39results about How to "Reduce hardware requirements" patented technology

The invention discloses an improved YOLOv11 crack identification and quantification method for a strip mine area and a slope thereof. The method comprises the steps of data preprocessing, model improvement and adaptability adjustment, initial identification model training, crack detection by applying a target identification model, local area identification and post-processing. According to the method, technical means such as a feature extraction network improvement method based on an AReLU function and an ACmix model and a self-adaptive local threshold segmentation algorithm based on a gray scale standard deviation are introduced, and efficient and accurate recognition of cracks is achieved. According to the method, the automation degree of crack detection is improved, the labor cost and the time cost are reduced, and the method is of great significance to safety monitoring of the strip mine area.

The invention discloses a current inversion method based on a magnetic sensor array, and the method comprises the steps: S1, setting a plurality of measurement points, uniformly dividing a conductor layout region in a to-be-measured part of a to-be-measured multi-conductor system into grids with M * N nodes, and numbering each grid node as (i, j); constructing a coordinate system and recording coordinates of all grid nodes; s2, establishing a finite element simulation calculation model, calculating a reference position coefficient matrix of all measurement points, and constructing a reference position coefficient matrix database; s3, calculating a conductor position coefficient matrix, and constructing a position coefficient matrix; s4, according to the magnetic field measurement value acquired by the magnetic sensor array, the conductor current is calculated through inversion; and S5, after resampling the magnetic field measurement value, substituting the updated measurement value and the position coefficient matrix stored in the step S4 into the linear equation set constructed in the step S4 to solve the current of each conductor. The method has the characteristics of rapid calculation and low operand while the accuracy is ensured, and the inversion calculation efficiency can be improved.

The invention provides a track direction and height detection method and a track detection device.The track direction and height detection method comprises the steps of data acquisition, device track acquisition, compensation value acquisition, real track acquisition and track direction and height data acquisition, angular velocity data is acquired through a gyroscope, angle data is acquired through an inclinometer, and the angular velocity data is acquired through an angle sensor; the contour data of the two steel rails are collected through the line laser sensor, and the mileage data are collected through the mileage detector, so that the three-dimensional track coordinates of the track detection device can be obtained based on the angular velocity data, the angle data and the mileage data; then, based on the high-precision line laser contour data, obtaining a feature point coordinate at a specified position on the inner side of the steel rail as a rail direction compensation value, and compensating a three-dimensional track coordinate of the device by using the feature point coordinate, so as to obtain a real track coordinate at the specified position on the inner side of the steel rail along the x direction and the top surface of the steel rail along the z direction; and accurate rail direction and height data can be calculated.

The invention discloses a tension monitoring method, device and system for a multi-component coupled chain transmission system. The method comprises the following steps: carrying out dynamic modeling and simulation analysis on the multi-component coupled chain transmission system; on the basis of analyzing the basic structure and the working principle of the multi-component coupled chain transmission system, the chain tension of the multi-component coupled chain transmission system is analyzed and calculated; finite element analysis is carried out on the multi-component coupled chain transmission system, and a virtual chain tension sensor is designed; and carrying out experimental research on the virtual sensor design method. Real-time monitoring of the chain tension is achieved through the virtual sensor, the hardware requirement and the system cost are reduced, and the monitoring precision and the real-time performance are improved.

The invention provides an apoplexy early warning system based on voice, images, motion data and electrocardiogram data. The apoplexy early warning system comprises a language barrier judgment module used for analyzing user voice definition and fluency based on user voice; the ataxia judgment module is used for analyzing the accuracy and continuity of the finger and nose actions of the user based on the video stream; the mouth inclination judgment module is used for analyzing the symmetry condition of the mouth relative to the central axis of the face based on the picture to obtain a mouth inclination judgment conclusion; the limb weakness judgment module is used for capturing action data in real time based on a built-in sensor and analyzing the action data; the electrocardio analysis module is used for analyzing the user electrocardio signals based on a neural network model to obtain an atrial fibrillation judgment conclusion; and the stroke risk comprehensive evaluation module is used for comprehensively evaluating the stroke risk score of the user based on a language barrier judgment conclusion, an ataxia judgment conclusion, a mouth inclination judgment conclusion, a limb weakness judgment conclusion and an atrial fibrillation judgment conclusion, and realizing risk layering.

