In order to solve the problem of low sensitivity in low-frequency vibration monitoring of bridges,a low-frequency and high-sensitivity fiber Bragg grating (FBG) acceleration sensor is proposed based on an equal strength cantilever beam.Twelve groups of FBG acceleration sensors are designed with the thickness of the cantilever beam and the weight of the mass block as the main parameters.The amplitude-frequency characteristics,sensitivity,and transverse anti-interference tests of the sensors are carried out by a vibration generator.And a set of sensors suitable for monitoring bridge vibration are obtained.The natural frequency of the sensor is 49 Hz,the operating band is 0—34 Hz,the sensitivity is up to 664.53 pm/g,the linearity is 99.9%,and the lateral anti-interference ability is strong.A novel and effective monitoring method is provided for bridge vibration by theoretical derivation and experimental verification.

The emission wavelength of semiconductor laser varies with the change of working current and temperature,which affects the effective linewidth and wavelength stability of the output laser,and cannot meet the requirements of the gain medium in the solid-state laser for the pump source wavelength and linewidth.In this paper,a self-developed reflective holographic volume Bragg grating (VBG) of 878 nm with a diffraction efficiency of 9.9% is designed.By using the VBG as a reflective cavity mirror,the laser emission wavelength can be locked around the designed 878 nm and an output linewidth is only 0.3 nm.The wavelength-current drift coefficient is 0.015 nm/A and the temperature drift coefficient is 0.007 5 nm/°C.An all-solid-state laser is constructed by using a wavelength locking semiconductor laser as the pump source.The self-developed 1 064 nm VBG with a diffraction efficiency of 98.71% and 94.32% is used as the front and rear cavity mirrors, respectively.A Nd∶YVO4 crystal with a doping concentration of 0.3% is used as the gain medium.After the spatial optical path tuning,a continuous and stable laser output with a central wavelength of 1 064.2 nm and a line width of 0.29 nm are obtained.

To solve the problem of the excessive cavity length demodulation error caused by the inability to determine the interference order due to the peak-to-peak method of demodulating the fiber-optic Fabry-Perot (FP) sensor.A peak-to-peak and interference order cavity length demodulation algorithm is proposed.The algorithm accurately determines the interference order of the selected peak by tracking the two farthest peaks in the reflection spectrum of the sensor,and then accurately solves the absolute cavity length according to the single-peak method.The accuracy of the cavity length solution is significantly improved by effectively eliminating the ambiguity of the interference order of the peak-to-peak method. Simulation analysis and experimental verification are carried out for FP sensors with cavity lengths of 30—100 μm.The simulated demodulation error is less than 0.7 nm,the experimental error is less than 1.4 nm,and the demodulation error is much smaller than that of the peak-to-peak method.

For light-weight low level light intensifying network,blurred image issue caused by inconsistent light intensifying degree in different area can occur when Zero-DCE handles the low level light image with a bigger brightness variation range.This paper introduces a self-adaptive loss function based on γ transform,on the basis of the original loss function,decreases the sensitivity of the network on image exposure difference and dramatically improves the low level light intensifying effect.In this method,CBAM module is added into the convolutional neural network (CNN) to increase the expression ability of the network to low level light image feature,in addition,the logarithm distance between the average value of gray level of the network intensifying image and the average value of intensifying feature image is selected as γ transformed self-adaptive factor,and finally,the gray level parameter distance between network intensifying image and γ transformed image is calculated.The experiment shows that the performance of this method is dramatically improved comparing to the original network,in which in aspect of image evaluation index,the error mean square is increased by 9.7%,the peak signal to noise ratio is increased by 13.8%,and the structure similarity is increased by 6.7%.

