99results about How to "Avoid error accumulation" patented technology

The invention relates to the technical field of air conditioner copper pipes, in particular to a multi-station synchronous correction control system for an air conditioner copper pipe. Comprising a copper pipe straightening unit used for straightening a to-be-treated copper pipe; the detection unit is used for collecting structural parameters of the to-be-treated copper pipe; the central control unit comprises a first processing module and is used for establishing a copper pipe analysis model; the first processing module is further used for setting a primary correction strategy according to the structure parameters of the copper pipe to be processed and the copper pipe analysis model; the first processing module is further used for setting working parameters of the copper pipe correction unit according to the primary correction strategy; the second processing module is used for establishing a linkage monitoring model; the second processing module is further used for judging whether a compensation correction instruction is generated or not according to the linkage monitoring model. By analyzing the structural parameters of the to-be-treated copper pipe, the correction efficiency of the system on the air conditioner copper pipe is improved. And a linkage monitoring model is additionally arranged, local deviation in the correction process of the to-be-treated copper pipe is early warned in time, and error accumulation is prevented.

The invention belongs to the technical field of remote sensing image processing, provides a remote sensing image blocking effect elimination method fusing adaptive weight and an anchoring mechanism, and aims to eliminate the blocking effect from a global level, take radiation consistency as a constraint condition, introduce the adaptive weight to carry out differential adjustment on constraint intensity of different overlapped regions, and improve the blocking effect elimination efficiency. Meanwhile, an anchoring mechanism is adopted, stabilization constraint is carried out on an overall solution space, finally, a global equation set is constructed, an iterative least square optimization strategy is used for solving, image radiation consistency and visual continuity are integrally ensured, and the quality of remote sensing images is improved.

This invention discloses a navigation method and system for offshore wind power booster stations based on multimodal information interaction, belonging to the field of artificial intelligence and robot navigation technology. It achieves visual-language modality alignment through the CLIP model, constructing an end-to-end navigation parameter generation model. This model can adaptively generate AGV motion parameters based on natural language commands and scene images, enabling efficient and accurate inspection of offshore wind power booster stations. By establishing a multimodal dataset and associating language commands with AGV motion parameters, this invention solves the error accumulation problem caused by multi-stage processing in traditional methods, improving navigation accuracy and adaptability. Furthermore, this invention employs a multi-task loss function optimization model, including contrastive loss and motion parameter regression loss, enabling the AGV to adaptively adjust motion parameters according to environmental changes, thus improving the efficiency and safety of offshore wind power booster station inspections.

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.

The invention provides a side wall positioning and mounting method and device for an assembly trolley of an overlapped assembly type subway station, and belongs to the technical field of underground engineering construction. The method comprises the following steps: an environment perception and initial positioning step: scanning a construction environment, and determining spatial positions of a side wall and an H-shaped steel groove relative to a preset BIM model through a target identification and point cloud registration technology; a transverse moving and coarse positioning step: planning a path based on a positioning result, and controlling the assembly trolley through real-time pose estimation and visual tracking to enable the side wall to move to a position above the target; an accurate positioning and leveling step: obtaining deviation and real-time attitude through visual detection and multi-sensor fusion technology, and driving an adjusting mechanism to carry out accurate alignment and leveling; a gap detecting and filling step: after the side wall falls in place, measuring the shape of the gap and guiding filling construction; according to the invention, global absolute position calibration is realized, error accumulation is avoided, construction safety is ensured through non-contact measurement, and positioning precision and reliability are ensured by using real-time feedback control.

The application provides a pressure regulating hydrogen storage device, a pressure regulating hydrogen storage method and a composite hydrogen delivery system, which can deliver hydrogen under pressure for a long time and meet the filling demand of hydrogen for hydrogenation load with large capacity. The pressure regulating hydrogen storage device comprises a hydrogen delivery pipeline, a pressure regulating hydrogen storage unit and a temperature regulating liquid bath unit. The pressure regulating hydrogen storage unit comprises a first pressure regulating hydrogen storage tank group filled with a first metal hydride hydrogen storage material and a second pressure regulating hydrogen storage tank group filled with a second metal hydride hydrogen storage material, which are sequentially connected in series along the hydrogen delivery direction of the hydrogen delivery pipeline. The platform pressure of the first metal hydride hydrogen storage material at room temperature is less than that of the second metal hydride hydrogen storage material at room temperature. The temperature regulating liquid bath unit comprises a first temperature regulating bath chamber containing the first pressure regulating hydrogen storage tank group, a second temperature regulating bath chamber containing the second pressure regulating hydrogen storage tank group and a temperature regulating liquid delivery tank in circulation communication with the first temperature regulating bath chamber and the second temperature regulating bath chamber, respectively.