The invention discloses an optical remote sensing image saliency target detection model and detection method, and belongs to the technical field of target detection. The saliency target detection model comprises a PVT backbone network module, an edge information extraction and fusion module BG, a feature enhancement module HAM, two space attention modules, a feature interaction module KTM and a feature fusion module. According to the invention, by introducing an edge detection module (BG) and a feature enhancement module (HAM), the target detection precision under a complex background is effectively improved. Compared with the existing CNN-based method, the method provided by the invention can better process targets with different scales and direction changes, reduces the probability of missing detection and false detection, and enables saliency target detection to be more complete and accurate.

The invention discloses a millimeter wave radar and monocular camera fused information acquisition device and method, and relates to the technical field of target detection. The device comprises a millimeter wave radar, a monocular camera and an adjusting bracket, wherein the adjusting bracket can adjust the positions and angles of the radar and the camera. And the data fusion module performs time synchronization and space alignment on the radar information and the image information, and obtains a target recognition result through a preset target recognition model. The acquisition method comprises the steps of timestamp matching, time compensation, joint calibration, coordinate transformation, image distortion correction and the like. Further, radar physical features are extracted by using a CFAR algorithm, image features are extracted by using a deep learning model, and effective features are obtained through PCA dimension reduction. And finally, an accurate target recognition result is obtained through feature splicing and optimization by using a support vector machine (SVM). According to the method, efficient fusion and target identification of millimeter wave radar and monocular camera information are realized, and the method has a wide application prospect.

The invention is suitable for the technical field of vehicle dynamics modeling and control, and provides a preview active suspension stiffness control method based on road surface grading and MAP mapping, and the method comprises the steps: collecting a road image through a vehicle-mounted visual sensor, dividing regions, and recognizing the road surface unevenness grade of each region through a convolutional neural network; obtaining the vehicle speed; and according to the current pavement grade and the vehicle speed, a vehicle speed-pavement grade-rigidity mapping MAP table generated in advance is inquired, and the target suspension rigidity is rapidly obtained and adjusted. According to the invention, the dependence on a high-precision sensor and a high-computing-power processor is reduced, and low-cost and fast-response rigidity optimization control of the preview active suspension is realized.

The invention discloses a non-cooperative moving target positioning method based on a multi-base positioning system. The method comprises the following steps: modeling the multi-base positioning system; dTD measurement information, DTD-rate measurement information and TDOA measurement information are extracted; dTD, DTD-rate and TDOA (time difference of arrival) measurement models are constructed; a TDOA measurement model is preprocessed by adopting a step-by-step estimation method, a generalized trust domain sub-problem is constructed, # imgabs0 # is solved, a DTD measurement model and a DTD-rate measurement model are preprocessed by adopting the step-by-step estimation method, a constrained weighted least square problem is constructed, the constrained weighted least square problem is relaxed into a convex positive semidefinite programming problem by utilizing a positive semidefinite relaxation method, and the convex positive semidefinite programming problem is solved. A CVX toolbox of Matlab is adopted to solve # imgabs1 # and # imgabs2 # for a positive semidefinite programming problem, or a joint estimation method is adopted to pre-process a DTD measurement model, a DTD-rate measurement model and a TDOA measurement model, a constrained weighted least square problem is constructed, a positive semidefinite relaxation method is utilized to relax the constrained weighted least square problem into a convex positive semidefinite programming problem, and the positive semidefinite programming problem is obtained. Solving # imgabs3 # for positive semidefinite programming problem by using CVX toolbox of Matlab

The invention relates to a remote sensing interpretation method, system and device and a storage medium, and the method comprises the steps: loading a sky map image on a base map, loading a to-be-interpreted remote sensing image on a first map layer, aligning the remote sensing image with the sky map image according to geographic coordinates, determining a target interpretation region in the remote sensing image, and carrying out the interpretation of the target interpretation region; generating position information for indicating the target interpretation area; obtaining a to-be-interpreted parameter of the target interpretation area, and sending the position information and the to-be-interpreted parameter to a server; wherein the to-be-interpreted parameters are used for indicating a target scene and a target interpretation model for interpreting the target interpretation area, an interpretation result of scene information the same as that of the target scene is efficiently and quickly interpreted in the server, and then the client displays the interpretation result in a second layer stacked with the base map and the first layer. According to the method, the information to be obtained can be quickly interpreted from the remote sensing image and visually displayed, and the interpretation process is carried out in the server, so that the hardware requirement of the client is reduced, and the applicability is improved.

