58 results about "Sequence reconstruction" patented technology

The invention discloses a prefabricated cabin welding seam quality detection method based on self-supervised learning, and the method comprises the following steps: collecting a multi-source welding seam image and process parameters, carrying out the synchronous calibration, and constructing a multi-modal observation matrix; the image is input into a DINOv2 model based on a visual Transform, dense features are obtained, and inter-frame registration and sequence reconstruction are completed; performing difference analysis on adjacent frames, extracting spatial changes and marking potential defects; the dense features and the process parameters of the corresponding time periods are fused, joint features are constructed and classified according to rules, and a preliminary result is output; inputting the defect area and the process parameters into an improved DEER model, extracting an influence path and amplitude, and generating a sensitivity score and a confidence coefficient; final judgment is given in combination with the preliminary result, the sensitivity and the confidence coefficient, and feedback is formed according to judgment and process difference. According to the method, high-precision and explainable detection and process optimization of weld defects are realized, and the method is suitable for online / offline quality control and tracing.

The invention discloses an intelligent strip mine slope rock mass monitoring method based on multi-source sensor fusion, and relates to the technical field of mine engineering safety monitoring, and the method comprises the following steps: laying a reference beacon and an asynchronous sampling array in a strip mine slope area, constructing a cross-scale time-frequency observation layer, obtaining ground surface high-frequency vibration data caused by a heavy mine car, and carrying out real-time monitoring on the ground surface high-frequency vibration data; and establishing a vibration fingerprint baseline, and calibrating the inherent frequency neighborhood of the slope rock mass. According to the method, coupling sensing is realized by constructing vibration fingerprints and inherent frequency neighborhoods, coherent decomposition and shadow sequence reconstruction are combined to decouple data, risk boundaries are established and time correction optimization is carried out, weight dynamic adjustment is fused to extract real deformation response, and finally false signals are suppressed by using a phonon structure and inverse phase micro-pulses. And precise monitoring and dynamic regulation and control of the slope state are realized.

The invention discloses a distributed sensor abnormal event identification method for intelligent traffic, and particularly relates to the technical field of traffic information perception and identification, and the method comprises the following steps: generating an abnormal information initial confidence value through deploying a sensor node with a confidence value dynamic adjustment function; during fusion processing, independent identification priorities of single node anomalies are reserved; after sequence reconstruction of a unified time reference, judging whether a response condition is met or not in combination with trajectory evolution; if yes, an abnormal response process is triggered, a space-time compensation set is constructed based on historical data of the sensing blind area for verification, and finally an abnormal intervention instruction is output and a traffic scheduling system is linked; according to the method, the sensing sensitivity, the fusion accuracy and the response timeliness of the abnormal information in a complex traffic environment are improved, the problem of an identification blind area that single-point abnormity is covered is avoided, state reconstruction and supplementary verification of the sensing blind area are realized, and the rapid intervention capability of an intelligent traffic system on emergencies is enhanced.

The invention relates to the technical field of medical information management, in particular to a chronic disease patient home nursing remote management system which comprises a volatility analysis module, a symbol recognition module, a correlation analysis module, a task decoupling module and a nursing task generation module. According to the invention, by analyzing the fluctuation amplitude of the continuously collected physical sign data in the specific time window, the short-term sudden change trend which cannot be revealed by single-time manually reported data can be explored, and the data change direction is further symbolized to identify a continuous change mode with clinical significance; and the time overlapping degree of the continuous sudden change time periods of different physical sign data can be calculated, so that the synchronous association reaction among various physical sign indexes is accurately identified, the potential composite physiological risk is revealed, and the dependency decoupling and execution sequence reconstruction of the high-risk task are realized based on the quantitative evaluation of the task emergency degree and the delay degree.

