325 results about "Data domain" patented technology

The invention discloses an energy consumption prediction and scheduling control method based on machine learning. According to the method, operation parameters, energy consumption curves and environment disturbance data of multiple devices are collected through a distributed sensing terminal, the data are input into a pre-trained machine learning model, and a probability prediction result of future energy consumption distribution is generated. On the basis of prediction, an intervention signal is applied in an equipment safety boundary, equipment response characteristics are obtained according to the difference before and after intervention, and an energy consumption risk map is constructed by combining the equipment response characteristics with a probability prediction result. And based on the energy consumption risk map, generating an extreme disturbance scene by using digital twinning, performing consistency check on a probability prediction result and a scheduling scheme in a data domain and a physical domain, and performing multi-stage scheduling in combination with task delays to generate a scheduling result. And finally, issuing the scheduling result to the equipment. The method can improve the accuracy of energy consumption prediction and the reliability of scheduling decision making, and is suitable for intelligent management of data centers, industrial production and high-energy-consumption scenes.

The invention relates to the technical field of information and data processing, in particular to an enterprise digitalization-oriented data management and application method, which comprises the following steps of: acquiring multi-source heterogeneous data and recording metadata; establishing a cross-department management system to standardize the treatment process; building an asset catalog based on the business section and the data field classification; formulating unified data description of a multi-level data standard system; designing a quality control rule to filter low-quality data; constructing a business index system and embedding the business index system into a business system; identifying a cross-system and cross-type association mode through a metadata association technology and a rule engine; and dynamically adjusting the treatment process by adopting a multi-objective optimization algorithm in combination with service feedback and treatment parameters. According to the method, the defects of existing data governance in comprehensiveness, depth and business adaptability are overcome, standardized governance and potential value mining of multi-source data are achieved, and effective data support is provided for intelligent decision making of enterprises.

A system and method for cross-domain knowledge transfer in federated compression networks. The system enables efficient lossless data compression across diverse data types by intelligently sharing compression strategies between domains. A cross-domain knowledge transfer system identifies relationships between different data domains, adapts compression parameters accordingly, and optimizes learning processes to maximize knowledge reuse. The architecture may include a knowledge repository for storing domain features and compression patterns, domain mapping components that identify similarities, and transfer learning optimization that enables efficient adaptation with minimal examples. This approach significantly accelerates model training for new domains while improving compression performance. Applications include satellite telemetry systems where efficient compression is critical for transmitting large information sets between distant locations. The system may employ probability prediction driven arithmetic coding paired with long short-term memory networks, enhanced by cross-domain knowledge sharing that adapts successful compression strategies from one domain to another while preserving domain-specific optimization.

The invention discloses a firmware-level security protection system and method for a hard disk of industrial automation equipment, particularly relates to the technical field of storage security protection, and is used for solving the problem of abnormal tampering of firmware. According to the method, instruction calling, interface access and interrupt timestamps are collected in a sliding window, a behavior sequence set containing an instruction sequence, calling frequency and sensitive marks is constructed and compared with a reference sequence library, firmware-level judgment on operation abnormity, access border crossing and tampering risks is triggered based on a transfer relation and an offset index, and the operation safety is improved. The method comprises the following steps of: dividing firmware into a control domain, a data domain and a configuration domain, realizing minimum permission and dynamic isolation by relying on state and permission mapping and permission register and interface shielding, forming a flow from detection to isolation to closed loop recovery by combining protected segment integrity verification and local copy hot switching, writing an exception into a protected loop log area for threshold updating, and performing protection on the protected loop log area. And the tamper-proof performance, the operation safety and the maintainability of the firmware are integrally improved.

The invention provides an automatic data management method and system based on a large model, and the method comprises the steps: obtaining a native data set corresponding to a to-be-managed data field, carrying out the cross-domain feature association mining of the native data set based on a pre-trained large model, generating a semantic feature map of data entities and an implicit association link between the data entities, and carrying out the data management through the implicit association link. Calling a pre-trained large model, performing quality dimension evaluation based on the semantic characteristic spectrum and the implicit association link, outputting a quality defect vector of a data entity, performing source tracing analysis based on the quality defect vector, and determining a data quality defect type existing in a to-be-treated data field and a cross-entity propagation path of the data quality defect; a strategy evolution module in the pre-training large model is driven to generate a self-adaptive governance strategy containing defect positioning anchor points, and the self-adaptive governance strategy is deployed to a data governance execution environment to complete automatic data governance, so that the governance strategy can better adapt to different data quality scenes; and the pertinence and effectiveness of data management operation are improved.

