25results about How to "Avoid error accumulation" patented technology

The present application relates to the technical field of precipitation forecast, solves the problem that the prior art is prone to underestimate the peak value of precipitation or produce false alarm under strong convective weather, and particularly relates to a short-impending precipitation forecast method fusing GNSS and water conservancy rain measurement radar, GNSS observation data and water conservancy rain measurement radar base data in a monitoring area are acquired, pretreated, corresponding atmospheric precipitable water distribution field and radar reflectivity factor distribution field are generated, based on a preset time sliding window, time sequence characteristics of the atmospheric precipitable water distribution field and the radar reflectivity factor distribution field are extracted, and the two are superimposed in channel dimension to construct a multi-channel spatio-temporal fusion input tensor. The present application overcomes the defect that the traditional extrapolation method rapidly decays with the prolongation of the forecast time, significantly improves the prediction accuracy of the rainstorm center intensity and the falling area, reduces the missed report and underestimation of the precipitation event, and solves the problem of low prediction accuracy of the traditional linear extrapolation algorithm under strong convective weather.

This invention provides a nonlinear time-series estimation method for the state of health (SOH) of lithium batteries, comprising: collecting feature data during the charge-discharge cycle of the lithium battery and generating a multi-channel standardized time-series dataset; extracting features and compressing dimensions of the data through an encoder; reconstructing and denoising the data through a decoder to output sequence features; mining the global correlation between feature sequences to obtain a context vector; further obtaining the SOH estimation result and synchronously updating the parameters. This method effectively suppresses model overfitting while introducing an improved KL divergence to prevent posterior collapse, achieving high-precision and robust estimation of the SOH of lithium batteries.

This invention discloses a method and apparatus for predicting the performance degradation of integrated circuit power supplies, relating to the field of integrated circuit technology. The method includes: smoothing and denoising the first health index data of the collected sample power supply using a Kalman filter algorithm to obtain a historical health index dataset; truncating the historical health index dataset into multiple first data sequences of preset window lengths, determining the true remaining service life label corresponding to each first data sequence, and generating a supervised learning dataset; determining the initial parameters of an LSTM-Attention prediction model based on a particle swarm optimization algorithm and the supervised learning dataset; iteratively training the model to obtain a target LSTM-Attention prediction model; and predicting the remaining service life of the collected second health index data of the power supply to be monitored based on the target LSTM-Attention prediction model, enabling early warning and switching processing of the power supply before a serious fault occurs.

This invention provides a traffic flow speed prediction method and system based on directed hypergraphs and attention mechanisms, belonging to the field of traffic state prediction technology. Based on directed hypergraphs, it constructs the relationships between nodes in a road network; aggregates directed hyperedge information to characterize the complex spatiotemporal features of the road network; builds an encoder and decoder architecture; integrates an attention mechanism, constructs an attention module, and introduces an improved dense connection structure to enhance the method's accuracy. This invention effectively overcomes the shortcomings of methods based on traditional graph structures, achieves the extraction and fusion of temporal and spatial features, and solves problems such as gradient vanishing in deep neural networks. It is applicable to multi-step traffic flow prediction in real road networks and achieves good prediction accuracy.

The present application provides a field core tank corner point extraction and perspective distortion correction method based on multi-scale Retinex fusion, comprising: S1. Core tank image preprocessing based on anisotropic diffusion and multi-scale Retinex fusion; S2. Core tank corner point extraction based on topological constraint Shi-Tomasi and graph cut optimization interaction; S3. Perspective correction based on thin plate spline transformation and Markov random field fusion; S4. Correction result precision verification and closed loop feedback based on Hough transform straight line fitting and geometric moment invariant fusion. The present application realizes accurate extraction of core tank corner points and efficient distortion correction of images, and simultaneously completes full-dimensional precision verification of the correction results, adaptive migration of processing parameters of batch images and lightweight field deployment of the algorithm.

This application provides a robot control method and electronic device. The method includes: a three-layer state machine based on a task layer, a sub-task layer, and an action layer; upon receiving a task pause signal, generating and storing context information based on the signal reception time; upon receiving a recovery signal, acquiring the current environment, time, and resource information, and determining whether the task can continue execution based on the context information; if execution is possible, acquiring the robot's current pose information, performing position correction based on the context information to obtain the continuation position, and finally controlling the robot to continue task execution from the continuation position based on the context information. This application achieves refined control and precise breakpoint continuation of robot tasks, effectively solving the problems of inconsistent states and position drift during task pause and recovery, avoiding invalid task execution, improving the flexibility, reliability, and accuracy of robot task execution, and optimizing the human-computer interaction experience.