The application discloses a kind of based on spot image moment feature telescope primary and secondary mirror calibration method, belong to telescope computer-aided alignment field.The specific steps of the calibration method are as follows, first, randomly give the free degree misadjustment of telescope system secondary mirror, utilize unbiased finite impulse response moment to extract the spot image feature of multiple field of view within reasonable range under the misadjustment state of telescope;Then utilize fully connected neural network to construct the model between the free degree misadjustment of secondary mirror and the multiple field of view spot image feature matrix;In actual use, after the multiple field of view spot image collected by CCD camera is extracted into feature matrix using unbiased finite impulse response, it is input into the trained neural network model to solve the misadjustment of secondary mirror, to guide the alignment of telescope system.This method aims to wavefront sensor as the image quality evaluation benchmark of telescope without, and without multiple iterations, to improve the efficiency of system primary and secondary mirror alignment and the imaging quality of system.

The present invention discloses a method for long-tail activity recognition using a dual-decoupled network on a mobile terminal, which specifically includes the following steps: collecting action signals and obtaining a training set; inputting the training set into a dual-decoupled network model, and realizing long-tail activity recognition through a mobile wearable recognition device; wherein the dual-decoupled network model is constructed by a neural network and is decoupled in a training phase and a test phase, respectively. The present invention enhances the feature extraction capability of the model by decoupling the training phase, i.e., separating feature extraction and activity classification, and adopts a multi-branch topology structure. At the same time, weight fine-tuning in the classifier phase can effectively improve the performance of the model in recognizing long-tail distributions; in the present invention, the decoupling of the training phase and the test phase significantly reduces the amount of model parameters, helps to reduce the hardware requirements of wearable recognition devices, and improves the real-time performance of activity recognition.

The invention relates to the technical field of energy storage equipment, and discloses an equalizing charging method for an energy storage device, the energy storage device and equipment, and the method comprises the steps: carrying out the pre-charging of the energy storage device, and carrying out the sorting of SoC values of a plurality of energy storage units on each phase of line under the condition that the DC side voltage is stable; performing partial dynamic cutting on the plurality of energy storage units on each phase of line according to the sorting result, and determining that the energy storage device completes pre-charging under the condition that the SoC values of the remaining energy storage units after cutting all reach a preset value; and performing grid-connected operation on the pre-charged energy storage device, performing in-phase equalization control on the plurality of energy storage units on each phase line according to the total SoC mean value and the in-phase SoC mean value of the energy storage device, performing inter-phase equalization control on the energy storage device according to the negative sequence component of the three-phase alternating current power grid, and quickly charging the energy storage device through a power ring. According to the invention, the over-charge or over-discharge phenomenon of individual energy storage units can be avoided, and the reliability of the energy storage device is greatly improved.

The present invention discloses a device and method for detecting and measuring surface defects of products, the device comprises a projector for projecting structured light onto the surface of the product to be inspected passing through the production line in real time, a camera for capturing the image of the surface of the product to be inspected with the projected structured light in real time, and a computer connected to the projector and the camera by a bus, the computer receiving the captured image from the camera, and detecting and measuring the surface defects of the object and quantifying the size specifications through the image processing algorithm preset by the computer. The machine vision detection method of device calibration, parameter learning, parameter setting, image correction, real-time detection and defect analysis improves the detection capability of three-dimensional defects on the surface of the object, especially compared with the traditional detection method, it has the improvement effect in many aspects such as accuracy, real-time performance, flexibility, etc., and reduces the hardware requirements for the computer, with outstanding cost performance.

The invention provides an airfoil profile streaming flow field rapid prediction method based on a Fourier neural operator, and belongs to the field of intelligent fluid mechanics. According to the method, machine learning and computational fluid mechanics are deeply fused, the future flow field state is rapidly predicted through a series of input and output data, and the calculation efficiency of the method is far higher than that of a traditional numerical method. According to the method, the calculation task of several hours or even several days of traditional computational fluid mechanics simulation can be shortened to the second level, the optimization iteration period of aircraft appearance design is greatly accelerated, and the dependence and cost of research and development on wind tunnel tests and high-performance calculation are remarkably reduced.

