27results about How to "Few parameters" patented technology

The application relates to the technical field of artificial intelligence optical character recognition, in particular to an end-to-end communication big data journey card recognition method, and the steps of the method comprise the following steps: adjusting and correcting a text box according to size information of the text box, so that the size of the text box meets the required angle and proportion of adjustment; performing separated feature extraction on the corrected text box through a preset neural network to obtain feature data of a font; and obtaining text data corresponding to the feature data through a Bi-GRU recurrent neural network combined with an activation function.

This invention relates to a practical laser spot localization method for array detectors based on radial basis function (RBF) neural networks, belonging to the field of detector technology. It utilizes RBF neural networks to achieve high-precision localization of laser spots. The RBF neural network is designed and trained, using the training set as the input layer and the output layer as the predicted spot position. The parameters and number of neurons are continuously trained and adjusted. The test set is input into the trained neural network to predict the spot position, and the effectiveness of the RBF neural network is tested. This invention has fewer parameters, lower computational load, higher accuracy, and is applicable to n*n multi-module array detectors. Furthermore, the experimental acquisition process in this application uses a quadrant detector, making the operation relatively simple.

The application discloses a non-uniformity correction method for a refrigeration infrared thermal imager in a wide temperature range. The wide temperature range is divided into different temperature subranges, and then imaging at different temperatures is carried out in the temperature subranges with the same integration time, so that the linearity of the images in the temperature subranges is ensured, and the images are made uniform in the temperature subranges through correction parameter modification. The application solves the problems of poor non-uniformity and poor image output quality of the refrigeration infrared thermal imager with temperature change, and has a simple correction mathematical model, less parameters and easy engineering implementation, thereby providing better support for improving the temperature resolution and image quality of the infrared thermal imager.

The present application relates to a kind of intelligent car light weight obstacle target detection method, comprising (1) collecting the original image of obstacle on road and making data set;(2) based on improved Yolov4, construct light weight network model;(3) build intelligent car target detection experimental environment, and utilize data set to carry out multiple training to light weight network model, obtain multiple network model training weights;(4) obtain the best network model training weight, and transplant to intelligent car target detection experimental environment;(5) utilize depth camera component to obtain the information outside intelligent car, and utilize the best network model training weight to detect and identify the obstacle in front of intelligent car.The present application enhances the identification efficiency of intelligent car by improving algorithm, aims at solving the problems, such as the real-time of detecting obstacle and low detection accuracy, due to the low algorithmic power of vehicle-mounted chip of intelligent car.

A base classification method comprises: acquiring a fluorescence image to be identified; identifying a position of each DNA nanoball in the fluorescence image; extracting a brightness data feature of the fluorescence image, the brightness data feature comprising brightness data of M dimensions corresponding to the position of each DNA nanoball; and inputting the extracted brightness data feature into a preset base identification model to obtain a base category corresponding to the fluorescence image. The application also provides a gene sequencer and a computer readable storage medium. The application can realize end-to-end classification from brightness data of a fluorescence image to a base category.

This invention discloses a fractional-order numerical analysis method and module for the deformation behavior of high arch dams, relating to the fields of service behavior analysis and computational mechanics technology for hydraulic structures. By constructing fractional-order characterization methods for decaying rheology, steady-state rheology, and accelerated rheology under three-dimensional stress states, fractional-order rheological mechanical element models of the dam body and foundation of high arch dams are established; using the incremental finite element method, based on t n The stress, viscoelastic strain, viscoplastic strain, and deformation at time t are calculated using the predifference formula. n The stress increment, viscoelastic strain increment, viscoplastic strain increment, and deformation increment within the time period are used to derive t. n+1 By repeatedly calculating the stress, strain, and deformation at specific moments, the stress, strain, and deformation of the high arch dam within the analysis period are obtained. This invention not only objectively characterizes the nonlinear rheological effects of the dam body and foundation of the high arch dam, but also achieves scientific numerical quantitative analysis of deformation behavior, providing technical support for safety monitoring and early warning of potential hazards in in-service high arch dams, and has broad application prospects and market potential.

