1391results about How to "Improve abilities" patented technology

The invention discloses a pulmonary infectious disease prediction system based on multi-modal data fusion, and aims to improve the accuracy and reliability of disease prediction through comprehensive analysis of multi-dimensional data. The system comprises a data acquisition module, a feature extraction module, a dynamic fusion module and a prediction modeling module. The data acquisition module is used for acquiring clinical and medical images and environmental factor data of a patient in real time; the feature extraction module extracts feature vectors of all data sources; the dynamic fusion module performs adaptive weight distribution on the multi-modal features and fuses the multi-modal features into multi-modal feature representation; the prediction modeling module is used for carrying out dynamic change trend modeling on the multi-modal feature representation through time sequence analysis so as to generate a probability prediction value of suffering from pulmonary infectious diseases and a disease trend prediction result of the patient; the invention provides a comprehensive and efficient disease prediction method by utilizing the collaborative advantage of multi-modal data and combining an attention mechanism and time sequence analysis, and can be widely applied to the fields of medical diagnosis and public health management.

The invention discloses an intelligent risk control system and method for cross-border e-commerce digitization. Comprising an identity verification module, a transaction monitoring and abnormal behavior detection module, a credit assessment and risk quantification module, an anti-fraud and anti-money laundering check module, a compliance auditing module, a user behavior analysis module, a risk control rule engine module and a comprehensive risk assessment module. The problems of how to make a risk control system capable of meeting individual requirements and having cost benefits for small and micro enterprises and how to effectively evaluate and improve the ability of salesmen in the aspect of risk control are solved, and the overall risk control efficiency and accuracy are further improved. Therefore, personalized and intelligent setting of risk control is realized, a visual portrait of the ability of a salesman is provided, intelligent assessment and standard improvement of the ability of risk control personnel are promoted, risk control optimization is realized on the whole, the risk management level of cross-border e-commerce is improved, and enterprises can develop steadily in a complex and changeable international trade environment.

The invention discloses an e-commerce intelligent operation monitoring and collaborative decision-making method based on big data analysis, and the method comprises the following steps: S1, collecting multi-source heterogeneous data, and constructing an original data set; s2, performing data cleaning and characterization processing to generate a feature data set; s3, constructing a knowledge graph and dynamically updating the knowledge graph by using the graph embedding model; s4, generating potential causal paths and weights in combination with the neural relationship reasoning model and the diffusion model, and introducing a microknowledge enhancement layer to generate causal enhancement features; s5, identifying abnormal behaviors in combination with a multi-level anomaly detection mechanism; s6, designing a dynamic game optimization model, determining a collaborative optimal solution under a multi-target conflict through Nash equilibrium, and processing a dominant and subordinate relationship in combination with a Stackelberg game to generate a collaborative optimization strategy; and S7, displaying the monitoring state and the collaborative optimization strategy through a visual interface. According to the invention, big data analysis, a knowledge graph and a dynamic game technology are combined, so that e-commerce intelligent operation monitoring and collaborative decision-making are realized.

The invention relates to a knowledge base question answering method and device and a computer readable storage medium, and belongs to the technical field of artificial intelligence. A query question input by a user is received; the query problem is analyzed, and keywords in the query problem are extracted; retrieving slice data matched with the query problem from a retrieval enhancement generation system on the basis of the text correlation between the keyword and the slice data; wherein a plurality of slice data are pre-stored in the retrieval enhancement generation system, the slice data comprise text slices and image-text slices, and the image-text slices comprise storage paths of images and image descriptions with text context information of the images; and inputting the retrieved slice data into a multi-modal large model, generating an answer related to the query question and an image matched with the answer, and outputting the question answer with the image. According to the method, the capability of an original retrieval enhancement generation system is remarkably improved, and the information requirement of a user in an image-text environment is more comprehensively met.

The invention relates to an intelligent monitoring method and system for an air environment monitoring station, and belongs to the technical field of environment monitoring. The method comprises the following steps: acquiring multi-modal monitoring data, performing data association analysis on the multi-modal monitoring data by adopting a multi-source data association model, and establishing an association relationship among the multi-modal monitoring data; performing data exception monitoring on the multi-modal monitoring data, and triggering and starting an exception type diagnosis event when data exception is monitored; extracting a first data subset related to an exception type from the multi-modal monitoring data; extracting a second data subset associated with the first data subset from multi-modal monitoring data; and performing fault diagnosis based on the first data subset and the second data subset, and outputting an abnormality diagnosis result. According to the method, the multi-source data association model is utilized, the association between different types of data can be effectively analyzed, detailed fault analysis and diagnosis are carried out, the fault response time is shortened, and the operation and maintenance efficiency is improved.