The invention discloses a DNA sequence reconstruction method and system based on multi-scale attention and contrast learning, and relates to the technical field of DNA storage data reconstruction. Comprising the following steps: collecting a plurality of DNA sequence copies, screening out abnormal length sequences, and constructing a standardized clustering data set; performing one-hot coding and filling processing on the DNA sequence; extracting context dependent features and cross-sequence variation features; an Inter-Sequence multi-head attention mechanism is constructed to calculate the similarity between the sequences, and a weighted sequence tensor is generated; a global dependency relationship in the sequence is extracted through an Intra-Sequence multi-head attention mechanism; local offset features caused by insertion and deletion errors are extracted through a multi-size convolutional network; inputting a double-layer long-short-term memory network for sequence-level modeling, and outputting base reconstruction probability distribution; and constructing positive and negative sample pairs, calculating comparison loss, combining cross entropy loss to form a joint loss function, and outputting a high-precision DNA sequence reconstruction result. The method has high accuracy and robustness under the conditions of complex noise and multiple types of errors.

The invention provides a Gaussian neural field dynamic scene reconstruction system based on depth consistency constraint, and relates to the technical field of computer graphics, and the system comprises an estimation module which generates a target frame initial depth map; the calculation module reconstructs the point cloud and obtains a point cloud normal direction and a pixel normal direction; the optimization module is used for iteratively correcting the initial depth based on the two types of normal consistency to obtain an optimized depth map; the alignment module is used for determining a scale parameter through regression by taking the first target frame as a reference, and carrying out scale transformation on the depths of other frames to form a consistent depth sequence; the reconstruction module is used for constructing or training a Gaussian neural field based on the sequence and outputting a three-dimensional representation; in addition, the calculation module can contain multi-dimensional wavelets and sparse reconstruction and is used for multi-scale noise suppression and direction weighted fitting. Reference frame selection is based on frame-level quality, geometry and scale stability indexes; according to the system, the intra-frame geometric credibility and the cross-frame scale consistency are improved, ghosting and tearing are reduced, and the stability and integrity of dynamic scene reconstruction are enhanced.

The application discloses a B cell immune repertoire antibody sequence feature analysis and specific screening method and related equipment, which can be applied to the technical field of data processing. According to the application, after the antibody sequencing sequence is subjected to germ line comparison and recognition to obtain corresponding first sequencing Fv sequence and germ line Fv sequence, the first sequencing Fv sequence is subjected to gene integrity filtering and structure integrity filtering, clonotype division and root node sequence reconstruction to obtain a second clonotype collection, and then a first lineage forest is constructed, and according to the second sequencing Fv sequence, the second clonotype collection or the first lineage forest, sequence feature analysis in multiple dimensions is carried out, so that the antibody sequence features can be analyzed from multiple dimensions such as sequence quality, sequence abundance, mutation degree, mutation preference and aggregation degree, the filtered BCR sequence has high affinity, and according to the sequence feature analysis results in multiple dimensions or the first lineage forest, a visualized graph is generated, so that relevant personnel can view the sequence features.

This application provides a Gaussian neural field dynamic scene reconstruction system based on depth consistency constraints, relating to the field of computer graphics technology. The system includes: an estimation module that generates an initial depth map of the target frame; a calculation module that reconstructs the point cloud and obtains the point cloud normal and pixel normal; an optimization module that iteratively corrects the initial depth based on the consistency of the two types of normals to obtain an optimized depth map; an alignment module that, using the first target frame as a reference, determines the scale parameters through regression and performs scale transformation on the depths of the remaining frames to form a consistent depth sequence; a reconstruction module that constructs or trains a Gaussian neural field to output a three-dimensional representation based on this sequence; in addition, the calculation module may include multi-dimensional wavelets and sparse reconstruction for multi-scale noise reduction and direction-weighted fitting of normals; the reference frame selection is based on frame-level quality, geometric, and scale stability indices. This system improves intra-frame geometric reliability and cross-frame scale consistency, reduces ghosting and tearing, and enhances the stability and integrity of dynamic scene reconstruction.

