was obtained. Two modified lactate sensors were used for CA testing of lactate. Within the concentration range of 0.1 mmol·L^-1 to 20 mmol·L^-1, there was a good linear relationship between lactate concentration and steady-state current, with a correlation coefficient (R²) greater than 0.99 and good repeatability, demonstrating the reliability of the developed device. The lactate measurement system developed in this study not only provides excellent detection performance and reliability, but also achieves portability and low cost, providing a new solution for lactate measurement.

In order to ensure the uninterrupted communication between high-speed train and base station, driving safety and satisfying online experience of passengers, a dual-link switching algorithm based on CNN-WaveNet decision parameter multi-step prediction model is proposed to establish a two-hop relay communication system model between the high-speed train and the base station. Firstly, the switching algorithm uses convolution neural network (CNN) to extract the time sequence characteristics of decision parameters. Then, it learns the mapping relationship between feature information and decision parameters based on WaveNet and combining with rolling prediction method to realize multi-step prediction of decision parameters. Finally,dual-antenna communication mode is adopted to realize dual-link communication. The simulation results show that the proposed handover algorithm can improve handover trigger rate and handover success rate.

To address the issue of low measurement accuracy caused by noise interference in the acquisition of low fluid flow rate signals with ultrasonic Doppler flow meters, a novel signal processing algorithm that combines ensemble empirical mode decomposition (EEMD) and cross-correlation algorithm was proposed . Firstly, a fast Fourier transform (FFT) spectrum analysis was utilized to ascertain the frequency range of the signal. Secondly, data acquisition was conducted at an appropriate sampling frequency, and the acquired Doppler flow rate signal was then decomposed into a series of intrinsic mode functions (IMFs) by EEMD. Subsequently, these decomposed IMFs were recombined based on their energy entropy, and then the noise of the recombined Doppler flow rate signal was removed by cross-correlation filtering. Finally, an ideal ultrasonic Doppler flow rate signal was extracted. Simulation and experimental verification show that the proposed Doppler flow signal processing method can effectively enhance the signal-to-noise ratio (SNR) and extend the lower limit of measurement of the ultrasonic Doppler flow meter.

To solve the problems of redundant feature information, the insignificant difference in feature representation, and low recognition accuracy of the fine-grained image, based on the ResNeXt50 model, an MSFResNet network model is proposed by fusing multi-scale feature information. Firstly, a multi-scale feature extraction module is designed to obtain multi-scale information on feature images by using different scales of convolution kernels. Meanwhile, the channel attention mechanism is used to increase the global information acquisition of the network. Secondly, the feature images processed by the multi-scale feature extraction module are fused with the deep feature images through short links to guide the full learning of the network, thus reducing the loss of texture details of the deep network feature images, and improving network generalization ability and recognition accuracy. Finally, the validity of the MSFResNet model is verified using public datasets and applied to wild mushroom identification. Experimental results show that compared with ResNeXt50 network model, the accuracy of the MSFResNet model is improved by 6.01% on the FGVC-Aircraft common dataset. It achieves 99.13% classification accuracy on the wild mushroom dataset, which is 0.47% higher than ResNeXt50. Furthermore, the experimental results of the thermal map show that the MSFResNet model significantly reduces the interference of background information, making the network focus on the location of the main body of wild mushroom, which can effectively improve the accuracy of wild mushroom identification.

Photographs taken in daily life often became blurred due to shaking, out-of-focus, changes in depth of field, and movement of photographed objects. Aiming at this problem, a double-channel cyclic image deblurring method based on edge features was proposed. Firstly, image edge gradient operator was introduced as a threshold based on the rule that the maximum value of the image edge gradient will decrease after the blurring process, making the blurred image be divided into two channels: edge channel and non-edge channel. Secondly, a double-channel loop iteration network was designed, where the edge gradient was used in the edge channel to sample the main edge structure and bilateral filtering was used in the non-edge channel to extract the detailed texture feature information. Finally, the feature information extracted from two channels was cyclically iterated to obtain a clear image using the deblurring model with maximum a posteriori probability. The experimental results showed that the image evaluation indexes obtained by the proposed deblurring model were superior to those of other algorithms, and the edge structure and texture details of the image were effectively recovered with better performance.