The invention relates to the technical field of geological exploration, in particular to a geological exploration safety monitoring and early warning method and system, and the method comprises the following steps: carrying out the exploration region division based on a geological exploration data set, carrying out the risk assessment and density assignment of geological units in combination with the risk degree, and generating a geological risk density map. According to the method, through geological risk assessment and density assignment, exploration area division and risk assessment, and in combination with risk level judgment of geological units, safety monitoring and early warning in the geological exploration process are more refined and scientific, and by means of stress analysis simulation and terrain sensitivity analysis, the safety of the geological exploration process is improved. According to the geological disaster early warning method, the stress response of the geological unit under natural and artificial loads is revealed, the capability of identifying risk points and sensitive areas in an exploration area is improved, early warning is carried out on geological disasters in the early stage, the real-time performance and pertinence of geological disaster early warning are optimized, and losses caused by the geological disasters are reduced.

The invention relates to a multi-modal large model self-reinforcement learning method based on thinking chain guidance, and the core of the method is to solve the problem that a multi-modal model is complex and poor in reasoning capability at low cost through a model self-generated thinking chain iteration fine tuning model. According to the method, the multi-modal large model self-reinforcement learning method based on thinking chain guidance is proposed for the first time, the multi-modal model is guided to generate an accurate thinking chain through an existing visual reasoning question and answer data set, so that a high-quality complex reasoning fine adjustment data set is constructed, and the complex reasoning ability of the multi-modal model is iteratively enhanced.

The invention discloses an operation site safety management and control method and system based on a multi-modal large model, and the method comprises the steps: collecting site violation data, and carrying out the preprocessing, and obtaining multi-modal data; constructing a multi-modal large model; carrying out training and optimization on the multi-modal large model by utilizing the preprocessed multi-modal data; and using the optimized multi-modal large model to realize operation site safety management and control. According to the invention, a visible light image, a safety violation knowledge base and other modal data are fused, and a comprehensive safety monitoring and management platform is constructed by using a large model and a big data processing technology, so that real-time monitoring, early warning and intelligent response of potential safety hazards of a complex operation site are realized, and the safety management level of the operation site is improved.

The invention belongs to but not limited to the technical field of blind hole machining, and discloses a process method and system for laser composite efficient precision machining of blind holes, and the process method comprises a laser hole prefabricating stage and a laser fine hole repairing stage; a hole is preformed in the punching position of a workpiece through a large-pulse-energy laser beam, and fixed-point materials of the workpiece are rapidly removed; and then the hole is finely trimmed through ultra-short pulse width laser, the incident angle direction of a laser beam and the surface of the material is changed through a rotary cutting system, the hole wall and the hole bottom of the hole after the pre-hole stage are trimmed, fine trimming of the blind hole is completed, and the high-precision blind hole with the controllable taper is obtained. The method comprises the following steps: quickly removing a prefabricated blind hole in a workpiece material by using large pulse energy laser; adjusting process parameters, and quickly pre-processing a blind hole with a certain depth in a workpiece; and then the hole wall and the hole bottom of the blind hole are trimmed through high-speed rotating light beams and variable light beam incidence angles of the rotary cutting system, smooth hole wall, flat hole bottom and controllable taper of the blind hole are achieved, and the machining quality and efficiency of the blind hole of the difficult-to-machine material are improved.

The embodiment of the invention relates to the field of large model security, and discloses a large model optimization method, a related device and a storage medium. The method comprises the steps of determining an application field of a to-be-optimized student large model; calling a preset confrontation model to generate a problem set; inputting the question set into the student large model through a preset confrontation model to obtain an output result; outputting the output result to a preset scoring model for scoring to obtain a score; calling a preset teacher model to train the student large model according to a preset high-quality data set to obtain a training result and an optimization scheme; and respectively feeding back the score, the training result and the optimization scheme to the student large model, and continuously updating the problem set to train and score the student large model until the student large model is gradually iterated to be optimal, thereby obtaining the trained student large model. According to the embodiment of the invention, the overall optimization period of the large model is remarkably shortened, the optimization efficiency of the large model is improved, and the performance of the optimized large model is further improved.