The invention relates to the field of computer information security and access control, and discloses a block chain power data security sharing method. The method comprises the following steps: acquiring a data field list, a measuring point list and a configuration mapping table, executing naming specification, time anchor registration and de-identification processing, and forming a fixed field sequence and an on-chain index record structure; based on context legality verification and policy mapping, generating an access constraint policy and completing generation and binding of a one-time session token, solidifying a policy snapshot on a contract side and registering a validity period; zero-knowledge proof is constructed, witness materials are assembled, and dual-channel checking, event account falling and failure evidence registration are carried out through a verifiable session smart contract; and an end-to-end key negotiation and control channel is established, so that safe data minimum access and result sealing are realized. According to the method, credible sharing and traceable access of the power data among multiple subjects are realized, and the data security, compliance and chain governance capability are improved.

The invention provides a grain pile surface pest detection model training method, a detection method and equipment, and relates to the technical field of image recognition, and the training method comprises the steps: respectively generating different types of data samples based on grain pile surface image data marked with pest bounding box labels in a plurality of data fields in advance, respectively inputting the two model branches, so that the two model branches sequentially carry out backbone network feature extraction and backdoor adjustment on the data sample to obtain a target feature map and pest detection result data; and determining the causal feature alignment total loss corresponding to the current iteration round according to the target feature maps obtained by the two model branches so as to optimize the two model branches. According to the method, the domain generalization ability of the grain pile surface pest detection model and the adaptability to complex and changeable actual detection scenes can be effectively improved, the application convenience and effectiveness of the grain pile surface pest detection model can be improved, and then the accuracy and reliability of a grain pile surface pest detection result can be improved.

The invention is suitable for the technical field of radar signal processing, and provides a tomographic SAR super-resolution imaging method based on a structured sparse network, and the method comprises the steps: firstly obtaining multi-channel SAR observation data, constructing multi-channel observation vector MMV data in a data domain, and constructing a multi-pixel signal model corresponding to the MMV data based on a neighborhood pixel elevation consistency hypothesis. The method comprises the following steps: designing a kernel principal component analysis KPCA expansion network model, carrying out dimension reduction and enhancement on MMV data, introducing a norm compressed sensing model on the basis of the MMV data after dimension reduction and enhancement, solving the compressed sensing model by using an ADMM iterative algorithm, expanding the ADMM algorithm into a deep network, reconstructing an elevation spectrum, and obtaining an SAR image super-resolution three-dimensional reconstruction result. According to the method, the super-resolution performance and the solving efficiency of three-dimensional reconstruction can be effectively improved, and high-resolution three-dimensional reconstruction of a large-scale scene can be efficiently realized.

The invention belongs to the technical field of seismic exploration and artificial intelligence, and relates to a physical perception collaborative fusion-based full waveform inversion neural network construction method, which comprises the steps of required model data construction, full waveform inversion neural network construction, loss function construction, network pre-training, and network fine tuning and application. Physical feature perception and collaborative fusion are carried out through a U-Net network and an adaptive residual learning module, so that the problem that traditional full-waveform inversion depends on synthetic data training and is not matched with an actually measured data field is solved. The designed pre-training strategy not only can get rid of dependence on a large-scale synthetic data set, but also can effectively relieve the domain offset problem through fine adjustment of measured data, thereby improving the generalization ability of the model. And the adaptive residual learning module enables the neural network to adaptively and dynamically optimize feature expression, and improves inversion precision through collaborative fusion of a physical perception mechanism. According to the method, the physical consistency and imaging reliability of speed model prediction can be remarkably improved while the labor cost is reduced.

The invention relates to the technical field of visual recognition, and particularly provides a bottle sorting-oriented three-dimensional visual posture recognition method and related equipment, and the method comprises the steps: obtaining three-dimensional point cloud data and RGB image data of bottles on a conveyor belt, carrying out the preprocessing, feature extraction and geometric analysis of the three-dimensional point cloud data, and obtaining six-degree-of-freedom posture parameters; and the six-degree-of-freedom attitude parameters are projected to RGB image data, the overlap ratio of the projection contour and the actual bottle contour is calculated, and when the overlap ratio is larger than a preset threshold value, it is confirmed that the attitude is effective, and the corresponding attitude parameters are output to a mechanical arm sorting control system. Six-degree-of-freedom posture parameters of the bottle body can be obtained through posture alignment and template point cloud registration, reliability confirmation of a posture recognition result is achieved through overlap ratio verification in an RGB image data domain, and the defect that the posture recognition result is not reliable under the conditions that the posture change of the bottle body is large, the appearance of the bottle body is diversified and the bottle body is inclined or shielded and interfered is overcome. The problems of insufficient recognition precision and incomplete attitude parameters exist.