The application discloses an evaporation dish evaporation amount homogenization sequence reconstruction method based on multi-source data fusion, specifically, daily evaporation dish evaporation amount data of each station in a target area, meteorological data and station metadata are screened, automatic secondary quality control is performed on the screened data, and a high-quality basic data set is obtained; a random forest regression and recursive feature elimination cross-validation method is used for selecting features, and finally a partial least squares regression model is used for splicing and interpolation on the daily evaporation dish evaporation amount data, so that the evaporation dish evaporation amount reconstruction sequence is generated; the evaporation dish evaporation amount reconstruction sequence is subjected to homogenization test, breakpoints are identified and corrected, and the evaporation dish evaporation amount homogenization reconstruction sequence is obtained, so that the problem of discontinuous evaporation amount data sequence caused by non-climatic factors such as mixed use of large and small evaporation dishes is solved, and the limitations of traditional conversion coefficient method and multivariate regression equation, such as large difference between stations, insufficient consideration of meteorological conditions and the like, are overcome.

The invention discloses a multivariate data anomaly analysis method based on extended C-Means. The multivariate data anomaly analysis method comprises the steps that (1) anomaly analysis of multivariate time series data is achieved through an extended fuzzy C-Means algorithm and a particle swarm optimization algorithm; (2) putting forward an extended version of an Euclidean distance function and applying the extended version to a C-Means algorithm to calculate the similarity of each variable in multivariate time sequence evaluation; (3) adopting a particle swarm optimization algorithm as a fitness function to calculate an optimal square Euclidean distance; (4) providing a sequence reconstruction technology to quantify an anomaly score of an anomaly level of a data normal structure; and (5) when shape abnormity is detected, eliminating the influence of time migration on similarity evaluation by adopting an autocorrelation coefficient of a time sequence. The multivariate time series data anomaly detection method improves the accuracy of multivariate time series data anomaly detection, and is of great significance to construction of an intelligent damage data platform system.

The application discloses a robust multi-sequence reconstruction method based on maximum posterior probability in DNA storage, and comprises the following steps: decoding each sequence in a cluster by using an improved BCJR decoder; converting each sequence in the cluster into corresponding time sequence representation; determining the weight of each sequence; and deducing a MAP algorithm formula of robust multi-sequence decoding, and integrating the sequence weight into joint posterior probability calculation to inhibit the influence of outlying sequences. By proposing a base IDS channel model, the BCJR decoding algorithm is improved, so that the defects that the existing statistical inference sequence reconstruction method is only suitable for binary error correction code are overcome. Therefore, the application can process the quaternary error correction of the base form in the DNA sequence.

The present application belongs to the field of computer vision and artificial intelligence, and particularly relates to a four-dimensional dynamic scene representation and reconstruction method, system and device based on structural equivalence prior, aiming at solving the problems of poor spatio-temporal consistency, large calculation cost and low long sequence reconstruction accuracy in dynamic scene reconstruction. The present application comprises: decomposing the input four-dimensional voxel sequence into multiple two-dimensional feature planes containing spatial projection and spatio-temporal coupling; introducing structural equivalence prior to fuse the plane features, generating unified and continuous spatio-temporal latent space representation by establishing the internal correlation between spatial geometric gradient and time evolution; finally, using a decoding module containing residual connection to decode the latent space representation, and recovering the high-fidelity four-dimensional dynamic scene layer by layer. The present application effectively improves the consistency and accuracy of dynamic scene reconstruction by establishing an explicit spatio-temporal coupling relationship, while taking into account the calculation efficiency, and has broad application prospects in the fields of autonomous driving, augmented reality, etc.