Considering that the algorithm accuracy of the traditional sparse representation models is not high under the influence of multiple complex environmental factors, this study focuses on the improvement of feature extraction and model construction. Firstly, the convolutional neural network(CNN) features of the face are extracted by the trained deep learning network. Next, the steady-state and dynamic classifiers for face recognition are constructed based on the CNN features and Haar features respectively, with two-stage sparse representation introduced in the process of constructing the steady-state classifier and the feature templates with high reliability are dynamically selected as alternative templates from the sparse representation template dictionary constructed using the CNN features. Finally, the results of face recognition are given based on the classification results of the steady-state classifier and the dynamic classifier together. Based on this, the feature weights of the steady-state classifier template are adjusted in real time and the dictionary set is dynamically updated to reduce the probability of irrelevant features entering the dictionary set. The average recognition accuracy of this method is 94.45% on the CMU PIE face database and 96.58% on the AR face database, which is significantly improved compared with that of the traditional face recognition methods.

A control strategy of repetitive control without inductorance decoupling was proposed to address the problem of high total harmonic distortion(THD) rate of the network-side current caused by the reduced stability of the rectifier module of the DC charging pile under weak grid as well as the dead zone and nonlinearity of switching devices during charging. Firstly, the parallel repetitive control was constructed in the inner current loop, and the proportional-integral(PI)+ repetitive controller based on parallel structure was designed. For system compensation, a second-order low-pass filter was selected to correct the system, and the network-side current harmonics were actively suppressed without increasing the filtering device, which effectively improves the quality of grid-connected current. Secondly, based on the synthetic vector method, the controller parameters were designed to realize the elimination of main pole by establishing two synchronous rotation coordinate system vector differential equations, so as to realize the inductanceless decoupling to cope with the influence of network-side inductance fluctuation on the stability of the control system under weak grid. By theoretical analysis and simulation, the proposed control strategy was embedded into the self-developed digital signal processor for the rectifier module of DC charging pile, simulated dynamic and steady-state operation experiments were conducted, and comparative analysis was performed to prove the feasibility of the proposed control strategy.

With the continuous evolution of electronic technology, field-programmable gate array (FPGA) has demonstrated significant advantages in the realm of signal acquisition and processing, and signal acquisition plays a pivotal role in the practical applications of laser gyros. By analysis of the output signals from a laser gyro and an accelerometer, this paper presents a circuit design for signal acquisition of the laser gyro based on domestic devices. The design incorporates a finite impulse response (FIR) filter to process the gyro signal and employs a small-volume, impact-resistant quartz flexible accelerometer for signal aquisition. Simulation results demonstrate that the errors in X, Y, and Z axes fall within acceptable ranges while meeting filtering requirements. The use of FPGA for signal acquisition and preprocessing enhances configuration flexibility, which provides an idea and method for optimizing performance and processing signals in laser gyro applications.

In complex industrial scenes, it is difficult to acquire high-precision non-cooperative target pose under monocular visual servo control. This paper presents a new method of target extraction and high-precision edge fitting for the wheel of the sintering trolley in steel production, which fuses multiple target extraction algorithms adapting to the working environment of the target. Firstly, based on obvious difference between the pixels of the target image and the non-target image in the gray histogram, these pixels were classified and then segmented in intraclass, removing interference factors and remaining the target image. Then, multiple segmentation results were merged and a final target image was obtained after small connected regions were eliminated. In the edge fitting stage, the edge fitting method with best-circumscribed rectangle was proposed to accurately fit the circular target edge. Finally, PnP algorithm was adopted for pose measurement of the target. The experimental results showed that the average estimation error of pose angle γ with respect to the z-axis rotation was 0.234 6°, the average measurement error of pose angle α with respect to the x-axis rotation was 0.170 3°, and the average measurement error of pose angle β with respect to the y-axis rotation was 0.227 5°. The proposed method has practical application value.