The embodiment of the invention provides a multi-site collaborative electromagnetic spectrum situation prediction method based on multi-source data fusion and an attention mechanism, and the method comprises the steps: 1, collecting spectrum observation data, spatial data and meteorological data of multiple sites, and carrying out the data preprocessing; step 2, based on prior knowledge and GIS data, calculating a theoretical field intensity value of a known emission source at each monitoring station by using a radio wave propagation model; step 3, constructing a multi-branch fusion neural network MFNN based on 3D-CNN, LSTM, GCN and MLP, extracting joint feature representation of multi-source data in the step 1 and the step 2, and introducing an adaptive attention mechanism to perform weighted fusion on each branch feature; and 4, generating a predicted value of a future frequency spectrum state through a prediction output layer, and performing data inversion to obtain a physical power value to realize channel state inference. Structured information capable of directly driving a decision is output through data inversion and channel state inference.

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.

The application discloses a large language model test learning method, device, equipment and medium, test data is input into a large language model with low rank adaptive parameters, the prediction score of each token in the vocabulary at each generation step is obtained; the token-level evidence quality of the generation step is extracted based on the prediction score, and the token-level cognitive uncertainty of the generation step is calculated; the token-level cognitive uncertainty is smoothed, the stable area of the generation step is determined according to the smoothed uncertainty, and the mask of the model update window is constructed; the mask entropy loss is calculated based on the mask, and the low rank adaptive parameters are updated through back propagation. The application can solve the problem of adaptive drift in model updating in related technologies.

The invention discloses an event camera video reconstruction method and system based on active aperture modulation, and the method comprises the steps: introducing an aperture modulation strategy for the first time, and reconstructing an initial frame with good quality by periodically adjusting the opening and closing of an aperture and actively triggering a dense global event signal; the problem of low background reconstruction quality caused by sparse events in a static region in the prior art is solved, and good prior information is provided for subsequent dynamic scene reconstruction. By constructing a forward-reverse bidirectional network, rich intensity information is provided for a static scene, and the problems of background disappearance and error accumulation in long-time operation in a traditional method are effectively solved. And meanwhile, high-time-resolution capture of a dynamic region is kept, and high-fidelity and high-dynamic-range video reconstruction is realized.

The application relates to the field of wind power prediction, and discloses a wind power prediction method and system based on DIRMO and a differentiated target function. In order to solve the defects that a single model is used for prediction in the existing multi-step power prediction, time series modeling and feature processing cannot be simultaneously considered, the prediction accuracy is reduced, the model robustness is insufficient, and the search efficiency is low, the power and wind speed data after preprocessing are denoised, normalized and feature-extracted to obtain multi-scale features; the normalized power and wind speed data are subjected to time series modeling to generate a preliminary prediction result in the future; the multi-step prediction task is divided into several groups, and a multi-output problem is converted into a single-output problem for training and prediction; the hyperparameters of a LightGBM model are automatically optimized, the optimized hyperparameters are used for training the LightGBM models of the groups, and the final power prediction is carried out based on the preliminary prediction result of the GRU and the multi-scale features. The application is mainly used for predicting wind power.

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 discloses a radar echo extrapolation method based on dynamic weight loss, comprising: (1) constructing a dynamic loss function based on the prediction results of an existing short-term forecast model, wherein the existing short-term forecast model is a model that uses radar echo extrapolation to predict precipitation intensity; (2) collecting radar echo data and establishing a radar echo extrapolation dataset, dividing the radar echo extrapolation dataset into a training set and a validation set according to a certain ratio; (3) inputting the training set data into the short-term forecast model for model training, obtaining the dynamic loss function to optimize the short-term forecast model, and inputting the validation set data into the optimized short-term forecast model to obtain the prediction results. This invention effectively improves the prediction performance of current mainstream deep learning models based on radar data and effectively alleviates the cumulative error generated by existing radar echo extrapolation methods.

The invention provides a robot control method and device based on audio driving and a robot, and relates to the technical field of robot motion control. The method comprises the following steps: acquiring audio data, and extracting audio features of the audio data; inputting the audio features into a pre-trained audio motion alignment module for audio motion alignment processing, and outputting audio style latent variables subjected to motion prior alignment; obtaining a motion content latent variable corresponding to the current task semantics; inputting the audio style latent variable and the motion content latent variable into a pre-trained diffusion student strategy network, and generating a joint action instruction of the robot; and sending the joint action instruction to a joint driver of the robot so as to drive the robot to execute an action synchronous with the audio data. According to the embodiment of the invention, through end-to-end direct mapping, multi-stage error accumulation is avoided, and high-fidelity and low-delay synchronous control of the robot action and the audio data is realized.