The invention provides a dynamic adaptation method and system for a multi-mode power supply compatible charging pile system, and the method comprises the steps: building a multi-protocol physical layer compatible connection between a charging pile and a power supply module through a multi-communication interface layer, and receiving a data frame inputted by the charging pile; analyzing a baud rate, a check bit and an instruction set of the data frame, and identifying a charging pile protocol type and a power module protocol type; querying a matching mapping relationship between the charging pile protocol type and the power supply module protocol type in the dynamic adaptation matrix; if the matching mapping relation exists, data field conversion is executed, and a target protocol frame is generated; and if the matching mapping relation does not exist, starting a machine learning process, creating a temporary mapping rule, and effectively solving the problems of poor compatibility and high manual configuration cost of the charging pile and the power supply module in a traditional protocol mode through the modes of protocol analysis, intelligent matching and dynamic learning.

Disclosed herein are methods, systems, and computer-readable media for prompting a machine learning model to generate answer data based on a recording. Some embodiments involve preprocessing a prompt corresponding to a query for a first system by receiving the prompt and a timestamp corresponding to a time position of the query in a recording, acquiring a text transcript based on the recording, and selecting, based on the timestamp and the text transcript, a first data domain from the text transcript. Some embodiments involve transmitting at least one of the prompt, the text transcript, and the first data domain to a second system, the second system including a machine learning model. Some embodiments involve generating answer data corresponding to the prompt by querying the machine learning model with the prompt, receiving answer data from the machine learning model, and transmitting the answer data to the first system.

The invention discloses a multi-objective collaborative optimization pipe network automatic transmission and distribution control method, and belongs to the technical field of process control and system automation, and the method comprises the steps: obtaining pipe network operation data, generating robust features through adversarial domain adaptation training, monitoring environment mutation in real time, building a dynamic multi-objective optimization model through the robust features, and carrying out the optimization of the dynamic multi-objective optimization model. And generating a Pareto solution set by adopting a dynamic multi-objective evolutionary algorithm, and finally selecting an optimal control solution in combination with a collaborative balance index and a preference function. According to the method, data field differences are eliminated by adopting an antagonistic domain adaptation technology, robust features insensitive to working condition changes are obtained, and an optimal equilibrium solution among energy consumption, pressure safety and service quality can be tracked and determined in real time when a pipe network load and an operating environment dynamically change in combination with a dynamic multi-objective evolutionary algorithm responding to environment mutation, so that the service quality of the pipe network is improved. And the real-time performance, the robustness and the multi-target collaborative optimization level of automatic transmission and distribution control of the pipe network are improved.

The invention discloses a privacy protection unsupervised cross-domain image retrieval method and system, and belongs to the technical field of information security. Aiming at the problems of privacy leakage risk and difficult feature alignment in the existing unsupervised cross-domain image retrieval, the invention provides a dual-protection transformation strategy, and realizes local encryption of image features of each data domain by combining orthogonal projection transformation and a differential privacy mechanism. Furthermore, intra-domain discriminant reinforcement learning and differential privacy cross-domain alignment are respectively performed through a probability prototype optimal transmission mechanism, and an intra-domain and cross-domain loss function is constructed based on a contrast learning normal form, so that safe and effective model training is realized. And after training is completed, image similarity calculation is performed by adopting protected feature representation, so that cross-domain image retrieval under privacy protection is realized. According to the method, the accuracy and robustness of cross-domain retrieval are improved, privacy security and model performance are considered, and the method is suitable for an image retrieval task in a multi-party cooperation scene.