This invention discloses a medical image segmentation method based on dynamic deformable convolution and a sliding window adaptive complementary attention mechanism, which improves the perception of small lesions and highly deformed targets in medical images, as well as the ability to distinguish between segmented targets and the background. Dynamic deformable convolution, through task-adaptive learning, can flexibly change weight coefficients and deformation biases, enhancing the ability to express local image features and achieving adaptive extraction of spatial features. The sliding window adaptive complementary attention mechanism achieves cross-dimensional global modeling of medical images through a self-attention branch of adaptively learned weight coefficients. This mechanism overcomes the shortcomings of conventional methods in modeling cross-dimensional relationships between space and channels, and can capture long-distance cross-dimensional correlation features in images. Furthermore, the parallel interactive approach combines local and global features at different resolutions to enhance representation learning, maximizing the preservation of both local and global features in medical images.

This invention discloses a PolSAR image classification method and related apparatus based on a complex convolutional Kolmogorov-Arnold network, belonging to the field of land cover classification technology. The method includes: acquiring a PolSAR image to be processed and performing preprocessing; inputting the preprocessed PolSAR image into a preset complex convolutional Kolmogorov-Arnold network to obtain a classification result; wherein the preset complex convolutional Kolmogorov-Arnold network includes complex KAN convolutional layers, multi-branch complex KAN convolutional blocks, and CV-PolyLoss. This invention has promising prospects for practical remote sensing application deployment.

The application discloses a kind of knowledge distillation methods, devices and equipment of mask autoencoder and storage medium, the method is by respectively establishing the teacher model and student model of mask autoencoder, wherein the teacher model and the student model are all visual transformation model, and the size of the teacher model is greater than the student model;The teacher model is pre-trained;The student model is pre-trained based on the pre-trained teacher model, so that the student model learns data generalization ability from the pre-trained teacher model, and obtains better image features representing ability;The student model is fine-tuned based on the pre-trained downstream task training, and the student model can be deployed in the power edge side lacking of computing resources, while ensuring that the model accuracy does not decrease, the model parameter is reduced, and the real-time inference speed is accelerated.

This invention relates to the field of weld defect type identification technology. Existing weld defect detection methods fail to achieve intelligent and efficient detection due to the diverse types of weld defects. This invention provides a low-cost self-learning neural network design method for ultrasonic detection of weld defects. It performs high-dimensional spatial domain feature representation on the initial one-dimensional ultrasonic signal, enriches the feature expression of weld defect data, selects the feature domain with better performance, constructs an adaptive scaling network for weld defect detection, and uses a multi-objective, training-free network search and evaluation method to iteratively search and evaluate candidate networks for weld defect detection, obtaining a balanced solution on both network classification accuracy and parameter quantity. Ultimately, it achieves the selection of a weld defect type detection network with superior overall performance without any training.

The invention relates to a rice bacterial leaf blight severity estimation method based on a trigonometric vegetation index, and the method comprises the steps: obtaining an unmanned plane multispectral image and a hyperspectral satellite image of a target region, and carrying out the preprocessing; performing rice bacterial leaf blight disease classification on the field scale, and determining an optimal classification threshold value of disease classification; calculating a disease distribution severity index RBLBI in each satellite pixel; screening out a plurality of characteristic wave bands sensitive to the rice bacterial blight disease; and on the basis of the screened characteristic wave bands, constructing a trigonometric vegetation index BLBTVI. The method has the beneficial effects that the model provided by the invention is light in weight, few in parameters, insensitive to atmospheric residual error, illumination variation and BRDF influence, and higher in robustness and interpretability; after conventional atmospheric correction and cloud / water body masking are completed, a continuous grid and time sequence monitoring result can be directly generated on a satellite-borne hyperspectral product, and the manual patrol and data processing cost is remarkably reduced.