The present invention discloses a method and device for automatically counting algae, which obtains multiple sub-videos, wherein the sub-videos are obtained by photographing the algae in the water sample flowing through the microfluidic chip through a microscope and a camera; the algae concentration displayed in each sub-video is calculated, and the real concentration of the algae in the water sample is calculated based on the algae concentration displayed in each sub-video. The present invention does not need to observe all algae samples, but only needs to observe samples of some algae, and the concentration of algae in surface water can be calculated according to statistical laws, which overcomes the disadvantage that a zoom camera cannot be used when collecting video.

The present invention relates to the field of thermal simulation technology, and in particular to a thermal simulation method, system and circuit board for a circuit board. The method comprises the following steps: obtaining circuit board design data and modeling to obtain a three-dimensional structural model of the circuit board; performing accuracy estimation according to the circuit board design data to obtain thermal simulation accuracy data of the circuit board; dividing the three-dimensional structural model of the circuit board according to the thermal simulation accuracy data of the circuit board to obtain circuit board component grid data; assigning physical properties to the circuit board component grid data according to the circuit board design data to obtain physical grid unit data; performing thermal simulation of the circuit board according to the physical grid unit data to obtain circuit board thermal simulation data; visualizing the thermal simulation data of the circuit board to obtain a thermal stress distribution map, and optimizing the circuit board component design of the three-dimensional structural model of the circuit board according to the thermal stress distribution map to obtain an optimized three-dimensional structural model of the circuit board. The present invention can effectively improve the thermal management capability of the circuit board.

Embodiments of the present application provide an electric safety belt control method and device, electronic equipment and storage medium. The domain controller obtains the action information of the target scene for the electric safety belt controller, generates the corresponding action instruction according to the action information, and transmits the action instruction to the electric safety belt controller through the hard line. After that, the control result is obtained based on the received state information of the electric safety belt controller. Based on the above method, data transmission is carried out through the hard line between the domain controller and the electric safety belt controller. On the one hand, the additional consumption of human resources and hardware resources caused by the high development complexity during the development of the CAN bus is avoided. On the other hand, the bus transceiver required for logical conversion is cancelled, further reducing the hardware requirements for data transmission.

The invention discloses a positive and negative sample data processing method and system for a transformer substation patrol algorithm, and the method comprises the steps: setting a transformer substation fault type, determining a first type of parameters, determining a second type of parameters and a third type of parameters from high to low based on a logistic regression algorithm according to the weight of effect variables, and obtaining a positive and negative sample data according to a historical fault database. And performing positive and negative sample labeling on the visual fault sample, the parameter fault sample and the environment fault sample, constructing a model by adopting a convolutional neural network, inputting positive and negative sample data to perform model training, and inputting a first type of parameters, a second type of parameters and a third type of parameters in the real-time inspection data into the model to perform fault judgment. According to the method, positive and negative sample labeling is carried out on the three fault types, the strict requirement for patrol data sample processing is lowered, fault judgment is carried out on the three types of parameters which cause the highest effect quantity weight of the same fault type, and the reliability and accuracy of fault judgment are improved.

The invention discloses a method for realizing system hotspot account processing. The method comprises the following steps: receiving a transaction request for a target account; counting the access frequency, and dynamically identifying whether the target account is a hotspot account or not; when the target account is identified as a hotspot account, executing an atomic accumulation operation on a balance change increment value corresponding to the target account in the distributed cache, performing real-time balance verification based on an accumulation result and a pre-stored account reference balance, and returning a transaction processing success response after the verification is passed; the method comprises the following steps of: atomizing and acquiring an accumulated change increment value of each hotspot account from a distributed cache through a timed scheduling task, clearing the accumulated change increment value, and generating a batch updating instruction based on the acquired accumulated change increment value; and executing the batch update instruction in the database transaction. The computing load of high-frequency transactions is transferred from a database to a high-performance memory, so that the transaction processing capacity of the system is improved by magnitude, the response delay is reduced from the second level to the millisecond level, and the high-concurrency real-time scene is met.