Disclosed is a method comprising: receiving captured seismic data and generating first velocity and first pressure data therefrom; transforming, from a first data domain to a second data domain, the first velocity and first pressure data and thereby generate second velocity and second pressure data; mapping wavenumber gather data from the second data domain to a third data domain; determining envelope ratio scaling data and threshold value data using the second velocity and second pressure data; generating, threshold envelope data using a thresholding operator associated with the threshold value data; using coefficient data associated with the threshold envelope data to estimate vertical component data; transforming, from the third data domain to the first data domain, the vertical component data to generate temporal-spatial data; and generating, based on the temporal-spatial data, noise component data comprised in the captured seismic data.

The invention discloses a water quality prediction method based on fine tuning large language model knowledge driving, and belongs to the crossing field of artificial intelligence technology and environmental discipline. According to the method, a large language model is finely adjusted according to a predefined prompt template through professional knowledge in the water environment field, a knowledge extraction model suitable for the water environment field is formed, and a quintuple is extracted from a text to construct a water environment knowledge graph. Performing graph embedding learning on entities in the knowledge graph by adopting a GloVe model and a residual image attention network, and calculating correlation among water quality indexes through cosine similarity to obtain a knowledge domain correlation weight matrix; meanwhile, the Pearson's correlation coefficient, the weighted grey correlation degree analysis and the dynamic time warping correlation similarity coefficient are used for respectively calculating linear and nonlinear correlation among the water quality indexes, and a data field correlation weight matrix is obtained; and finally, constructing a KG-DW-CLA water quality prediction model jointly driven by knowledge and data to perform water quality prediction. According to the method, the problems that a traditional model is insufficient in domain knowledge utilization and single in data correlation modeling are solved, the accuracy and scientificity of water quality prediction are improved, and the method has high generalization ability and application value.

The invention discloses a data reproduction method and device, a storage medium and computer equipment. The method comprises the steps of obtaining picture information generated based on original data in a specific data field; performing optical character recognition on the picture information, and performing structured data extraction on an optical character recognition result through a preset formatted data extraction model to obtain structured information corresponding to the picture information; and performing original data reproduction according to the structured information and the picture information through a preset visual model to obtain reproduction data corresponding to the original data. According to the method, the limitation of acquiring data from pictures traditionally is broken through, and the acquisition efficiency and accuracy are improved by means of the OCR and formatted data extraction model; the original data are completely and consistently reproduced by combining the visual model with the structured and picture information, so that the automatic processing of data reproduction is realized, and the efficiency and accuracy of data reproduction are improved.

A system and method for neural network confidence regularization is disclosed. A classification system uses a neural network training model to generate prediction. The confidence of the neural network training model is adjusted based on feature prevalence. The system processes mixed data types (binary, categorical, continuous, and date) through type-specific transformations and tensor construction. A tempering factor is calculated from the unweighted sum of features and applied to intermediate neural network outputs. This tempering mechanism reduces model confidence when several low-weight features are present, enabling faster convergence, better generalization, and improved classification accuracy compared to standard neural networks, particularly for complex non-linear relationships in tabular data domains.

The invention discloses an HSI and ranging data cross-domain classification method and system based on domain expansion and feature decoupling, and aims to solve the problems of single mode, insufficient domain offset simulation and weak generalization ability in the prior art. Preprocessing the HSI and LiDAR data to obtain a source domain sample, and generating an expansion domain sample through frequency domain amplitude disturbance and adversarial domain expansion; extracting basic features by using a multi-stream convolutional neural network, and separating domain specific features and cross-modal and cross-domain domain invariant features; and adaptively fusing the three types of features through an attention network, and inputting the preprocessed target domain data into a trained model to obtain a classification result. Through multi-modal fusion, mixed domain expansion and feature decoupling, the cross-domain classification precision and generalization ability are improved, fine adjustment of target domain data is not needed, application is flexible, and the method is suitable for multi-scene remote sensing ground feature classification.

The invention provides a multi-modal knowledge graph generation method and system for hardware design analysis, which break through design, verification and physics, realize information barriers among the three traditional EDA (Electronic Design Automation) fields, perform semantic fusion and association on information from different data fields, create a unified and queriable hardware design knowledge representation, and improve the reliability of hardware design analysis. And a coherent and operable unified knowledge model is formed. A multi-modal design knowledge graph is created by integrating a simulation result (dynamic data) and a physical implementation feature (performance data) into a static structure model, and a comprehensive design model with structure, behavior and performance features is analyzed in a unified framework. An engineer does not need to manually switch between different tools and data formats, but can access all-dimensional information about a static structure, a dynamic behavior and a physical property of a design through a single query interface.

