809 results about "Network module" patented technology

The invention relates to the technical field of warehouse-in and warehouse-out management, in particular to a second-hand car warehouse-in and warehouse-out management system and method. The system comprises a vehicle RFID tag module, an intelligent electronic price tag module, a warehouse-in management module, a warehouse-out management module, an inventory monitoring and alarming module, a multi-point RFID positioning network module and an inventory map visualization module, and a unique encrypted RFID tag can be configured for each second-hand vehicle; binding an intelligent electronic price tag for each second-hand vehicle, and receiving verification to generate a verification result; collecting label information when the vehicle enters the parking lot; the label information is verified when the vehicle leaves; receiving the verification result of the intelligent electronic price tag and the vehicle in-library data of the database to lock the position of the abnormal vehicle; dense RFID reader nodes are arranged in the market to collect and output positioning data; and receiving the positioning data and the vehicle state information, generating a real-time inventory distribution map, and providing a scheduling and site planning decision basis for a manager. The scheduling and planning efficiency of managers can be improved.

The invention discloses a lightweight multi-source unmanned aerial vehicle target detection method and system based on a DEYOLO framework, and relates to the field of target detection, and the method comprises the steps: obtaining an unmanned aerial vehicle visible light image and an unmanned aerial vehicle infrared image which are registered, and inputting the images into a pre-trained target detection model; the model comprises a double-flow feature extraction network module which is used for extracting an unmanned aerial vehicle visible light image and an unmanned aerial vehicle infrared image to obtain a visible light feature map and an infrared feature map; the bimodal adaptive feature weighting module is used for performing bimodal adaptive feature weighting and adding on the visible light feature pattern and the infrared feature pattern to obtain fusion features; the lightweight bimodal attention enhancement module is used for performing feature enhancement on the fusion features; and the detection head is used for detecting the enhanced features. According to the method, the calculation complexity is effectively reduced, and the detection precision and the reasoning speed of the model on the low-slow small target and the robustness of the model on a complex scene are remarkably improved.

The invention discloses an image recognition system for industrial part defect detection, and particularly relates to the field of part defect detection, which comprises a multi-axis controllable light source array module, a high-speed polarization camera module, an edge computing node module, a double-branch semantic segmentation network module and a physical constraint post-processing module, according to the invention, through combination of time-sharing stroboscopic illumination and polarization image sequence acquisition, multi-dimensional perception of surface topography and material differences is realized; generating an elevation map and a normal map by using photometric stereo solution, constructing a differential rendering layer reverse matching CAD model, and extracting flash sensitive features; a double-branch U-Net network is adopted to fuse geometric and polarization characteristics, the characterization capability is enhanced through a trans-attention mechanism, and a pixel-level mask is output; and finally, mapping a two-dimensional result to a three-dimensional coordinate system by means of calibration parameters, carrying out geometric verification in combination with a tolerance zone and a height threshold value, and automatically generating a structured defect report containing position, size, grade and visual information.

The invention discloses a time-space alignment-based vehicle-road cloud integrated end-to-end automatic driving device and method, and aims to solve the problems of limitation of a single vehicle intelligent end-to-end algorithm and incompatibility of fusion after vehicle-road collaborative perception. The device comprises a roadside device and a vehicle end, the roadside device extracts roadside features and transmits the roadside features, and the vehicle end compensates transmission delay through a space-time alignment network module, projects the roadside features to a unified BEV space, and inputs the roadside features into a prediction planning network module after dynamic fusion. The method comprises the steps of road side processing, vehicle end processing, space-time alignment, feature fusion and end-to-end planning, and multi-task loss function optimization is adopted. According to the invention, deep fusion of vehicle and road information is realized, heterogeneous sensor scenes are adapted, the vehicle end load is reduced, the long-tail scene sensing and planning precision is improved, and the reliability and safety of automatic driving are enhanced.

The invention belongs to the field of visual tracking, and relates to a multi-target visual tracking method based on cooperation of a dynamic neural field and a neural network, which takes a cross-modal cooperation architecture as a core and comprises a dynamic neural field module based on multi-target trajectory maintenance and shielding matching and an improved MoESDQ neural network module. Meanwhile, a collaborative decision-making mechanism is designed, when the activation peak value of the dynamic neural field is attenuated to a preset threshold value, neural network feature matching is triggered, and disappearance target reproduction correlation is achieved based on cosine similarity. The objective of the invention is to solve the visual tracking capability under the condition of scene and target motion change in a monitoring range, for example, under an intelligent traffic intersection scene. The problems of high ID switching rate, multi-target misassociation and low tracking precision under a real-time tracking background caused by scene change or frequent shielding of vehicles and pedestrians, similar target appearances, transient disappearance and reproduction of the targets and sudden illumination change are solved.