The present application relates to a method, device, equipment and medium for verifying the commissioning status of power equipment, and relates to the technical fields of power engineering auditing and artificial intelligence. It includes: parsing the submitted review materials using a multimodal document parsing model; inputting the dispatching operation power time series into a variational autoencoder model for sequence reconstruction calculation to obtain the reconstruction error and generate a time series anomaly feature component; generating a trajectory space anomaly feature component based on the spatio-temporal data of project personnel clock-in and the benchmark coordinates of the equipment ledger, and at the same time performing error level analysis on the on-site commissioning image data to generate an image anti-counterfeiting anomaly feature component; submitting the above multi-source heterogeneous data to a multimodal large model for cross-comparison and logical reasoning, breaking the limitation of only relying on the submitted review materials by a single party, identifying forgery behaviors from the source. With the equipment identifier as the only primary key, in the face of false commissioning scenarios with strong concealment and across multiple links, a multi-dimensional cross-verification closed loop is formed to identify deep造假行为 is misspelled. It should be "deep造假行为 is misspelled. It should be "deep forgery behaviors" here. So the corrected translation is as follows: The present application relates to a method, device, equipment and medium for verifying the commissioning status of power equipment, and relates to the technical fields of power engineering auditing and artificial intelligence. It includes: parsing the submitted review materials using a multimodal document parsing model; inputting the dispatching operation power time series into a variational autoencoder model for sequence reconstruction calculation to obtain the reconstruction error and generate a time series anomaly feature component; generating a trajectory space anomaly feature component based on the spatio-temporal data of project personnel clock-in and the benchmark coordinates of the equipment ledger, and at the same time performing error level analysis on the on-site commissioning image data to generate an image anti-counterfeiting anomaly feature component; submitting the above multi-source heterogeneous data to a multimodal large model for cross-comparison and logical reasoning, breaking the limitation of only relying on the submitted review materials by a single party, identifying forgery behaviors from the source. With the equipment identifier as the only primary key, in the face of false commissioning scenarios with strong concealment and across multiple links, a multi-dimensional cross-verification closed loop is formed to identify deep forgery behaviors.

The application discloses a method and device for constructing an extended image sequence, and relates to the technical field of image sequence reconstruction, and mainly aims to improve the low authenticity of the simulated object extended image sequence constructed at present, so that the difference between the simulated object and the real projected object is large, and the authenticity of the forward projection is poor. The method comprises the following steps: obtaining tomography data of a target projected object; performing tomography image reconstruction processing on the tomography data to generate a tomography reconstruction image sequence of the target projected object, the tomography reconstruction image sequence of the target projected object being used for representing the maximum image sequence reconstructed under the current reconstruction condition; and constructing an extended image sequence of a simulated projected object matched with the target projected object based on the tomography reconstruction image sequence of the target projected object, the extended image sequence being used for representing the extended image sequence of the simulated projected object which is not reconstructed and needs to be virtually constructed under the current reconstruction condition.

The present application relates to the technical field of multi-agent enterprise autonomy management, in particular to a human-computer dual-sovereignty enterprise autonomy management system and method based on four AI centers; the system comprises a planning center, a decision-making center, an execution center and an evolution center; the planning center constructs a hierarchical constraint matrix group and performs reversible constraint reduction processing, generates a folding governance core and a constraint source mapping chain; the decision-making center generates an item execution sequence, a resource allocation table and an artificial confirmation node table; the execution center forms an execution state record set; the evolution center calculates a constraint bearing deviation vector and performs inverse reduction deviation write-back processing, outputs a constraint correction vector, an authorized boundary correction table and a governance sequence reconstruction instruction. The present application is used for realizing constraint reduction, execution control and deviation write-back correction in the enterprise autonomy management process.

The invention discloses a method and a system for realizing DNA storage by aiming at multi-rule rotation coding of Chinese texts, and relates to the technical field of DNA storage. Encoding the Chinese text by using a five-stroke font input method; mapping is carried out according to the positive and negative code tables, and an interval balance and dynamic detection strategy is introduced in the mapping process to control GC content distribution in intervals and the probability of occurrence of homopolymers between the intervals. Scattering and recombining the sequence by using block coding, and compressing by using RLE coding, wherein the RLE coding generates a sequence file and a run-length file; performing GC content constraint on the sequence file by using cross coding; and br compression is adopted for the run-length file to further improve the compression ratio. And generating a DNA sequence through rotary coding. According to the method, five-stroke coding, positive and negative code table mapping and sequence reconstruction strategies are introduced, so that the information storage density is higher, the local GC content of the DNA sequence is more stable, the homopolymer length is smaller, the unexpected motif proportion is lower, and the data has higher reliability and safety in the storage process.