The invention discloses an OTDR (Optical Time Domain Reflectometer) system, which comprises a signal generator; a circulator; a variable optical attenuator which attenuates the received backward light and outputs an attenuated optical signal; the single-photon detector is used for converting the received attenuated optical signal into an electric signal; the photon counter is used for counting the received electric signals; and a controller that restores the backward light incident on the variable optical attenuator on the basis of the count signal and the attenuation amount of the variable optical attenuator. According to the OTDR system, the variable optical attenuator is used for attenuating the backward light, and the attenuated optical signal enters the single-photon detector, so that the saturation of the single-photon detector is avoided, the anti-saturation capability of the single-photon detector is improved, and the technical problem that the anti-saturation capability of the single-photon detector is poor is solved. Moreover, because the single photon detector is accurate in detection, the counting signal output by the photon counter can be ensured to be accurate, and the controller can also be ensured to accurately restore the back light emitted into the variable optical attenuator.

The invention discloses an end-cloud cooperative vehicle-mounted interaction method for multiple task scenes. The method comprises the following steps: (1) training a fine tuning model in an edge executor of a vehicle-mounted end; (2) the vehicle-mounted terminal receives a user instruction and converts the user instruction into a task text instruction; (3) calculating the execution confidence coefficient of the task text instruction, if the execution confidence coefficient is greater than a threshold, entering step (5), and if the execution confidence coefficient is less than the threshold, entering step (4); (4) the cloud planner receives the original task, decomposes the original task into sub-tasks through a cloud large model, converts the sub-tasks into task text instructions and then sends the task text instructions to the edge executor; (5) the edge executor receives a task text instruction, sends the task text instruction into the fine tuning and enlarging model and outputs user instructions meeting different scenes; a scene-related vehicle end API interface is called, and a processing result is returned; and (6) summarizing the returned processing result by the edge executor, and making a final response to a user instruction. The problem that calling interfaces of the vehicle-mounted external tool are not uniform can be solved, and the efficiency of processing a multi-task scene is improved.

The invention belongs to the technical field of automobile information safety fuzzy testing, discloses an automobile information safety fuzzy testing method based on a large model, and develops optimization application research of LLMs in an automobile software fuzzy testing scene by utilizing the advantages of an artificial intelligence technology in the aspects of data processing, natural language interaction, generalization ability and the like. The concealment and effectiveness of the test set generated by the fuzz test are optimized, and the fuzz test process efficiency is improved. Through the implementation of the invention, related key technologies are applied to the fuzz test of the automobile, and other related products and enterprises in China are radiated by writing technical standards, so that the fuzz test system can be used as an application demonstration and sample plate of the fuzz test of the intelligent driving automobile. According to market analysis, fuzzy test calculation is carried out on 100 thousand automobiles every year according to the method, and the data safety and the system safety are improved.

The invention provides a night infrared detection method and system based on an improved lightweight YOLOv8 model, and relates to the technical field of deep learning and target detection. The method comprises the following steps: firstly, acquiring an infrared image data set, preprocessing the infrared image data set, and dividing the preprocessed infrared image data set into a training set, a verification set and a test set; an improved lightweight YOLOv8 model is constructed; according to the improved lightweight YOLOv8 model, on the basis of a YOLOv8 model, a DWR module used for improving the multi-scale and feature extraction capability of an infrared image, a CGAFusion module used for improving the feature fusion capability of the image, and an LSDECDDetect module are added. Training the improved lightweight YOLOv8 model by using the training set, and verifying the effectiveness of the improved lightweight YOLOv8 model by using the verification set in the training process to obtain a trained improved lightweight YOLOv8 model; and inputting the test set into the trained improved lightweight YOLOv8 model to obtain an infrared detection result, thereby improving the vehicle detection accuracy under the condition of poor visibility at night.

The embodiment of the invention provides a self-adaptive detection and detection system for low-altitude flight inspection, and aims to improve the efficiency and safety of unmanned aerial vehicle inspection operation. Firstly, a plurality of unmanned aerial vehicle description information sets, unmanned aerial vehicle cluster description information and flight inspection task description information are acquired; the information covers performance parameters and task requirements of the unmanned aerial vehicles in different inspection scenes, then the information is loaded into a target adaptive inspection decision model for knowledge learning in advance, and target estimation confidence is generated by analyzing the information so as to evaluate the success probability of the unmanned aerial vehicle cluster in executing a specific inspection task. And when the target estimation confidence meets the set requirement, the corresponding unmanned aerial vehicle cluster is intelligently scheduled to execute the inspection task, so that efficient completion of the task is ensured, and the adaptability and flexibility of the unmanned aerial vehicle inspection operation are effectively improved through intelligent evaluation and dynamic scheduling.