The invention provides a carrier communication method for a laser energy transmission system with a dispersed structure, which is characterized in that a semiconductor laser is adopted as an electro-optical conversion device, a power controller and a signal modulator are combined at a transmitting end, and a modulation signal is superposed on a laser beam while energy transmission is realized, so that the laser energy transmission system with the dispersed structure is realized. Energy and information superposition is realized; at a receiving end, a photovoltaic array is adopted to convert received laser beam energy into electric energy, modulation information is extracted at the same time, and energy and information separation is achieved; a plurality of independent energy transmission links are arranged, each link independently transmits energy and information, and each link is connected to a photovoltaic array to realize energy conversion and information demodulation. The system is particularly suitable for high-power energy transfer occasions, synchronous transmission of energy and information is achieved, the overall transmission rate and data throughput of the system are greatly improved, the requirement for high-frequency modulation equipment is reduced, the requirement for hardware is low, and good system compatibility and expandability are achieved.

The invention relates to the technical field of intelligent recognition, in particular to a voiceprint-based vehicle load variation recognition method and system, and the method comprises the following steps: collecting sound signals generated when a vehicle passes through a deceleration strip when the vehicle is driven into and out of a specific place, a driving-in sound signal and a driving-out sound signal of the same vehicle are associated in combination with license plate recognition; the collected sound signals are preprocessed; extracting a time domain feature, a frequency domain feature and a time-frequency domain feature of the preprocessed sound signal; on the basis of the extracted features, feature vectors of the driving-in sound signal and the driving-out sound signal are generated respectively; and calculating the Euclidean distance between the feature vectors of the driving-in sound signal and the driving-out sound signal by adopting an Euclidean distance multi-scale entropy algorithm, and judging whether the load of the vehicle varies or not according to the Euclidean distance in combination with the driving speed of the vehicle. By adopting the method, the vehicle load variation can be effectively identified, and the method has the characteristics of low deployment cost and high robustness.

The invention provides a reverse analysis method and system for physical layer characteristics of a radio frequency chip. The method comprises the following steps: configuring a software environment of a radio frequency chip with unknown physical layer configuration characteristic information, after the software environment is configured, performing code burning on two development boards carrying the radio frequency chip, and taking one development board after code burning as a transmitting end and the other development board after code burning as a receiving end; controlling the transmitting end to perform forward operation in the data processing flow, and controlling the receiving end to perform reverse operation in the data processing flow; and performing reverse analysis according to all data generated by the transmitting end and the receiving end in the data processing flow to obtain physical layer characteristic information of the radio frequency chip. According to the technical scheme provided by the invention, the reverse analysis efficiency and accuracy of the physical layer characteristics of the radio frequency chip are improved.

This application provides a receiver image compensation method and system without a training sequence, relating to the field of wireless communication technology. After obtaining the intermediate frequency (IF) signal and its conjugate signal, the method calculates image coefficients based on the correlation between the IF signal and the conjugate signal, then synthesizes an image compensation matrix according to the image coefficients, and finally uses the image compensation matrix for image compensation. The method can analyze random spatial radiation signals received by the receiver in free space, utilizing the conjugate relationship between the image signal and the main signal, and obtaining the image compensation coefficients through the conjugate correlation of the IF signal. This method can analyze the image coefficients without a dedicated training sequence, reducing the complexity and hardware requirements of image compensation.

The invention discloses an automatic cleaning tank for a gravure roller of a printing machine, and relates to the technical field of corollary equipment of the printing machine, the automatic cleaning tank comprises a mounting box, an ultrasonic plate washing tank is arranged on one side of the mounting box, and the upper end of the ultrasonic plate washing tank is flush with the ground; the multiple positioning holes are evenly formed in the bottom of the ultrasonic plate washing tank in an array mode, positioning guide mechanisms are fixedly arranged at the center positions of the upper ends of the positioning holes, placing bases can be installed in the positioning holes, and shaftless gravure rollers are placed on the placing bases; according to the shaftless gravure roller cleaning device, complete sets of shaftless gravure rollers are sequentially placed on the placing seats, then the shaftless gravure rollers and the placing seats are sequentially transferred to the positioning holes in the ultrasonic plate cleaning tank through the transferring mechanism to be cleaned, the overall cleaning speed and the production efficiency are greatly improved, meanwhile, due to the fact that a plurality of plate rollers can be treated at a time, the cleaning efficiency is greatly improved, and the production cost is reduced. The space and energy input of the ultrasonic plate washing tank are fully utilized, and the cleaning cost of the unit plate roller is reduced.