This invention belongs to the technical field of superconducting motor quench detection, and more specifically, relates to a superconducting motor quench detection method based on terminal voltage differential and unsupervised autoencoder. The method includes: acquiring the terminal voltage signal of a superconducting magnet; suppressing induced voltage interference through moving average differential processing; and obtaining a preprocessed signal after filtering; inputting this signal into a pre-trained unsupervised autoencoder model, which consists of an encoder and a decoder. The encoder extracts deep features through a convolutional neural network, a bidirectional long short-term memory network, and an attention mechanism; the decoder performs sequence reconstruction; and by calculating the reconstruction error and comparing it with an adaptive threshold, the quench state is detected and located. This invention solves the problem that the quench judgment threshold often relies on empirical settings and is difficult to adapt to the complex and ever-changing dynamic operating conditions of the motor.

The invention discloses difunctional glutathione synthetase based on ancestor sequence reconstruction and application of the difunctional glutathione synthetase, and belongs to the technical field of enzyme engineering and bioengineering. A difunctional glutathione synthetase Anc427 with an amino acid sequence shown as SEQ ID NO: 1 is obtained through an ancestral sequence reconstruction technology, and compared with St-GshF, the obtained difunctional glutathione synthetase Anc427 has the advantages that the melting temperature is increased to 56.2 DEG C from 45.4 DEG C through a stability test and an amplification reaction; at the temperature of 40 DEG C, the half-life period is prolonged from 173.29 min to 3465.74 min; the half-life period at 50 DEG C is 266.6 min, and the thermal stability is remarkably improved. Meanwhile, the activity of the bifunctional glutathione synthetase Anc427 is improved by 17% compared with that of St-GshF, and the conversion rate under the substrate concentration of 100 mM is improved by 13.6%. The industrial application value is extremely high.

This invention relates to the field of data processing and discloses a knowledge-guided adaptive tunnel time-series data augmentation method, system, terminal, and storage medium. The method includes: parsing the original time-series data during tunnel construction to construct a knowledge-guided objective function; subsequently performing variational mode decomposition and controllable reconstruction on the original sequence to obtain a preliminary augmented sequence; introducing explicit constraints on distortion disturbances and prediction delays, and employing a multi-objective optimization strategy to minimize the sequence reconstruction error and lag order, obtaining the final augmented sequence through iterative solution; collecting multi-source time-series data from the tunnel site, inputting the augmented sequence into a prediction model to obtain the predicted value for the next time step; subsequently obtaining the actual value at that time step and concatenating it with historical sequences, and performing augmentation and prediction again on the updated sequence to achieve rolling augmentation and rolling prediction of the sequence. This invention provides an efficient and reliable data augmentation method for time-series prediction tasks in the tunnel field.

The invention provides a metamorphic rock boron enrichment mechanism comprehensive investigation system based on a micro-area in-situ technology. The system comprises a sample target preparation and characterization module, a combined micro-area in-situ analysis module, a data integration and processing module and a space-time correlation and evolution sequence reconstruction module which are connected in sequence. The method comprises the following steps: establishing a uniform space coordinate system, and sequentially calling an electronic probe, a laser ablation mass spectrum, an ion probe and other micro-area analysis instruments to carry out collaborative analysis on the same sample target to obtain data of mineral components, trace elements, boron isotopes and zircon U-Pb-Hf isotopes. The system carries out automatic fusion and mineral generation intelligent identification on multi-source data, couples a geochemical model and a machine learning algorithm, constructs a boron element geochemical behavior evolution sequence model taking absolute time as an axis, and realizes quantitative description of the whole process of source, migration and enrichment of boron in an open system. According to the method, the systematicness and the accuracy of the study on the boron enrichment mechanism of the metamorphic rock are remarkably improved.

This invention relates to the field of artificial intelligence technology and discloses a speech extraction method, apparatus, device, and medium. The method includes: first, acquiring reference speech of the target speaker and mixed speech of all speakers; preprocessing and encoding the reference speech and mixed speech to generate two discrete token sequences; fusing the two discrete token sequences to form a fused discrete token sequence; using a language model to predict the fused discrete token sequence to generate candidate discrete token sequences for the target speaker; calculating the probability distribution of the candidate token sequences using a linear classifier and selecting the sequence with the highest probability as the target discrete token sequence; and then reconstructing the target discrete token sequence into a speech waveform to obtain the speech of the target speaker. This invention transforms the complex audio generation problem into a classification problem, simplifying model training; and utilizes the sequence modeling capability of a language model to capture long-term dependencies between speech tokens, achieving high-quality speech reconstruction.