The invention discloses a construction method, system and equipment of a power dispatching knowledge question-answering system and a storage medium. The construction method comprises the steps of obtaining a power dispatching semantic document needing to be processed by the power dispatching knowledge question-answering system; sequentially carrying out character-based segmentation, proposition conversion and proposition combination on the power dispatching semantic document on the basis of the large language model and the related cue words to obtain a plurality of text blocks; based on the large language model and the related prompt words, identifying and extracting entities related to power dispatching and relationships among the entities from each text block; constructing a local graph for each text block on the basis of the large language model and the related prompt words, the entities extracted from each text block and the relationship among the entities; traversing each text block, and combining all the local graphs into a global graph; and on the basis of the large language model and the related prompt words, according to the question type queried by the user, a reply text is correspondingly generated by matching a minimum local graph which is most related to the global graph in the global graph.

The invention discloses a network strategy intention translation system and method based on retrieval enhancement. The system comprises an intention acquisition module, a retrieval enhancement module, an intention translation module, a network strategy generation module, a network strategy verification module, a network strategy issuing module and the like. The method comprises the following steps: (1) an intention acquisition module acquires a network strategy intention input by a user in a natural language; (2) a retrieval enhancement module retrieves a retrieval document highly related to the business scene; (3) an intention translation module generates an expected intermediate intention template; (4) a network strategy generation module converts the expected intermediate intention template into a specific network configuration strategy; (5) a network strategy verification module verifies the network configuration strategy; and (6) the network strategy issuing module converts the network configuration strategy into a specific network configuration command and pushes the specific network configuration command to corresponding network equipment.

The invention discloses a flight space directional training auxiliary system and method, and the system comprises a flight task simulation module; the meteorological environment simulation module is used for simulating a meteorological scene corresponding to a flight task, and simulating the meteorological scene by establishing a multi-dimensional meteorological model and combining meteorological parameters; the meteorological environment parameter control module comprises a time step length control module which adopts a preset time step length to input meteorological parameter observation data collected in real time to the meteorological parameter setting module; the specific meteorological event control module is used for transmitting meteorological parameters of a specific meteorological event area to the meteorological parameter setting module when the airplane triggers a specific meteorological event; and the training scheme generation module is used for generating a proper training scheme according to the historical training information of the flight personnel. According to the system and the method, a real and personalized simulation training environment can be provided for the flight personnel, and the professional skills and the capability of coping with various flight conditions of the flight personnel are effectively improved.

The invention discloses a knowledge archiving method, device and equipment, a storage medium and a computer program product, and relates to the technical field of data processing, the disclosed knowledge archiving method comprises the following steps: performing service type classification on an unstructured target demand document by adopting a vector machine algorithm to obtain a first classification result; a basic large model is adopted to generate a document abstract of the target demand document, service type classification is conducted on the target demand document according to the document abstract, a second classification result is obtained, and the basic large model is a large language model obtained through training of historical demand documents and historical document abstracts; determining a target classification result of the target demand document based on the first classification result and the second classification result; and filing the target demand document, the document abstract and the target classification result into a preset knowledge base in a structured form. According to the method and the device, the unstructured business requirement document can be efficiently archived into the structured document knowledge.

The invention provides a training method, device and equipment for a synthetic data generation network and a medium, and the method comprises the steps: carrying out the training process of gradually increasing the task difficulty in three stages from the cross-view consistency training to the space-time consistency training, and then carrying out the combined prediction fusion training in combination with a multi-modal large language network. The network is enabled to fully learn at each stage, so that the capability of the synthetic data generation network is gradually enhanced, the problem of overfitting or training possibly caused by one-step training is avoided, and the synthetic data generation network can fuse multi-modal information such as images and texts. More real and diversified synthetic data with dynamic prediction capability are generated, and the training efficiency is improved.