This invention relates to the field of medical image analysis technology, specifically to a method and system for dry eye detection based on multimodal prior and generalized contrastive learning. The method includes: first, acquiring typical images, descriptive text, and relationship graphs of various dry eye diseases; extracting and stitching multimodal features to construct a medical prior knowledge base; second, extracting visual features from the anterior segment of the eye image and retrieving corresponding semantic features from the knowledge base according to disease labels, then projecting and aligning the two; third, constructing a medical similarity matrix based on multimodal features, mining difficult negative samples, and performing generalized contrastive learning on positive and ordinary negative samples to train a visual feature extraction network; finally, inputting the test image into the network for tear river segmentation, calculating the tear river height and pupil-tear river distance, and combining the cosine similarity of visual features and various semantic projection features to output the probability that the test image belongs to each dry eye disease level. This invention improves the accuracy of dry eye detection.

This invention provides a method for identifying ground faults in distribution network lines based on a time-series large model, comprising: Step S1: data acquisition and spatiotemporal alignment; Step S2: multi-scale sliding window segmentation; Step S3: calculation of physical enhancement features; Step S4: construction of a training sample set; Step S5: construction of a time-series large model with physical information enhancement; Step S6: model training; Step S7: real-time acquisition and identification of ground faults. This invention achieves high-precision, low-false-alarm ground fault detection, accurately identifies fault types and faulty lines, and effectively solves the identification challenges of extreme scenarios such as intermittent arc grounding and high-resistance grounding.

The application discloses a kind of measurement method and measuring device of tire capsule center line skew amount, including placing tire to be measured, make its axis with first direction parallel;Along the first direction, the initial position of the indicating mark along the first direction is set as reference zero point, by moving preset indicating mark, make it with the one side edge of the capsule center line in the tire inner align, and again;Under the condition that the relative position of tire and indicating mark in the plane perpendicular to first direction is unchanged, tire is turned over;Again, the indicating mark is moved along the first direction, make it with the same side edge of the capsule center line in the tire inner after turning over align;The displacement value of indicating mark relative to reference zero point in the first direction is obtained at this time, take the absolute value of the value as the capsule center line skew amount of tire at current measurement position, half.This application converts arc length measurement into height difference measurement, easy to operate, without positioning tire center, direct reading, greatly improve measurement efficiency and accuracy.

Embodiments of the application disclose a camera calibration simulation method and device, electronic equipment and a computer readable storage medium, relating to the technical field of camera calibration; the method comprises: constructing a simulation environment containing a calibration board and a plurality of virtual cameras, generating a plurality of groups of candidate pose sequences by using a Monte Carlo sampling method; performing grid uniformity evaluation and concave hull coverage evaluation to obtain an optimal pose sequence, generating simulation calibration data based on the optimal pose sequence; constructing a camera co-view weight map according to the calibration board corner points co-observed by the plurality of virtual cameras, determining a calibration transmission link; performing joint calibration on the plurality of virtual cameras according to the calibration transmission link by using the simulation calibration data to obtain a calibration result; and calculating the difference between the calibration result and a calibration true value to obtain a calibration evaluation report. In this way, the present scheme can accurately quantify the calibration accuracy based on the calibration true value and clearly determine the influence of the calibration board pose to provide guidance for the calibration board pose.

The application discloses a well-seismic error analysis method of depth domain seismic data, an electronic device and a medium. The method comprises the following steps: obtaining full-area time domain seismic data, depth domain seismic data, drilled well logging curve data and drilled well layered data; obtaining a time domain synthetic seismogram based on the time domain seismic data and the drilled well logging curve data; obtaining time domain seismic data wave group characteristic information corresponding to a target layer section according to the time domain synthetic seismogram; identifying the time domain seismic data wave group characteristic information on the depth domain seismic data, and reading corresponding underwater true vertical depth; obtaining well-up true vertical depth according to the drilled well layered data and the depth domain seismic data; and obtaining well-seismic error of the depth domain seismic data according to the underwater true vertical depth and the well-up true vertical depth. In this way, reliable well-seismic error can be obtained.

The utility model discloses a turnover type clamping mechanism for machining a disc brake of an electric vehicle, which relates to the technical field of turnover type clamping mechanisms and comprises a fixed mounting seat. The upper surface of one end of the fixed mounting seat body is fixedly connected with the main turnover mechanism, and the upper surface of the other end is slidably connected with the auxiliary turnover mechanism; an electric vehicle disc brake is arranged between the main turnover mechanism and the auxiliary turnover mechanism; the main turnover mechanism and the auxiliary turnover mechanism are correspondingly arranged and are partially structurally connected; according to the utility model, the base, the main support frame and the auxiliary support frame form a stable support system; the first pneumatic clamping jaw and the second pneumatic clamping jaw are matched from the two sides to clamp the electric vehicle disc brake, external force generated in the machining process can be effectively counteracted, it is guaranteed that the electric vehicle disc brake is stable and does not move in the machining process, the machining precision is improved, and the size precision and the form and location tolerance of all parts of the electric vehicle disc brake are easier to control within the standard range.