The application relates to the field of data processing and discloses a knowledge-guided adaptive tunnel time series data enhancement method, system, terminal and storage medium.The method comprises the following steps: analyzing original time series data in a tunnel construction process and constructing a knowledge-guided target function; then performing variational mode decomposition and controllable reconstruction on the original sequence to obtain a preliminary enhanced sequence; on this basis, explicit constraints of distortion disturbance and prediction delay are introduced, and a multi-objective optimization strategy is adopted to minimize sequence reconstruction error and lag order, and the final enhanced sequence is obtained through iterative solution; multi-source time series data of a tunnel site are collected, the enhanced sequence is input into a prediction model to obtain a next-time prediction value; then, the real value at this time is obtained and spliced with a historical sequence, and the updated sequence is subjected to enhancement and prediction again, so that rolling enhancement and rolling prediction of the sequence are realized. The application provides an efficient and reliable data enhancement method for a time series prediction task in the tunnel field.

Provided in the embodiments of the present application are a sequence reconstruction model training method and apparatus, and a sequence reconstruction method and apparatus, which methods and apparatuses are used for performing training on the basis of normal samples and abnormal samples so as to obtain a model that can be used for sequence reconstruction. Specifically, the capability of the model for reconstructing a normal sequence can be improved on the basis of adversarial learning. The sequence reconstruction model training method comprises: acquiring a first training set, wherein the first training set comprises normal samples and abnormal samples, the normal samples and the abnormal samples each comprising at least one element arranged in the chronological order of generation, and the abnormal samples each comprising a sequence having an abnormal point; and then, using the first training set to update an initial sequence reconstruction model, so as to obtain an updated sequence reconstruction model, wherein the sequence reconstruction model comprises an encoding module, a first decoding module and a second decoding module, the first decoding module being configured to use the normal samples to perform a first update on the encoding module, and the second decoding module being configured to perform a second update on the encoding module on the basis of the abnormal samples by means of adversarial learning.

The application is suitable for the technical field of medical image analysis and artificial intelligence assisted diagnosis, and provides a breast cancer neoadjuvant chemotherapy efficacy prediction method based on DCE-4DNeRF, first, a DCE-4DNeRF model is used, non-uniformly sampled original DCE-MRI sequences are reconstructed into uniformly sampled sequences which are continuous in time and space through spherical harmonic functions and projection mechanism; second, a tumor perception prediction network is constructed, the network introduces biological position coding based on anatomical prior, and combines a differentiable sampling mechanism, so that the network can automatically focus on the key tumor area without manual segmentation; finally, the dynamic spatio-temporal features of the two stages before and after chemotherapy are fused to predict the efficacy. The application effectively overcomes the data time inconsistency, realizes the end-to-end and label-free accurate prediction, significantly improves the prediction performance, and provides a reliable basis for clinical individualized treatment decision.

The application discloses a V2X trust management system based on Mamba and a block chain, which utilizes a deep neural network Mamba deployed at a base station to detect whether a basic safety message broadcast by a vehicle is abnormal in an unsupervised sequence reconstruction manner. A detection result is converted into a 'positive interaction' or 'negative interaction' record representing a vehicle behavior and stored in a block chain module jointly maintained by the base station. When it is necessary to evaluate vehicle trust, the base station inquires all interaction histories of the vehicle from the block chain module and calculates a dynamic trust value of the vehicle based on a Beta distribution expectation. The application innovatively combines a powerful abnormal detection capability of deep learning and decentralized and tamper-proof characteristics of a block chain, realizes effective identification of attacks in a vehicle network and distributed trust management, and solves a single-point failure problem of a centralized scheme.