The application discloses a kind of high similarity electric power tool identification method and system based on deep learning, belong to target detection technical field, its method includes, obtains high similarity electric power tool image dataset and carries out pretreatment;The image dataset after pretreatment is input to improved YOLOv8 model and is trained;The improved YOLOv8 model includes sequentially connected main network, neck network and detection head, wherein the detection head is used to receive the feature map of different scales output by neck network, and utilizes shared convolution layer output detection and classification result;The improved YOLOv8 model trained is used to detect the electric power tool image to be detected by the above-mentioned method and system, and the target detection efficiency and precision are improved.

This invention discloses a method, system, and electronic device for detecting road defects in ground-penetrating radar (GPR) images, belonging to the field of computer vision and road defect detection technology. The method includes: acquiring and preprocessing GPR images; constructing a defect detection model, which includes a multi-scale vertical / horizontal strip convolution module, an adaptive gated routing module, and a bidirectional attention fusion module, used for extracting directional features, dynamically fusing multi-scale features, and adaptively fusing directional features, respectively; training and using the model for defect identification and localization. This invention reduces computational complexity through strip convolution and achieves adaptive feature fusion through gating and attention mechanisms, significantly improving the detection accuracy and efficiency of multi-scale, directional defects in GPR images.

The application discloses a gravity and magnetic inversion method based on grid merging and BTTB matrix fast forward, relates to the technical field of gravity and magnetic exploration, and comprises the following steps: establishing a basic forward and inversion grid and a basic grid kernel function matrix; initializing a coarse grid model and a conversion matrix; coarse grid inversion; adaptive grid subdivision based on a model gradient and establishment of a new conversion matrix; fine grid inversion; iterative convergence and output. The application has the beneficial effects that the BTTB kernel function matrix is calculated based on the basic grid in advance, the conversion from fine scale model parameters to coarse scale model parameters is realized through the conversion matrix, the iterative closed loop process of coarse grid inversion, adaptive grid subdivision and fine grid inversion is adopted, the criterion of adaptive grid subdivision is to compare the gradient eigenvalue of the current inversion model parameters with a preset threshold value, and in the grid inversion of different scales, the BTTB kernel function matrix generated by the same set of basic grid in advance is used to realize fast forward through the processing of the conversion matrix.

The application discloses a lightweight dual-prediction branch semantic segmentation water body extraction deep learning method and system, first, the coding part of the lightweight dual-prediction branch semantic segmentation model is constructed, three convolution blocks in the coding structure are designed based on the visual perception, and high, medium and low resolution feature maps are extracted respectively; secondly, the decoding part of the lightweight dual-prediction branch semantic segmentation model is constructed, the dual-prediction branch is designed, the medium and low resolution feature maps are fused and used as one of the discriminant features, input into one of the prediction branches, the high resolution feature map is input into the other prediction branch, and the maximum confidence of the two prediction branches is used as the final prediction probability; finally, the model is trained by using a satellite remote sensing image labeled data set and the adaptive learning rate. The deep learning model provided in the application can effectively solve the water body extraction in a large range of satellite remote sensing images, and provide an intelligent method for dynamic change monitoring of water resources.

The application relates to an effluent BOD concentration prediction method based on a WSFA-AFE ILSTM neural network and relates to the field of artificial intelligence.The application proposes a WSFA-AFE method aiming at the problem that the input characteristic variable and input history step length are difficult to determine when a neural network is used to predict a multivariate time sequence of effluent BOD.The method can adaptively extract dynamic characteristic variables in the multivariate time sequence, so that the neural network can better predict the effluent BOD concentration.The application proposes an ILSTM neural network aiming at the problems that the standard LSTM neural network has a large number of structure parameters and the training process is time-consuming.The application simplifies the recursive term weight in the structure equation, reduces the number of required training parameters in the network, and accelerates the convergence speed through a parameter updating algorithm.The application realizes efficient, accurate and low-cost prediction of effluent BOD concentration at future time according to the data collected in the sewage treatment process.