The application discloses a renewable energy power generation uncertainty tracking method based on a recursive Gaussian mixture model, comprising the following steps: 1, obtaining a deterministic prediction curve of renewable energy power generation, forming a joint observation set containing predicted and actual power generation; 2, setting the cluster number as k, performing a k-means algorithm to obtain k clusters; 3, based on the k clusters, calculating initial parameters of each Gaussian component in the Gaussian mixture model; 4, updating the Gaussian mixture model parameters by using an expectation maximization algorithm to obtain starting optimal parameters; 5, constructing a recursive updating probability model, and based on newly-arrived samples, recursively updating the probability model; 6, using the updated Gaussian mixture model, calculating the conditional probability distribution of the actual power under the condition of the latest predicted power, so as to realize real-time quantification and tracking of the renewable energy power generation uncertainty.

The invention provides a document image correction method and device based on depth guidance, equipment and a medium, and relates to the technical field of document image correction, and the method comprises the steps: obtaining a color image and a depth image of a to-be-corrected document; splicing the color image and the position prompt image to form a multi-channel input tensor; inputting the multi-channel input tensor into the corrective neural network model to obtain a dense displacement field; wherein the correction neural network model comprises a plurality of cascaded geometric attention blocks, and each geometric attention block is configured to generate a rotation position code and a decomposition attention mask of depth perception based on the depth image, and perform geometric self-attention calculation of row-column decomposition on the feature map by using the code and the mask; and finally, sampling the color image according to the displacement field to generate a corrected document image. Geometric modeling is guided through depth information, a row-column decomposition attention mechanism is adopted, the complex deformation correction precision is remarkably improved, the calculation complexity is greatly reduced, and high efficiency and robustness are achieved.

The present application relates to the technical field of underground engineering ventilation and heat and humidity environment control, and discloses a tunnel heat exchange real-time prediction method and system based on physical guidance machine learning, comprising: obtaining standardized simulation data related to tunnel heat exchange and target tunnel measured data and preprocessing to obtain first time series data; based on air enthalpy and annual cycle, day cycle time characteristics, the first time series data is subjected to physical enhancement feature construction to obtain model input features including encoder input features and decoder input features; an LSTM-Seq2Seq-Attention prediction model is constructed, and a two-stage training strategy of simulation data pre-training and target tunnel measured data fine-tuning is used to train the prediction model; the trained prediction model is used to output tunnel outlet temperature and humidity prediction results at multiple future time points at one time; according to the comparison result of the prediction result and the same period measured data, effective measured samples are screened to obtain a model update data set, and the prediction model is periodically fine-tuned and dynamically updated.

The utility model relates to the technical field of material distributing devices, and provides an automatic sorting device for an automatic detector, which comprises a first collecting bin and a second collecting bin which are oppositely arranged, the material distributing part is rotationally arranged between the first collecting bin and the second collecting bin and provided with a first part and a second part which are arranged in a lifting mode, and a containing groove is formed between the first part and the second part and used for providing support for materials; the material distributing piece is provided with a first working position and a second working position, and the first working position is used for receiving materials; when the material distributing piece rotates to the second working position, if the first part descends, the materials fall into the first collecting bin, and if the second part descends, the materials fall into the second collecting bin; compared with a traditional sorting mode depending on a multi-degree-of-freedom mechanical arm or a clamping structure, the automatic sorting mode for the materials has the advantages that the operation steps of grabbing, moving, releasing and the like of the materials are omitted, the treatment period of a single material is effectively shortened, and the operation efficiency of the whole equipment is improved.

The invention discloses a flange production quality screening method and system based on computer vision, and relates to the technical field of computer vision, and the method comprises the steps: obtaining a to-be-detected image of a target workpiece, carrying out the image preprocessing, and obtaining a standardized detection image; the technical key points are as follows: by establishing a quantitative correlation model of curvature and reflection characteristics, spatial directional compensation on visual offset caused by oil film reflection is realized, and the segmentation accuracy under a complex reflection condition is remarkably improved; through a cooperative compensation double-closed-loop mechanism, the two-dimensional visual information reversely improves the three-dimensional deformation modeling precision, the overall measurement stability of the scheme is improved, and joint inversion and mutual feedback correction of optical distortion and geometric projection errors are realized. The related technical problems of edge drift, region missegmentation and size measurement systematic deviation caused by two disturbance coupling are solved, and a more effective guarantee is provided for automatic quality inspection in a current dynamic industrial scene.