The invention discloses a side information enhancement sequence recommendation method and system based on causal intervention depolarization, and relates to the technical field of recommendation. According to the method, the recommendation accuracy and robustness are improved by capturing the evolution laws of the user interests on different time scales. By introducing a trainable frequency domain filtering weight, noise interference is effectively suppressed, and the stability and generalization ability of user interest representation are enhanced. The influence of false correlation is eliminated by introducing a learnable causal mask matrix and a differential fusion mechanism, depolarization of systematic interference factors is realized, and the fairness and causal rationality of a recommendation result are improved. And specialized modeling of different user behavior modes is realized through the global-local double-gating hybrid expert network module, so that the expandability and the response speed of the system are improved. And deep interaction of cross-modal features is realized through a dual-path feature weighted fusion module, so that the comprehensiveness of user interest description and the personalized level of recommendation are remarkably improved in a multi-source information fusion scene.

The invention relates to a meteorological deduction method and device fusing physical constraints and a neural network, and the method comprises the steps: obtaining multi-source meteorological data, and constructing a spatial-temporal feature input tensor; the spatio-temporal feature input tensor is subjected to standardization processing and then input into a deep learning network model, and a future weather prediction result is obtained; the model extracts time sequence evolution features and space attention features through a neural network module and a space attention module respectively, and integrates the time sequence evolution features and the space attention features in a splicing form; for a forecast task of a future gamma day, a deep learning network model and a physical mode are adopted for prediction respectively, and a splicing time point is determined according to an error minimum principle, so that splicing of prediction results is carried out; when the physical mode is used for prediction, the improved regional numerical weather prediction model is used as a basis, atmospheric basic equation sets are integrated, and weather prediction at future moments is carried out. Compared with the prior art, the method has the advantages that the atmospheric physical law and data driving advantages are fused, and the extreme weather prediction precision and stability are improved.

The invention discloses a digital twinning multi-agent reinforcement learning intelligent decision-making system and method with a secure memory playback mechanism, and the system comprises a digital twinning module which is used for constructing a virtual model and synchronizing the virtual model with a physical entity in real time; the multi-agent reinforcement learning module is used for carrying out strategy learning based on a constrained Markov decision process and balancing performance and safety through a Lagrange multiplier; the safe memory playback module is used for weighting and playing back the experience samples according to the risk and the timeliness so as to improve the learning safety; the reversible grey influence network module is used for causal modeling and reasoning and enhancing decision interpretability; the double-loop self-constraint control module ensures that a control action is always in a physical safety boundary through a barrier function and safety projection; and the convergence and stability criterion module is used for verifying strategy security convergence and system asymptotic stability. According to the method, the problems of strategy border crossing, virtual-real mismatching and the like in the high-risk manufacturing process are solved, and multi-target optimal control under the safety constraint is realized.

The invention is suitable for the technical field of unmanned aerial vehicle cluster cooperative control, and provides an edge calculation distribution system and method for unmanned aerial vehicle cluster cooperative tasks, and the system comprises a multi-heterogeneous neural network module, a local fusion module, a global fusion module, a conflict resolution module, a task distribution module, and a self-adaptive optimization module. According to the invention, by constructing a multi-level semantic feature extraction and fusion architecture and combining an intelligent conflict resolution mechanism and a dynamic task allocation algorithm, efficient intelligent collaborative operation of an unmanned aerial vehicle cluster in a complex and changeable environment is realized, all core algorithms can be executed in a distributed manner in the local of the unmanned aerial vehicle through a distributed edge computing architecture, and the distributed edge computing architecture can be applied to the unmanned aerial vehicle cluster. Dependence on a central node is remarkably reduced, communication overhead and delay are greatly reduced, cooperative work can still be maintained even if part of nodes fail, and perfect balance between robustness and performance is achieved.