This invention discloses a method and system for tracing the source of harmonics in a distribution network, belonging to the field of distribution network monitoring technology. The method includes: S1, constructing an initial distribution network harmonic causal graph; S2, sensing real-time changes in the distribution network topology and dynamically updating the initial harmonic causal graph to obtain an updated harmonic causal graph; S3, decomposing the updated harmonic causal graph into multiple single-frequency causal subgraphs, performing causal intervention on each subgraph to separate background harmonics from user-side harmonics, calculating the contribution of each user-side harmonic source, and outputting the tracing results. This invention uses a causal graph neural network to replace traditional statistical correlation analysis, thereby effectively distinguishing between causal relationships and spurious correlations in harmonic propagation. By decomposing the harmonic causal graph into multiple independent single-frequency causal subgraphs, accurate decoupling of broadband coupled harmonics is achieved, avoiding error accumulation caused by frequency domain decomposition.

This invention provides a dedicated measuring tool and method for measuring the length of a step inside a deep hole. The measuring tool includes a gauge cylinder with a through hole; a dial indicator whose measuring rod is inserted into and fixed in a through hole at one end of the gauge cylinder; a measuring rod inserted into a through hole at the other end of the gauge cylinder, with a support block and a small measuring block fixedly disposed at its middle part and exposed end, respectively; a large measuring block fixed at the end of the gauge cylinder away from the dial indicator, with its inner end located inside the gauge cylinder as a limiting element for the support block; a spring sleeved on the measuring rod; and a reference block with a corresponding through hole featuring a step, the length of which is set as the maximum limit dimension of the step length inside the hole of the workpiece to be measured. This invention achieves measurement by directly measuring the length of the step inside a deep hole, which reduces the measurement difficulty, avoids error accumulation, and effectively solves the problem of accurately measuring the length of the step inside a deep hole of a bushing with precision requirements.

The present application relates to the technical field of wheelchair control, and particularly relates to an electric wheelchair walking assisting control method and system based on road condition self-adaptation, which comprises the following steps: acquiring internal electrical signals and bilateral rotation speed feedback signals, collecting acceleration and angular velocity information of the wheelchair; performing fusion processing on the acceleration and angular velocity information to obtain a pitch angle; constructing an intention feature quantity reflecting a user's pushing intention and load change through an intention recognition algorithm to generate a reference assisting force; performing feedforward compensation processing on the reference assisting force to adapt to the type of the driving road to obtain feedforward compensation torques of bilateral motors; performing synchronization dynamic calibration on the feedforward compensation torques of the bilateral motors through a synchronization calibration algorithm; superimposing the reference assisting force and the calibrated feedforward compensation torques through a torque calculation model to obtain target torques and obtain motor control instructions. Through the fusion of motor operation parameters and IMU attitude data, the user's pushing intention is accurately recognized and the road condition is self-adapted, and the safety is significantly improved.

This invention discloses a three-dimensional positioning method for aerial targets based on distributed ground radar cooperation, comprising: constructing a cross-correlation function by transmitting and echoing signals, and obtaining the range estimate of each ground radar to the target through quadratic fitting; constructing a nonlinear spherical equation system, introducing auxiliary variables for linearization, and iteratively solving it using the Lagrange multiplier method to obtain the initial three-dimensional position estimate of the target; extracting the initial horizontal coordinates, obtaining the altitude correction contribution of each radar based on the geometric relationship between the radar and the target, adaptively selecting the radar with the largest contribution and satisfying the decoupling premise as the optimal correction radar, using its range information to decouple from the spherical range equation to obtain the altitude closed-form solution, and fusing the horizontal coordinates and the corrected altitude to obtain the final three-dimensional position. This invention overcomes the accuracy degradation problem caused by the lack of altitude dimension information and ill-conditioned matrix in traditional algorithms, and significantly improves the three-dimensional positioning accuracy of distributed ground radar for aerial targets.

The application discloses a voltage transformer error prediction method based on modal decomposition and a gated recurrent unit, and comprises the following steps: 1, collecting historical error data of a capacitive voltage transformer to construct a time series data set; 2, using an improved modal decomposition method, dynamically adjusting a noise amplitude coefficient based on signal frequency domain characteristics, and repeatedly decomposing to obtain a mean modal component; 3, calculating the information entropy value of each component, classifying the components according to the entropy value and a time-frequency diagram, and using an adaptive wavelet packet threshold denoising method for high-entropy components; 4, designing an error signal prediction model enhanced by a non-local attention mechanism, and dynamically fusing local time sequence characteristics and global context information through a gating coefficient; 5, establishing a sub-modal parallel training framework, modeling and predicting each component respectively, and fusing and outputting; and 6, model performance evaluation and verification, and verifying the effectiveness of the method. The method effectively improves the prediction accuracy of the output signal of the voltage transformer and can effectively adapt to the signal prediction demand under complex working conditions.