The invention relates to the field of network security analysis, in particular to a network security situation analysis and prediction method and system based on AI, and the method comprises the steps: obtaining network source data in real time, and analyzing the behavior characteristics of the network source data; constructing an AI analysis model, inputting the behavior characteristics into the AI analysis model, and analyzing danger coefficients of the behavior characteristics through the AI analysis model; when the analysis result is that the danger coefficient is greater than the preset safety threshold, judging that the behavior characteristics meet an early warning condition, and executing an early warning strategy based on the early warning condition; when the analysis result is that the danger coefficient is smaller than the preset safety threshold value, the monitoring duration is determined according to the danger coefficient; constructing a situation prediction model, inputting the behavior characteristics of which the danger coefficients are smaller than a preset safety threshold value and the monitoring duration into the situation prediction model, predicting the network situation of the behavior characteristics in a time range of the monitoring duration through the situation prediction model, and outputting a prediction result; and generating network security situation feedback based on a prediction result.

The invention belongs to the technical field of computer vision and artificial intelligence, and discloses a lightweight wisdom classroom learning behavior detection method and system, and the method comprises the steps: extracting student behavior features in a classroom image, and obtaining preliminary multi-scale features; designing a deep convolutional neural network to perform deep spatial-temporal feature representation on the extracted features, and introducing a focusing modulation module to enhance the features; designing a dynamic detection head to fuse and process features of different scales; and calculating the difference between a prediction result and a real label by using the optimized classification loss function and bounding box loss function, optimizing model parameters, and outputting a final student behavior detection result through the model. And finally, the system performs detailed analysis on the detected student behaviors and generates a visual classroom behavior report. According to the method, more accurate student behavior detection in a classroom scene is realized, a complex classroom environment and diversified student behavior modes can be effectively processed, and the method has relatively high practicability.

The invention provides an efficient article sorting method and system of an intelligent warehousing system, and relates to the technical field of intelligent warehousing systems.The method comprises the steps that warehousing real-time environment data and a target image set for target objects to be sorted are obtained; inputting the target image set into a preset object classification model to determine a target classification corresponding to the target object; determining at least one candidate free goods allocation matched with the target classification, and determining a target transportation path of the target object according to each candidate free goods allocation, the storage real-time environment data and the target classification; determining a target clamping strategy matched with the target classification; and the target path and the target clamping strategy are sent to the AGV and the mechanical arm correspondingly, so that the AGV and the mechanical arm cooperatively sort the target objects. Therefore, the intelligent level of the intelligent warehousing system is remarkably improved, and automatic sorting tasks for various objects can be completed more intelligently.

The invention relates to the technical field of safety control, in particular to an intelligent control method and system for disaster coping with a wind-resistant and fire-resistant rolling shutter door, and the method comprises the following steps: extracting and calculating a wind speed difference value and a temperature gradient value based on a wind speed sensor signal and a temperature sensor signal of the rolling shutter door, carrying out the attribution classification through combining the load pressure data at the two sides of a door curtain, and comparing and judging with the node state of the roller shutter door to obtain a node disaster response priority index. According to the method, the wind speed difference value and the temperature gradient are calculated by extracting wind speed and temperature signals, multi-node cooperation is ensured, transmission shaft operation parameters, infrared signals and the pressure state are analyzed, the operation state is mastered in real time, the power of the transmission shaft is dynamically adjusted to improve adaptability and stability, output of a driving motor and configuration of sealing parameters are optimized, and the sealing performance is improved. The roller shutter door can keep balance between operation and sealing performance under strong wind and high temperature, wind resistance and fire resistance are remarkably improved, disaster loss is reduced, and response efficiency and safety protection capacity are enhanced.

The invention provides a text structured extraction method and system, a terminal and a medium, and particularly relates to the technical field of natural language processing and information retrieval, and the scheme comprises the following steps: obtaining an analysis text of a target text, adding an index number based on the analysis text, and splicing to obtain a target spliced text; obtaining a plurality of training data sets according to the plurality of texts; performing fine tuning on a preset large language model by using the plurality of training data sets to obtain a fine-tuned large language model; and based on the target spliced text, performing processing through a fine tuning large language model, and generating structured content of the target text. According to the scheme, the index numbers are fused into the text paragraphs, the key information of the target text can be pointed through the index numbers, the ability of a large language model to accurately and comprehensively understand the whole text structure is improved, it can be ensured that the structured content of the generated target text keeps the hierarchical structure of all the paragraphs in the text, and the user experience is improved. And the readability and accuracy of the structured content of the target text can be improved.