The invention discloses an underwater acoustic target classification and identification method based on multi-domain feature deep fusion and multi-platform sonar data learning. The method comprises the following steps: firstly, preprocessing target tracking beam data collected by a multi-platform sonar, and constructing a time domain waveform sample set and a time-frequency spectrogram sample set; secondly, constructing an underwater acoustic target classification and recognition network model which comprises a primary feature extraction parallel double-branch network module and a feature deep fusion and classification recognition module, the primary feature extraction parallel double-branch network module extracts time domain and time frequency features respectively, and the feature deep fusion and classification recognition module realizes feature deep fusion and classification through a bidirectional cross attention mechanism; training the model by adopting a multi-stage training strategy, wherein the training comprises pre-training, special training, end-to-end training and fine tuning and incremental training; finally, unknown target data is preprocessed and then input into the trained model for forward reasoning, and a classification recognition result is output according to the majority voting principle. According to the method, target multi-domain information can be fully mined, multi-source data are effectively utilized, and the model generalization ability and the recognition precision are improved.

The invention discloses a coal mine intelligent coal face control system based on digital twinning. The system comprises an underground entity unit, a ground virtual unit and an interaction unit connecting the underground entity unit and the ground virtual unit. The physical unit acquires data through the sensing network module and uploads the data through the communication module; the virtual unit builds a three-dimensional dynamic model through a twinborn body building module, virtual and real consistency is guaranteed through a model dynamic correction module, then an optimal control strategy is generated through simulation optimization of a twinborn data analysis and decision-making module, and finally an instruction is issued to an execution mechanism module through a control module to form closed-loop control; according to the method, the digital twinborn body completely corresponding to the physical working face is constructed, simulation operation and verification can be carried out on the whole process of the coal mining technology and equipment linkage logic in a virtual space in advance, and therefore potential action interference and logic conflicts are recognized and eliminated before an actual control instruction is issued, the effect of pre-judgment in advance is achieved, and the real-time performance of the coal mining technology is improved. And the look-ahead decision-making capability of the system is greatly improved.

The invention relates to the field of intelligent manufacturing, and provides an intelligent production line autonomous optimization system and method based on digital twinning, and the system comprises a data collector, a data processing module, a digital twinning model, a topology perception parameter self-optimization module and a multi-objective optimization decision engine. The intelligent agent network module is used for constructing an intelligent agent network structure based on a topology theory and accurately describing a complex interaction relationship between equipment; a pure complex-driven knowledge migration mechanism is adopted, and cross-device optimization experience efficient sharing is achieved; a topology collaborative multi-objective optimization protocol is designed, three objectives of capacity, quality and energy consumption are balanced, the system carries out a large number of simulation experiments in a virtual environment constructed by digital twinning, an optimal parameter combination is generated through a reinforcement learning algorithm, and an optimization result is fed back to an actual production line. The system adaptability is improved by 50%-80%, and the learning efficiency is improved by 3-5 times.

The invention discloses an aircraft cluster multi-task scheduling system based on a cooperative game and a working method thereof. The system comprises a simulation environment module, a hierarchical strategy network module, a centralized value network module and a training and execution module. The simulation environment module is used for constructing a multi-agent air combat confrontation environment and generating states, rewards and interaction data required by training; the hierarchical strategy network comprises a shared space-time representation encoder f theta (.), a high-layer strategy network pi H (aHz) and a low-layer strategy network pi L (aLz; aH); the centralized value network is used for receiving global state information in a training stage, estimating the overall return of our agent cluster, calculating a dominant function, and realizing the optimization of the network by minimizing the value loss; and the training and execution module optimizes a centralized value network parameter and hierarchical strategy network parameters theta H and theta L by using a global state St in a centralized training stage, and outputs an air combat decision ai according to local observation independent decisions of each agent in a distributed execution stage.

The invention relates to the technical field of stress monitoring, in particular to a piezoelectric ceramic system for stress monitoring, which comprises a distributed sensing network module, a stress signal processing module, a stress feature extraction module, an intelligent positioning module and a monitoring feedback module. Mechanical stress is converted into charge signals through a piezoelectric ceramic sensor array and a sensor group, spectrum correlation cross validation is carried out on multiple paths of signals through high-fidelity conditioning and digital processing, and abnormal data are effectively eliminated; vibration mode characteristics are extracted through collaborative time-frequency domain analysis, and damage risk parameters are calculated to distinguish damage types; the intelligent positioning module determines a damage area and endows different confidence levels with the damage area in combination with a sensor spatial topological relation and an abnormal association mode; and the monitoring feedback module executes graded early warning. According to the method, the reliability of monitoring data, the sensitivity of early damage identification and the accuracy of damage positioning are remarkably improved, and intelligent graded early warning based on risk confidence is realized.