This application discloses a learning method, apparatus, device, and medium for testing a large language model. Test data is input into a large language model with low-rank adaptation parameters to obtain the prediction score of lexical units in the vocabulary at each generation step. Based on the prediction scores, the lexical-level evidence quality of the generation step is extracted, and the lexical-level cognitive uncertainty of the generation step is calculated. The lexical-level cognitive uncertainty is smoothed, and a stable region of the generation step is determined based on the smoothed uncertainty. A mask for the model update window is constructed. Based on the mask, the mask entropy loss is calculated, and the low-rank adaptation parameters are updated through backpropagation. This application can solve the problem of adaptive drift in model updates in related technologies.

This invention provides a method and system for depicting brain midline CT images based on a 2.5D architecture and multiple attention mechanisms. The method includes: constructing a brain midline depiction network, namely GUP-Net (Global Uncertainty-Location Network), and introducing a three-dimensional spatial correction strategy, a Global Slice Attention (GSA) module, a Dynamic Uncertainty Awareness (DUA) module, and a Position Attention (PA) module; the GSA module is used to fuse cross-slice features to enhance key layer recognition; the DUA module optimizes feature extraction during downsampling to reduce lesion interference; the PA module improves spatial localization accuracy during upsampling; and the GUP-Net is trained using three-dimensionally corrected brain CT images to obtain the depiction model. This invention effectively solves the problems of distortion, blurring, and displacement of the brain midline caused by lesions, improving depiction accuracy and robustness.

This invention discloses a transfer alignment method based on a right-invariant error model using Lie groups, belonging to the field of inertial navigation systems. The method includes: first, referencing the navigation coordinate system, remodeling the right-invariant attitude and velocity errors based on Lie group theory to replace traditional transfer alignment errors, and incorporating sensor bias, lever arm and installation angle errors between the master and sub-inertial navigation systems to fully consider transfer alignment error sources; second, re-deriving the system state equations within the Lie group framework to decouple computational parameters from the system model; then, based on an "attitude + velocity" measurement matching scheme, re-deriving the system measurement equations within the Lie group framework; finally, addressing the nonlinear characteristics of the measurement equations, combining an unscented Kalman filter to accurately estimate the system state. The transfer alignment method proposed in this invention fully considers the objectively existing error sources between the master and sub-inertial navigation systems, improves system error modeling within the Lie group framework, and ensures that the system model accurately reflects the actual situation.

The present application relates to the technical field of automatic driving, in particular to a depth map completion method based on double-branch gradual fusion, comprising: processing color images and laser radar sparse depth images of the same target scene through a double-branch depth completion network based on mixed attention fusion and gradual refinement; training the double-branch depth completion network based on mixed attention fusion and gradual refinement using a training set in a depth completion dataset, obtaining network weight parameters after each round of training, and verifying the network weight parameters through a verification set, selecting the network weight parameters with the smallest error as the optimal network weight parameters; and performing depth completion on the double-branch depth completion network using a test set and the optimal network weight parameters to obtain a target dense depth image. The present application optimizes feature processing by extracting, fusing and refining the features of sparse depth images and color images, and finally obtains a dense depth image with rich details and accurate depth values.

The application provides a robot dynamic error compensation method based on a capsule neural network, adopts Newton-Euler method to establish a robot dynamics model, and performs linearization and dynamics parameter identification; uses joint motion data as input to obtain predicted torque for model training by using the linearized dynamics model after the dynamics parameter identification, and calculates error between the predicted torque and actual torque; regards normalized parameters as vectors in a vector space, calculates inner products of the vectors to form a Gram matrix and converts the Gram matrix into a two-dimensional image; learns error between the predicted torque and the actual torque by using a capsule neural network, and obtains a trained capsule neural network; obtains real-time prediction error by using the trained capsule neural network, and compensates real-time predicted torque obtained by the linearized dynamics model after the parameter identification by using the real-time prediction error. The application realizes high-precision and adaptive dynamic error compensation.

The application discloses a kind of measurement methods of alignment accuracy, measuring device and computer readable storage medium, in which the measurement method of alignment accuracy includes: with the reference frame line of transparent substrate and limit post as reference, the first surface of product to be detected is towards transparent substrate and is fixed on transparent substrate;Determine the first relative position information of first feature point on second surface and reference frame line set position;Rotating product to be detected and transparent substrate, determine the second relative position information of second feature point on first surface and reference frame line set position, wherein, second surface is oppositely arranged with first surface;Determine the alignment accuracy of first feature point and second feature point using first relative position information and second relative position information.The method improves the efficiency of measuring alignment accuracy, avoids the cumulative error caused by multiple clamping.