The application provides a method and device for predicting the trajectory of an object, and belongs to the technical field of computer vision and image processing. The method comprises the following steps: reconstructing a multi-view time sequence two-dimensional image sequence of an acquired target object to obtain an original particle point cloud sequence, and generating a sparse key point cloud sequence after sampling; constructing a particle graph sequence based on the spatial position of a key point at each moment in the sequence, and performing spatial semantic completion; performing object-level dynamic space-time aggregation on the key point feature sequence after completion to obtain aggregated key point features, and inputting the aggregated key point features into a particle graph converter to output key point update features after long-range force propagation modeling by the converter; and decoding the key point displacement at a future moment based on the update features, and generating the future motion trajectory of the target object according to the key point displacement. A device for predicting the trajectory of an object is also provided based on the method. The application can simultaneously predict the fine appearance and stable and accurate motion trajectory of an object.

The invention discloses a time-space consistency data reconstruction method based on a generative adversarial network, and relates to the field of time-space data processing and modeling, and the method comprises the steps: firstly obtaining target time-space data containing a plurality of time slices, and constructing a missing mask corresponding to the target time-space data; extracting spatial texture and time evolution features through a spatial-temporal feature coding module of three-dimensional convolution, and respectively inputting coding features into a local detail reconstruction path and a global time sequence reconstruction path to obtain a local reconstruction result and a time sequence reconstruction result; dynamically fusing the outputs of the two paths by adopting an attention mechanism to generate preliminary reconstruction data of the current time slice; according to the method, true and false discrimination is performed on a true sequence and a reconstructed sequence of a continuous time slice through a time sequence consistency discrimination network, and parameters of a generation network and the discrimination network are updated under an adversarial training framework, so that the generation sequence has stronger cross-frame consistency and structural stability; and automatic completion can be realized only by inputting the spatio-temporal data containing the missing region and the mask thereof.

The embodiment of the invention provides a training method and device of a battery performance degradation trend prediction model, and the training method of the battery performance degradation trend prediction model comprises the steps: obtaining the discharge capacity data of a battery in a historical period, and dividing the discharge capacity data into a plurality of data blocks according to different time scales; inputting the plurality of data blocks into a sequence reconstruction network of a to-be-trained prediction model for processing, and calculating a first loss value according to an output result of the sequence reconstruction network and a predetermined first label; determining training sample data of the current training round according to the training stage to which the current training round belongs; inputting the training sample data into a prediction network of a to-be-trained prediction model for processing, and calculating a second loss value according to an output result of the prediction network and a predetermined second label; and adjusting model parameters of the to-be-trained prediction model based on the first loss value and the second loss value, and generating a battery performance degradation trend prediction model.

The present application relates to the field of computer vision, and discloses an end-to-end online three-dimensional reconstruction method based on three-dimensional Gaussian blocks. The method takes three-dimensional Gaussian blocks as basic units, and alternately executes intra-block loop and inter-block loop processes. The intra-block loop uses block memory reading, Gaussian regression and block updating modules to perform pose-free end-to-end reconstruction and block updating based on the current frame and memory features. The inter-block loop starts when the blocks are filled, aligns the coordinate systems of adjacent blocks through explicit geometric guidance, and transfers appearance features using implicit appearance guidance. The present application realizes real-time online reconstruction of pose-free video streams, effectively solves the problems of geometric drift and appearance break in long sequence reconstruction, and balances the granularity and generalization of reconstruction.

The application provides a knowledge and data double-driven broken mill equipment health management industry model construction method, and belongs to the technical field of artificial intelligence.In the application, firstly, broken mill equipment multi-source knowledge and operation and maintenance data are acquired, knowledge extraction, sequence reconstruction and unified vectorization processing are performed, a mixed vector with semantic alignment is formed, differences of heterogeneous data are eliminated, and standardized input is provided for a model;then, a neural engine is constructed based on a Transformer, a symbolic reasoning engine is constructed in combination with an expert rule base and forward reasoning, an interactive checkable double-engine collaborative reasoning architecture is formed, data generalization and mechanism constraint are considered;next, real-time operation and maintenance data and query input architecture are constructed, a result is generated by the neural engine, and the symbolic engine is checked, iteratively modified, and a reliable reasoning sample is obtained;finally, a health management instruction set is constructed, a low-rank adaptive fine-tuning model is adopted, RAG retrieval and thinking chain prompting are fused, and finally integrated and highly reliable broken mill equipment health management industry model construction is completed.