The invention relates to the technical field of disaster monitoring, in particular to a convective heavy rainfall monitoring and early warning method and system based on satellite remote sensing, and the method comprises the following steps: calculating the cloud top brightness temperature change of a convective cloud cluster in image data based on the detection data of GPM, FY-3G and H8 satellites, analyzing the time sequence of the cloud top brightness temperature according to the cloud cluster dynamic tracking data, and calculating the time sequence of the cloud top brightness temperature; and obtaining a cloud top temperature feature set. According to the method, through the time sequence analysis of the cloud top brightness temperature, tiny meteorological changes can be rapidly captured, the early-stage potential rainfall risk identification capability is enhanced, the classification process of weather conditions is optimized through the introduction of principal component analysis, the risk assessment accuracy is improved, and through the application of the Kalman filter, the risk assessment accuracy is improved. The dynamic tracking of the cloud cluster is more accurate, the VAR model provides higher reliability in predicting the future meteorological threshold, the deep Q network is combined with the real-time satellite data, the early warning information is updated in real time, and an instant and effective disaster prevention countermeasure is provided.

The invention discloses a VSG-VDG-based autonomous control method and device for an AC-DC hybrid micro-grid interconnection converter. The equipment comprises a direct-current micro-grid, an alternating-current micro-grid and an interconnection converter connected with the direct-current micro-grid and the alternating-current micro-grid. The direct current microgrid is provided with a direct current side VDG control module which is used for controlling a direct current converter to adjust direct current power supply output by collecting internal voltage and current signals; the AC micro-grid is provided with an AC side VSG control module which is used for controlling on-off of a switching tube of a three-phase full-bridge inverter by collecting internal voltage and current signals so as to adjust AC three-phase voltage output. The interconnection converter is provided with an interconnection converter control module which is used for connecting the direct-current bus and the alternating-current bus through the bidirectional interface converter and adjusting energy balance in the subnet through bidirectional energy flow. According to the method, the robustness of the AC / DC micro-grid system is enhanced, and the fault self-healing capability of the system under extreme conditions and the transient and steady-state stability of a controlled system are improved.

The invention provides a social worker service method and system based on long-term diseases, and the method comprises the steps: determining a patient and case data thereof, and selecting an evaluation service group; performing pathological assessment and psychological assessment on the patient through the assessment service group; determining a service object according to an evaluation result; determining whether the family member is included in a second service object needing social worker service; determining a current social worker service stage of the first service object and generating a demand for a social worker; according to the requirements, the schedules of the social workers and the skill labels of the social workers, one or more social workers meeting preset conditions are searched for providing social worker services for the patients and / or family members. According to the method, the specific requirements of the patients and the family members are accurately analyzed through the intelligent matching algorithm, high-quality and personalized social work services can be effectively provided for the long-term disease patients and the family members thereof, it is ensured that each patient and the family members thereof can obtain customized service schemes, and the specialty and pertinence of the social work services are greatly improved.

The invention relates to a vector propulsion spherical airship with an adjustable structure. The spherical airship is mainly composed of a vector propelling device, an airship capsule, an airship framework, a flight control module, a posture detection module, an environment monitoring module, a battery and the like. By adjusting the thrust direction of the vector propulsion device and the included angle between the airship framework and the capsule body, accurate control over the attitude and the flight path of the airship is effectively achieved. The design has the advantages of being small in size and compact in structure, and the propelling device is installed on the framework with the adjustable included angle, so that the control process is more flexible. Meanwhile, the propelling device achieves dynamic position adjustment through a framework structure and is prevented from being located in the airship movement direction, and potential safety hazards of a traditional external propeller are overcome. Although only two propelling devices are configured, the system can still provide fine control capability through seven controllable degrees of freedom, and the precision and flexibility of flight control are remarkably enhanced.

The invention discloses a knowledge distillation method driven by an interference generative adversarial network. The method comprises the following steps of: firstly, acquiring an image classification data set formed by three-channel RGB (Red, Green, Blue) format images; secondly, an interference generative adversarial network AD-GAN is constructed, the AD-GAN comprises a teacher model, a student model and an interference generation model, input of the three models is images in the image classification data set, output of the teacher model and the student model is confidence vectors of image category prediction, and output of the interference generation model is images with interference; finally, adversarial training is conducted on the three models, one training batch is divided into an interference generation model training stage and a student model distillation training stage, independent optimizers are used for training of the interference generation model training stage and the student model distillation training stage respectively, the interference generation model serves as a generator in adversarial training, and the teacher model and the student model serve as a discriminator together. According to the method, the generalization ability and the anti-noise performance of the student model are enhanced, and a practical and efficient solution is provided for a resource-limited scene.