This invention relates to the field of multi-source data processing technology, specifically disclosing a data governance method and system for multi-source heterogeneous data fusion. The method includes receiving unstructured data fed back by a data acquisition agent, performing virtual mapping on the unstructured data, recursively adjusting the access parameters corresponding to the data source node, and generating simulation data based on the virtual mapping when a data source node experiences a data upload anomaly. It also includes receiving data increment density and data hotspot distribution fed back by the data acquisition agent, planning the scheduling path of the unstructured data parsing end based on the data hotspot distribution, and receiving the parsing results fed back by the parsing end. This invention acquires unstructured data by deploying edge nodes, identifies the unstructured data, constructs a data domain model, synchronously determines the data state of the unstructured data, marks hotspot areas in the data domain model, and then determines the parsing order, enabling important data to be parsed faster.

The invention relates to a service providing method for a vehicle. The method comprises the steps that a service request of a passenger of the vehicle is acquired; analyzing the service request to determine a request intent of the occupant; determining at least one involved data field according to the request intention; according to the at least one data field, related data is extracted from the corresponding data field of the vehicle for fusion processing to generate a fusion data table; and based on the fused data table and according to the request intention, executing a corresponding vehicle-mounted application operation to respond to the service request. The invention also relates to a service providing device for a vehicle, a computer program product and a vehicle. According to the invention, the multi-source heterogeneous data generated in the use process of the vehicle is fully utilized, and the unified data table is formed through cross-domain fusion, so that the intelligent response can be made to the complex query request of the passenger. The large table structure after data fusion is beneficial to structured retrieval and correlation analysis, and provides data support for subsequent execution of vehicle-mounted application operation, so that man-machine interaction is simplified, and user experience can be improved.

The application discloses a STCA-MIMO radar anti-interference method based on nonlinear time delay, which comprises the following steps: calculating target echo signals based on nonlinear time delay of transmitting array elements; carrying out digital mixing and matching filtering on the target echo signals to obtain preprocessed signals; determining a first transmitting steering vector and a first receiving steering vector according to the preprocessed signals; calculating a covariance matrix of STCA-MIMO radar system receiving data; calculating an adaptive weight vector by using a robust direct data domain beam forming technology based on the covariance matrix and an actual receiving-transmitting joint steering vector of a target; calculating a first receiving-transmitting joint steering vector of the STCA-MIMO radar system; and carrying out beam forming by using the first receiving-transmitting joint steering vector and the adaptive weight vector. The application can effectively suppress interference signals in the main lobe of the STCA-MIMO radar, and solves the problem of the decline of interference suppression performance in the prior art when the interference is located near the distance grating lobe of the main lobe of the real target.

The invention relates to the technical field of data transmission, and discloses a cross-domain model communication method, device and equipment and a computer readable storage medium. The method is applied to a cross-domain model communication system, the system comprises a user domain, a data domain and a gatekeeper, and the method comprises the following steps: sending request information of a model in the data domain to the gatekeeper through the user domain, and performing security auditing on the request information through the gatekeeper; when the security check is passed, sending the request information to a data domain through a gatekeeper, and generating response information based on the request information through a model; sending the response information to a gatekeeper through the data domain, and performing security auditing on the response information through the gatekeeper; and when the security check is passed, the response information is sent to the user domain through the gatekeeper. When the model carries out data transmission between the user domain and the data domain, the transmitted data is subjected to security auditing through the gatekeeper, and the data transmission is carried out only when the security auditing is passed, so that the security of cross-domain communication of the model is improved.

The application discloses an energy consumption prediction and scheduling control method based on machine learning. The method collects the running parameters, energy consumption curves and environmental disturbance data of multiple devices through a distributed sensing terminal, inputs the data into a pre-trained machine learning model, and generates a probability prediction result of future energy consumption distribution. On the basis of prediction, an intervention signal is applied within the safety boundary of the device, the device response characteristics before and after intervention are obtained, and an energy consumption risk map is constructed in combination with the probability prediction result. Based on the energy consumption risk map, an extreme disturbance scenario is generated using digital twinning, and the consistency of the probability prediction result and the scheduling scheme is checked in the data domain and the physical domain. Combined with the task delay, a multi-level scheduling result is generated. Finally, the scheduling result is issued to the device. The application can improve the accuracy of energy consumption prediction and the reliability of scheduling decision, and is suitable for intelligent management of data centers, industrial production and high energy consumption scenarios.