The application discloses a mechanical structure of a flexible power-assisted exoskeleton system and a control method thereof, and the mechanical structure comprises an actuating unit box, a driving device, a main controller, a power module, an intention detection component and a power-assisted unit; the body function of a user can be enhanced under different environments, including climbing, crawling, going upstairs and downstairs; the control method is characterized in that an inertial measurement unit is fixed on each main limb of the human body, and the real-time measured limb posture data are transmitted to the main controller; the main controller performs mean filtering and preprocessing on the data, so that the credibility of the data is enhanced; the main controller obtains the corresponding human motion phase according to the posture analysis of different human motions; finally, the main controller sends a driving signal to the driving device according to a preset power-assisted strategy, and the motor rotates according to the signal, so that power assistance is achieved. The application has the advantages of simple algorithm, high real-time performance, strong stability, low hardware cost, low hardware demand and easy engineering application implementation.

This invention discloses a distributed beamforming method, system, and apparatus for low-Earth orbit satellites, comprising: acquiring ground user location information and satellites to be communicated with; calculating the transmission paths between multiple satellites and ground users and obtaining the superimposed signal of electromagnetic waves at the ground end; calculating the superimposed power distribution of the ground electromagnetic waves based on the superimposed signal; constructing an objective function that maximizes the average received power of all ground users and the fairness of ground users based on the calculated superimposed power distribution of the ground electromagnetic waves; calculating the optimal satellite and its corresponding initial transmission phase within the selectable set, indirectly controlling the transmission distance from the satellite to the ground end to control the final phase of the electromagnetic waves received by the receiver, thereby maximizing the overall power gain of multiple ground receiving terminals and the fairness among each individual ground user, providing ideal power gain regardless of the location of the ground user, and simultaneously covering multiple users.

The application belongs to the technical field of electric power intelligent inspection, and particularly relates to a lightweight electric power fitting detection method and system based on dual priori enhancement and a medium. The method comprises the following steps: acquiring an electric power fitting image of a power transmission line as input; constructing a target detection network; generating high-dimensional static semantic feature embeddings corresponding to various categories according to prompt words of electric power fitting category names and physical attributes; mapping the high-dimensional static semantic feature embeddings to a space matching the feature dimension of the target detection network; completely replacing text features extracted by a general CLIP text encoder by using semantic features output by a semantic projection layer after dimension reduction; directly inputting the semantic features after dimension reduction into a detection head, decoding and outputting the final detection result of the electric power fitting by the detection head; and converting the model into a lightweight neural network inference engine format and deploying the model on an edge device for real-time inference. The application significantly enhances the detection capability of small targets in a complex background.

The embodiment of the invention discloses a memory exchange method and device, a terminal and a computer program product, and belongs to the technical field of memory exchange. The method comprises the following steps: allocating a memory exchange partition from a reserved space of a physical dynamic partition, wherein the reserved space is a space reserved for system upgrading in the physical dynamic partition; and under the condition that a memory exchange requirement exists, performing memory exchange between the memory and the memory exchange partition. According to the scheme provided by the embodiment of the invention, idle storage space can be effectively utilized, and storage space waste is avoided; moreover, a terminal which is not provided with a special physical partition serving as a memory exchange partition can also realize memory exchange by adopting the scheme provided by the embodiment of the invention, so that the hardware requirement during memory exchange is reduced, and the terminal application range of a memory exchange technology is expanded.

The invention discloses a small sample image classification migration method based on low-frequency enhancement, and the method comprises the steps: enabling an existing image classification model to be divided into an image feature encoder and a classifier, setting a convolution kernel adapter for each convolution kernel in the image feature encoder, taking the parameters of the convolution kernel adapters and the classifier as learnable parameters, and carrying out the recognition of the learnable parameters. Other parameters are fixed as parameter values obtained by pre-training of the image classification model, then multiple rounds of training are carried out, and in each round, sampling is carried out from a small sample set of each image classification task to obtain a support set and a query set for training; and after the training is finished, selecting the model parameter with the maximum average prediction precision as the final image classification model parameter according to the average prediction precision of each round. According to the method, the learning speed of small sample image classification migration can be effectively improved, the video memory consumption is reduced, and the learning robustness is improved.