Hyperspectral image classification methods based on the classical convolutional neural network (CNN) have some problems,such as insufficient expression of key detail features and a large number of samples for training.Aiming at these problems,this paper proposes a hyperspectral image classification model with multi-scale features and dual-attention mechanism.Firstly,using 3D convolution,the spatial-spectral features of images can be directly extracted,and transposed convolution is adopted to get more details of the feature map.Then,a feature extraction module is built through convolution kernels of different sizes to achieve multi-scale feature fusion under different receptive fields.Finally,the dual-attention mechanism is designed to suppress the confused regional features and highlight the distinguishing features.The experimental results on two hyperspectral images show that when 10% and 0.5% samples are randomly selected as training samples for each class of ground object,the overall classification accuracy of the proposed model is improved to 99.44% and 98.86%,respectively.This model can obtain higher classification accuracy than some mainstream deep-learning classification models.Since the model can focus on more important detailed features during feature extraction,the classification effect is improved.

Facing the trend of miniaturization, multilayer,and high integration of print circuit board (PCB),to address the problems of missed detection,difficult feature extraction,high false detection rate,and poor detection performance of current PCB defect detection methods,this paper proposes a PCB small target defect detection method based on the improved YOLOv5 algorithm.It first uses the density-based spatial clustering of applications with noise (DBSCAN)+dichotomous K-means clustering algorithm for PCB small target defect characteristics to find a more suitable anchor frame.It then improves the feature extraction layer,feature fusion layer,and feature detection layer of the YOLOv5 network to enhance the extraction of key information and strengthen the fusion of deep and shallow information.This reduces the false and missed detection rate of PCB defects to improve the detection performance of the network.Finally,relevant comparative experiments are conducted on the publicly available PCB dataset.The results show that the improved model has an average accuracy (mAP) of 99.5% and a detection speed of 0.016 s.Compared with the Faster R-CNN, YOLOv3,and YOLOv4 network models,the detection accuracy is improved by 17.8%,9.7% and 5.3%,respectively,and the detection speed is improved by 0.846 s,0.120 s and 0.011 s,respectively,which satisfies the requirements of high precision and high-speed detection of PCB defects in actual industrial production sites.

To solve the problem that the location of broken pipes in a factory cannot be accurately determined by machine vision,a pipeline edge detection method based on improved Canny operator with adaptive threshold segmentation is proposed.The method processes the acquired images in terms of filtering method,gradient direction and threshold segmentation.Firstly,sampling-adaptive median filtering+bilateral filtering is used instead of Gaussian filtering in the traditional Canny operator to reduce the loss of image edge information and remove the noise in the image.Then,the gradient amplitude is calculated to detect the edge information in different directions.Finally,to avoid the ineffective manual selection of thresholds,the OTSU threshold segmentation algorithm is used for adaptive selection of thresholds.Experiments show that the method improves the image signal-to-noise ratio by 28.22%,the number of edge points by 39.97%,the number of four-connection channels by 11.52% and the number of eight-connection channels by 5.92% compared to the conventional Canny operator.The extracted features are complete and have good continuity, enabling effective detection of breakage in pipeline images.

To solve the problem of high trash content and poor quality of the output cotton in the cleaning and combing process,a local motion impurity ratio control optimization method is proposed by combining the improved Gaussian mixed model (GMM) and frame difference method.Firstly,the principle of the cleaning machines and the characteristics of the trash is analyzed cotton.Secondly,by extracting the key frames of the video,and combining the improved GMM and the frame difference method,the target is accurately extracted by the "with" operation of the image sequence,and then by designing the GMM classifier to obtain the cotton impurity rate for analysis.Finally,it is compared with the traditional detection algorithm for validation.Experiments show that the improved algorithm is better than the traditional algorithm in terms of effectiveness as well as practicality.At the same time,it can meet industrial high precision and real-time requirements by introducing closed-loop control.