The system inputs at least one security log file from a first data domain into a first machine learning (ML) model. The system compute, using an output from the first ML model, a weighted sentiment value for one or more of the multiple cybersecurity events. The system detects, using a second ML model, an anomaly in a first cybersecurity event of the multiple cybersecurity events. The system correlates, using a third ML model, the anomaly in the first cybersecurity event to a different anomaly in a second cybersecurity event associated with a second data domain. The system determines, using a fourth machine learning model, a cybersecurity action to minimize a cyberattack. The system determines, using the fourth model, a predicted impact to the telecommunication network associated. The system generates a prioritization ranking of every cybersecurity action and executes each cybersecurity action based on the prioritization ranking.

The invention relates to the technical field of industrial communication and FPGA virtualization, in particular to an FPGA-based virtual Ethercat slave station function implementation method, which comprises the following steps of: pre-storing a plurality of sets of independent virtual slave station parameter configuration tables in a BRAM (Border Random Access Memory) of an FPGA, switching responses according to time slices in a sequence of the configuration tables during topology scanning of a master station, and simultaneously injecting programmable transmission delay to simulate physical link difference; in the operation stage, a data frame logic address is extracted through a hardware analysis engine, multi-slave-station data field modification is completed in a single cycle through a multi-channel parallel processing unit, protocol compatibility is guaranteed through a mailbox command agent, and finally a CRC check code is updated and a response is returned according to a virtual topological sequence. According to the method, a single FPGA supports multiple virtual slave stations, the deployment cost of the multiple slave stations is remarkably reduced, the hard real-time performance and the protocol compatibility are guaranteed, the resource utilization rate is improved, the slave stations can be dynamically added and deleted by updating the configuration table, and the core problems that an existing scheme is high in cost, poor in real-time performance and weak in expansibility are effectively solved.

The invention discloses a jungle scene illumination and clustering degree dual-adaptive animal detection method, and relates to the technical field of image recognition, and the method comprises the following five steps: obtaining a balanced training set with a source mark through data field reforming and rare class condition synthesis in combination with triple quality inspection and progressive matching; implementing scene consistency enhancement of label synchronization, linkage of geometry and illumination disturbance, border-crossing repair and failure rollback; an illumination adaptive converter is embedded in the middle layer of the trunk in parallel, gating residual error fusion is carried out, and a brightness agent amount is exported; according to the brightness proxy amount and the target scale, monotone mapping of a target confidence threshold and an initial intersection-to-union ratio threshold is established, and neighborhood consistency screening is combined; and softening non-maximum suppression by combined scheduling of the clustering degree and the brightness proxy amount, and outputting a detection result. The link gives consideration to low illumination, small target and cluster shielding, improves sample availability and candidate cleanliness, and realizes cooperation from data to a cluster level.

The application relates to the field of data asset management, and discloses a data asset right confirmation management method and system based on a unified data standard. The method comprises the following steps: standard ownership information and data domain verification information are respectively constructed for a plurality of data assets of a plurality of asset nodes, the standard ownership information comprises a plurality of right confirmation field information, and the data domain verification information comprises a plurality of data domain field information; the number of data assets in each data domain is acquired; when the number of first data assets in a first data domain is greater than a preset number, the first data domain is divided into a plurality of sub-data domains each comprising a plurality of first data assets; each sub-data domain corresponds to a plurality of sub-asset nodes; and the data domain field information in the data domain verification information of the sub-data assets is updated according to the ownership relationship between the sub-data assets and the sub-asset nodes. The application realizes standard and normative management of data and facilitates clear property right definition of data assets.

The invention provides an image reconstruction method based on a learnable projection operator and an iterative optimization framework, and belongs to the technical field of medical imaging. The method comprises the following steps: acquiring sparse projection data, recovering processed data by using a projection domain sub-network, and converting the projection data into image domain data through an inter-domain data conversion operator. And then, the image domain sub-network performs optimization processing on the image domain data, converts the optimized data back to the projection domain, and performs iteration until a convergence condition is satisfied. The invention also comprises a consistency constraint framework to jointly optimize the two sub-network modules. In addition, the inter-domain data conversion operator module comprises a position parameter learnable module and a reverse line integral calculation module, so that efficient data domain conversion is realized. According to the image reconstruction method based on the learnable projection operator and the iterative optimization framework, the radiation dose can be reduced, and meanwhile, a high-quality CT image can be provided.