The application provides a method and system for constructing a large power grid frequency response model considering voltage characteristics. On the basis of a classical system frequency response model, the influence of load voltage characteristics on power system frequency is considered, and a frequency response model capable of reflecting the joint influence of prime movers and governors and loads is constructed. Through model analysis, the calculable part of the model parameters is determined, and the number of parameters to be identified is reduced. Finally, according to the measured large power grid data, all the remaining parameters of the frequency response model considering voltage characteristics are identified based on a particle swarm algorithm. The application can solve the problem that the classical system frequency response model does not work well when voltage drops, accurately simulate the frequency drop process of a large power grid, and improve the accuracy and applicability of the classical frequency response model. The model can quickly and accurately calculate the dynamic response process of the system frequency of a power system under different power disturbances, which is helpful for frequency safety analysis and control of a large power grid.

The invention discloses a point cloud registration method and system based on multi-source semantic information and a plane normal vector. The point cloud registration method comprises the following steps: acquiring first point cloud data of one site and second point cloud data of another site in a target area; semantic segmentation is carried out on the first point cloud data and the second point cloud data, and segmented components are classified; for a component of a target category, extracting a plane component in the target category; acquiring multi-dimensional feature data of the planar component; performing plane component matching of the first point cloud data and the second point cloud data by using the multi-dimensional feature data; calculating a rotation matrix by using the matching component; and performing global multi-plane consistency evaluation on the rotation matrix to obtain an initial registration matrix. The method has the unification of high registration success rate, high robustness and relatively high calculation efficiency, so that a reliable initial pose is provided for point cloud precise registration and multi-site fusion in a complex city scene.

This invention provides a method for rapidly evaluating the storage capacity of a gas storage facility converted from a constant-pressure water body. The method includes: Step 1, calculating the effective trap size and pressurization coefficient based on a geological assessment of the effective trap; Step 2, calculating the available gas storage capacity from gas injection and water displacement based on the assessment results of the effective trap size; Step 3, calculating the available gas storage capacity from gas injection and compression of rock and residual water based on the assessment results of the effective trap pressurization coefficient; and Step 4, calculating the storage capacity of the gas storage facility converted from the constant-pressure water body. This method provides highly accurate calculation results for the gas storage capacity of such facilities, enabling rapid evaluation and providing an accurate capacity calculation method for the preliminary evaluation of gas storage facilities converted from constant-pressure water bodies.

This application discloses a method, apparatus, equipment, and storage medium for analyzing the energy storage capacity of a distribution network. This application relates to the field of distribution network management technology. The method includes: acquiring historical operating information corresponding to a target distribution network; predicting the output of the configured energy in the target distribution network during a target time period based on the historical operating information, obtaining predicted output data; and estimating the energy storage capacity corresponding to the target distribution network based on the predicted output data and the cost parameters corresponding to the target distribution network. This application obtains historical operating information corresponding to the target distribution network, predicts the predicted output data of the target distribution network during a target time period based on this historical operating information, and then analyzes the energy storage capacity configuration corresponding to the target distribution network practically and conveniently based on the predicted energy output characteristics and the cost parameters of the target distribution network. This analysis method requires fewer parameters, is easy to obtain, and is simple and efficient, effectively expanding the applicable scenarios for distribution network energy storage capacity analysis.

The present application relates to the technical fields of hydraulic fracturing slug dosage design, in order to solve the problems of complex calculation process and long time in slug dosage design, the present application provides a kind of optimization design method of hydraulic fracturing slug dosage based on frog leap algorithm, according to the fitness function of constraint condition and optimization target, constraint condition includes slug dosage constraint condition;Optimization variable is continuously iterated and optimized according to the fitness function by using frog leap algorithm, and optimization variable includes the slug fracturing fluid dosage of each section, until the convergence condition is met, so as to search the global optimal solution in the feasible region of slug fracturing fluid dosage.It also provides a system for implementing the method of the present application, including feasible region setting module, optimization variable setting module, constraint condition setting module, fitness function setting module and optimal solution search module.The frog leap algorithm used in the present application is easy to program and realize, and the calculation speed is fast, the optimization ability is strong, and the efficiency of slug dosage optimization design is guaranteed.