This application relates to the technical field of computer vision, and in particular to a method and system for identifying paint formulation components based on multi-angle and multimodal measurement. The method includes: acquiring multimodal measurement data of a paint sample under test from multiple angles, including image data, spectral data, and texture data; extracting and fusing features from the multimodal measurement data to obtain a formulation-level feature vector; determining the types of colorants contained in the paint sample and the presence confidence of each colorant based on the formulation-level feature vector and a pre-constructed colorant standard feature library, using an attention mechanism and a multi-label classifier; and for colorants whose presence confidence meets preset conditions, performing physically constrained regression prediction based on the formulation-level feature vector, the colorant standard feature library, and the colorant presence confidence to determine the content proportion and content prediction confidence of the colorants. This application improves the formulation identification effect and accuracy under multiple colorant conditions.

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.

The application belongs to the technical field of detection, and particularly relates to a karst cave three-dimensional model automatic construction method based on multi-source point cloud fusion, which comprises the following steps: step S100, multi-hole multi-source sensor cooperative arrangement and original detection data acquisition; step S200, single-hole detection data preprocessing and karst cave space contour point cloud data extraction; step S300, multi-hole point cloud constrained by control points is registered and globally fused, the karst cave space contour point cloud data of all detection drill holes are registered and fused, and global karst cave point cloud and fusion quality signals are output; and step S400, a karst cave three-dimensional grid model or a fusion quality insufficient prompt information is output. The present application executes a three-level registration fusion strategy, and has high multi-hole registration precision. The present application effectively solves the error accumulation problem in multi-hole data fusion, ensures uniform distribution of overall fusion precision, and is suitable for large-scale chain-bead string-shaped karst cave detection scenes with many caves and many drill holes.

This invention discloses an installation guidance system and method for prefabricated composite slabs in irregular-shaped formwork systems for public buildings, relating to the field of building technology. It includes a data acquisition module, a calculation and decision-making module, an installation guidance module, and a geometric compensation connection module. The data acquisition module acquires on-site control points, 3D point clouds, and calibration target data, establishes on-site-BIM coordinate correspondence, and outputs original measurement data and quality indicators. The calculation and decision-making module performs rigid body registration and deviation analysis on the point cloud and BIM, solves for the optimal installation posture and tolerance allocation of components, and generates installation command parameters. The installation guidance module projects installation commands to the site using laser, total station, or AR methods, providing real-time prompts for positioning, fine-tuning, and tightening steps and values. The geometric compensation connection module achieves XYZ and angle fine-tuning through adjustable brackets, elliptical slot connections, and edge compressible layers, and reliably locks and cures after grouting while meeting load-bearing requirements. This invention features high accuracy.

The utility model discloses a flywheel drilling special machine, including fixed part and processing part, the fixed part includes Y direction sliding seat and chuck setting on Y direction sliding seat, and the axis of chuck is along Y direction setting, and the fixedly connected with positioning portion has on the chuck, and the positioning portion includes the locating block that can move along Z direction, and the processing part includes the drilling device of being located fixed part both sides, and the X direction sliding seat is equipped with below the drilling device, and the drilling device includes a plurality of rotation hole mechanism, and the flywheel drilling special machine adopting this scheme has the advantages of high processing precision, balanced processing accurate, high processing efficiency.

The application discloses a kind of fusion driving and cascading error attribution hydroelectric generating set multi-subsystem parameter identification method, comprising: construct the high-fidelity simulation model of four cascading subsystems of local control unit LCU, speed regulator, electro-hydraulic servo system, hydroelectric unit;Design dynamic fusion coefficient with iteration attenuation, measured signal and model simulation signal are fused as the input of each subsystem, realize the transition from data-driven to model consistency-driven;Each subsystem output error is decomposed into upstream propagation error and its intrinsic error, and the optimization objective function with intrinsic error as the core is reconstructed;Improved particle swarm optimization algorithm, based on intrinsic error dynamic distribution search weight and introduce propagation error suppression factor, realize multi-subsystem parameter collaborative optimization.The application can effectively suppress cascading error accumulation, improve the physical interpretability and identification accuracy of parameter, and is suitable for hydroelectric generating set digital twin and online control scene.

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.

The application provides an instrument identification method, computer equipment, a storage medium and a computer program product. The method first crops the instrument region in the inspection image, removes background interference, and obtains an image highlighting the instrument. Then, the cropped image is input into an end-to-end correction model to directly generate a front view image through a single forward propagation, avoiding error accumulation of traditional multi-stage methods and improving the robustness of the correction from the architecture. In the model training stage, the reading result of the front view image output by the correction model is taken as a supervision signal by a multi-modal large language model, which on the one hand distills the geometric and semantic understanding ability of the large model to the lightweight model, and on the other hand makes the correction model directly optimize the reading accuracy instead of only focusing on image similarity. This training strategy significantly improves the performance of the model under complex working conditions such as inclination, occlusion and dramatic changes in illumination. Finally, based on the corrected front view image, the reading recognition is performed, which greatly improves the accuracy and stability of the instrument reading.

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.