Aiming at the problems of difficult deployment and access of surveillance system server, as well as high operation and maintenance cost, a remote surveillance camera is designed based on RK3566 chip, which is controlled and transmits data via email platform. Firstly, to address the impact of environmental factors such as weather and light on image quality, a deep neural network (DNN) image exposure correction network is employed to rectify images with abnormal exposure. Additionally, a back propagation (BP) neural network is utilized to fit a curve relating the brightness difference to the gamma value of images before and after exposure correction, thereby adjusting the gamma value of the camera. Secondly, to enhance the precision of YOLOv5 algorithm in differentiating between anomalies in nighttime imagery, infrared image data are employed, and a context-aware light-weight label assignment head and coordinate attention mechanism are incorporated into the model to augment the model’s detection accuracy and recall rate for small targets. Furthermore, to meet the demand for reporting of abnormal situations in unattended environments, an automatic target identification and reporting process has been designed which combines YOLOv5 algorithm with the frame-difference motion detection algorithm. The camera has been tested for compatibility with the current mainstream commercial email platforms. The mean time required for transmitting a single image file via the email platform is less than 10 s, while the mean time for transmitting a short video is less than 60 s. The BP network’s average training loss is 0.015, and the average testing loss is 0.013, which basically meets the precision requirements for gamma adjustment. The improved YOLOv5 algorithm achieved an mAP@0.5 of 91.5% and a recall rate of 85.5%, effectively enhancing the accuracy of small object detection.

The purpose of this study is to analyze the galloping characteristics of the catenary positive feeder in fluctuating wind areas considering dynamic-wind angle of attack and aerodynamic damping. Firstly, the flow field model of the catenary positive feeder was established, the fluctuating wind field was simulated by Davenport wind power spectrum and linear filtering method, and the wind speed at inlet in calculation domain was controlled by editing the profile file to simulate and calculate the aerodynamic characteristics of the positive feeder in the fluctuating wind area. Then, taking the positive feeder as the research object, the mathematical model of actual structure and the corresponding finite element model were established. By applying the wind load to the finite element model, the influence of aerodynamic damping caused by the self-movement of the positive feeder on the galloping response was analyzed, and the frequency domain characteristics of galloping displacement of the positive feeder considering aerodynamic damping were studied. Finally, the calculation method of aerodynamic damping by the Guidelines for Electrical Transmission Line Structural Loading (ASCE No.74) was used for the galloping response of the positive feeder and compared with the proposed method. The results show that when considering aerodynamic damping, the galloping amplitude of the positive feeder decreases significantly, and the first-order resonance effect on the vertical displacement and horizontal displacement decreases significantly. The galloping trajectories calculated by the two methods are consistent. Therefore, this study is of great significance to further clarify the ice-free galloping mechanism of the catenary positive feeder in violent wind areas.

The application of lithium metal anodes is hindered by low Coulombic efficiency (CE), serious lithium dendrites and volume expansion. An MnO/Polypropylene (PP) composite separator was developed to regulate lithium metal deposition behaviors through in situ forming stable artificial solid electrolyte interface (SEI) passivating layers. The concentration of MnO in the cells can be maintained at a constant based on quite low solubility of MnO in the liquid electrolyte, and the dissolved MnO can be reduced to produce Li₂O and Mn metal nanoparticles, which can not only function as nucleating seeds of lithium metal deposits but also repair the broken SEI layer. Dendritic-free Li deposits can be obtained by simple separator coating. It can also improve the electrochemical performance of lithium metal batteries. And it is benefit for applications of Li metal anodes.