To solve the problems of installation errors,time and labor consuming of manual detection,an image recognition model for single thermal battery defects based on transfer learning and convolutional neural network (CNN) is proposed.First,the images of the dataset are preprocessed by cropping and adding noise,etc.The visual geometry group 16 (VGG16) network is used as the backbone architecture of the model,and a selective kernel (SK) convolution is used after the bottleneck layer.Then,global average pooling (GAP) layer and Dropout layer are added,and L2 regularization and other fine-tuning operations are also added,an defect recognition model Q-VGGNet for single thermal battery is got.Finally,pre-training learning is performed on the dataset ImageNet,and the learned weight parameters are transferred to the model Q-VGGNet.The testing results show that the overall recognition accuracy of the six net-work models for the defect images on the dataset can reach 98.39%,94.44%,97.27%,96.34%,93.71% and 95.61%,respectively.The recognition accuracy rates of the Q-VGGNet network model for qualified images and the three types of defective images (negative electrode missing,tab broken,and current plate missing) can reach 99.6%,95.9%,99.6% and 98.4%,respectively.The results show that this method can detect thermal battery defects more accurately and quickly, and has good defect diagnosis ability.The accuracy is improved nearly 3% higher than the traditional method,and a good solution for manual detection of single thermal battery defects is provided.

Aiming at the problem of person re-identification (person re-ID) caused by complex background and object occlusion,angle transformation and pedestrian posture change in real environment,a person re-identification model based on efficient channel attention (ECA) and poly-scale convolution (PSConv) is designed.Firstly,the residual network is used to extract the global features,and a feature fusion module based on PSConv and ECA is added at the end of the network.The global features are fuzed with the global features extracted from the module to get a new global feature,and then the new global feature is segmented to obtain local features.Finally,the new global feature and the local feature are fused to get the final feature, and the loss function is calculated.The experiment is verified on Market1501 and DukeMTMC-reID data set.Rank-1 and mean average precision reach 94.3 % and 85.2 % respectively on Market1501 data set,and 86.3 % and 75.4 % respectively on DukeMTMC-reID data set.The results show that the model can deal with the complex situation in the actual environment,enhance the discrimination of pedestrian features,and effectively improve the accuracy and precision of pedestrian recognition.

Water-jet guided laser machining technology has the advantages of low thermal effect,high precision and no tool damage,and has been applied to the machining of high-precision components. In order to reduce the diameter of laser focusing spot to facilitate the accurate coupling between laser and water-jet and improve the quality of laser hole,a method of reducing the diameter of laser focusing spot based on the principle of double-slit interference is proposed.The geometric and mathematical models of traditional water-jet guided laser and double-slit interference water-jet guided laser drilling are constructed. Based on COMSOL simulation,the effects of traditional and double-slit interference water-jet guided laser drilling on the evolution of laser hole profile,the distribution of heat affected area and the size of recast layer are compared and analyzed under the same laser beam parameters and processing conditions.The results show that compared with traditional water-jet guided laser,the double-slit water-jet guided laser processing technology can effectively reduce the diameter of laser hole,the cone angle,the width of heat-affected area and the size of recasting layer,and further reduce the diameter of laser focusing spot under the existing conditions.

In the face of a large number of consensus nodes in the food alliance blockchain network,due to the low efficiency of the traditional practical Byzantine fault tolerance (PBFT) consensus algorithm,communication energy consumption is too high,which greatly increases the risk of information leakage and data fraud.To solve the above problems,this paper proposes a PBFT optimization consensus algorithm based on agglomerative hierarchical clustering (AHC).Firstly,the AHC algorithm is used to classify and cluster all the consensus nodes. Secondly,PBFT consensus occurs in all clusters in parallel.Finally, the message agreement is reached through the consensus of the master nodes in the cluster.The experimental results show that the improved algorithm can effectively reduce the energy cost and improve consensus efficiency and throughput.

In order to solve the problem that the detection efficiency of quantum states at the receiving end is low when phase coded continuous variable quantum key distribution (CVQKD) is transmitted over a long distance,this paper proposes to use naive Bayes classifier at the receiving end to improve the system performance.NB classifier first trains quantum states of labeled categories,learns the distribution of quantum states of different categories,calculates the prior probability and likelihood probability of each category,and then calculates the posterior probability of substates to be measured belonging to each category based on the prior probability and likelihood probability,and determines the category of substates to be measured according to the magnitude of the posterior probability.The simulation results show that the improved scheme can improve the system performance by reducing the probability of the test state being incorrectly measured at the receiving end.When the excess noise is 0.01,the safety distance of the improved scheme can exceed 250 km.