This invention provides an epileptic seizure prediction system based on CNN and self-attention mechanism, belonging to the field of signal feature analysis. It includes: a sample acquisition unit for acquiring a sample dataset; a first transformation unit for performing an S-transform on the sample channel signals to obtain corresponding sample feature sets; a sample feature matrix determination unit for determining a sample feature matrix based on each sample feature set; a network construction unit for constructing an initial prediction neural network; a training unit for training the initial prediction neural network; a signal acquisition unit for real-time acquisition of the EEG signal to be tested; a second transformation unit for performing an S-transform on the EEG signal to be tested to obtain a test feature set; a test feature matrix determination unit for determining a test feature matrix based on each test feature set; and a prediction unit for determining whether the current time period is a pre-seizure phase based on the epileptic seizure prediction model. This improves the accuracy and efficiency of epileptic seizure prediction.

The present disclosure relates to an active return control method and device, an electric power steering system and a vehicle, wherein the method comprises: acquiring a steering wheel angle; calculating an active return demand torque by using a cascade control algorithm according to the steering wheel angle; and controlling a motor to output a corresponding torque according to the active return demand torque, so as to return the steering wheel.

The application discloses an HDR imaging method based on a Bayesian neural network, which comprises the following steps: firstly, mapping input LDR images with different exposures to a linear domain; secondly, connecting the LDR images with different exposures and the images in the linear domain, and inputting the images into a feature extractor to obtain features; thirdly, obtaining new features through a convolution attention module and a convolution layer; and finally, obtaining an HDR image through a feature fusion module. The application can reduce network parameters, expand a receptive field, make network calculation faster, and improve performance.

The invention discloses a gold thread segmentation detection method and system in chip AOI detection, electronic equipment and a computer readable medium, and the method comprises the following steps: obtaining a chip image; performing welding spot positioning on the chip image to obtain a first welding spot coordinate and a second welding spot coordinate of a gold wire; calculating the length L and the inclination angle of the gold thread according to the coordinates of the first welding spot and the second welding spot; the gold wire is divided into a plurality of sections, each section corresponds to a detection area, the detection area is formed by connecting indented end points, and the indented end points are obtained through calculation according to coordinates of a first welding spot and a second welding spot, a preset indentation proportion, the length and the angle; for each section of gold thread, performing threshold segmentation by using the gray threshold corresponding to the section, and extracting the region of the section of gold thread; and judging whether the gold thread areas are communicated or not and whether the whole gold thread is communicated or not so as to determine whether the gold thread has defects or not. According to the invention, the gold thread detection algorithm complexity and cost are reduced.

The application belongs to the field of image fusion, and discloses a kind of image fusion method based on double decoder, for solving the problem that the feature extraction capability and fusion effect of complex multi-modal image processing of camera shooting of different imaging modes based on deep learning-based image fusion method are poor, the application includes: multi-modal image A1, A2 is extracted through large receptive field feature extraction module, then respectively through two interactive decoding modules, and in the decoding process, the decoding information between two different modal decoding modules is spliced on the channel to interact and fuse, finally reconstructs the fusion image C, calculates the loss of fusion image C and multi-modal image A1, A2, updates network model parameters.The application can effectively realize the fusion of complex multi-modal image, has the characteristics of good feature information extraction, small parameter quantity, high reconstruction precision, strong robustness and the like.

The application belongs to the field of microwave technology, and particularly relates to a W-waveband ultra-wideband waveguide filter, the filter structure comprising a resonant cavity and metal-loaded capacitance, wherein a downlink filter port and an uplink filter port are arranged at the front end of the cavity; the resonant unit is composed of nine resonant cavities, and two adjacent resonant cavities are connected through direct coupling; a metal column is inserted into the H plane of each resonant cavity through the capacitive loading technology, so as to realize ultra-wideband, high out-of-band suppression and miniaturization; the structure of the application is composed of nine rectangular resonant cavities offset coupling, a circular metal is inserted into the resonant cavity, and the overall structure is symmetrical about the fifth resonant cavity; the structure has fewer parameters in simulation optimization, reduces the design time, and improves the design efficiency.