The invention discloses an intelligent optical sensing system based on multispectral fusion, and relates to the technical field of optical sensing, and the system comprises a multispectral image collection module which is used for synchronously collecting original image data of a target scene under three spectral channels of visible light, near-infrared and short-wave infrared; the space-time registration and preprocessing module is used for carrying out high-precision space-time registration and radiation correction on the images of different spectrum channels; the self-adaptive feature extraction and fusion module is used for extracting multi-scale spectral features from the registered multi-spectral image and dynamically selecting and weighting a fusion strategy according to scene content; and the lightweight decision network module outputs a final target identification and state discrimination result. According to the technical scheme, the time-space consistency of multispectral data acquisition can be realized, the cross-band registration precision is improved, the robustness under the conditions of low contrast, shielding and severe weather is enhanced, meanwhile, the calculation complexity and memory occupation are greatly reduced, and the method is suitable for edge calculation equipment with limited resources.

The invention relates to the technical field of batteries, in particular to a battery health state training method and prediction method, and the training method comprises the steps: obtaining a training feature vector of a battery; performing iterative training on the initial prediction model of the SOH of the battery through the training feature vector until a current total loss function value obtained by a total loss function corresponding to the iterative prediction model is not greater than a total loss function threshold value, and taking the current prediction model as a target prediction model; in each iterative training process, inputting each training feature vector into an SOH mapping network module to obtain an SOH current predicted value and an implicit state vector; and inputting each training feature vector, the SOH current predicted value and the implicit state vector into a degradation dynamic network module to obtain a physical degradation rate, and obtaining a total loss function value according to the physical degradation rate and the total loss function. The target prediction model obtained through the training method improves the prediction precision of the battery health state.

The invention provides a composite material mechanical property prediction method based on data mining, and belongs to the technical field of composite material mechanical property tests.The method specifically comprises the steps that parameter information and mechanical properties of a composite material are collected to form a data set, an initial prediction model is constructed, and the initial prediction model comprises a conversion module and an initial prediction network module; the conversion module is used for correcting the thermal conductivity of the composite material sample and calculating the bearing capacity of the composite material sample, and training and optimizing the initial prediction model by using the data set to obtain a final prediction model; and collecting parameter information of the to-be-predicted composite material, preprocessing the parameter information, inputting the preprocessed parameter information into the final prediction model, and outputting the mechanical properties of the to-be-predicted composite material by the final prediction model. Through the treatment scheme, the accuracy of predicting the mechanical property of the composite material is improved.

The invention discloses an agricultural machinery intelligent scheduling method based on reinforcement learning and dynamic game optimization, and aims to improve the working efficiency and resource utilization efficiency of agricultural machinery. According to the system, a millimeter wave radar, a 4G / 5G network module and a Beidou positioning system are arranged for agricultural machinery, so that real-time acquisition and primary processing of agricultural machinery position, operation state and environment data are realized, and the data are uploaded to a cloud for further analysis. Secondly, designing a multi-level decision-making mechanism comprising a deployment center, agricultural machinery and task points: in the first layer, the deployment center allocates tasks according to factors such as task priority, agricultural machinery load conditions and terrain resistance; in the second layer, a single agricultural machine optimizes path planning according to a specific terrain so as to realize obstacle avoidance. Besides, a revenue model comprehensively considering the job completion degree, the energy consumption and the delay time is constructed and converted into a reward function suitable for being used by a multi-agent reinforcement learning algorithm so as to promote the agents to make an optimal decision. Finally, the invention provides a set of complete intelligent agricultural machinery scheduling management system from data collection, on-line fine tuning, emergency processing and man-machine collaborative management, so that the agricultural production efficiency is effectively improved, manual intervention is reduced, and the development of intelligent agriculture is promoted.

The invention relates to a hyperspectral remote sensing image water body extraction method based on CNN-Transform mixed architecture, and belongs to the technical field of remote sensing image semantic segmentation. Comprising the following steps: preprocessing a hyperspectral image; constructing a feature extraction network, and obtaining multi-scale spectrum-space features; an enhanced multi-scale context attention module is designed, and local detail features of multi-scale cavity convolution and global context features of cross window global attention are fused; an enhanced residual feedforward network module is designed, and the local feature extraction capability is enhanced by using depth separable convolution; and constructing a dual-path feature fusion module, and fusing deep semantics and shallow detail features through a space attention path and a channel interaction path. According to the method, the problem of missegmentation caused by easy confusion of the water body and the shadow / building in the hyperspectral image and the problem of difficulty in effective fusion of local details and global context information are effectively solved, and the fine water body extraction precision and the boundary definition are remarkably improved.

The invention relates to the technical field of information security, and discloses a signature verification system based on a digital certificate. The system comprises a certificate state acquisition module, a distributed verification network module, a local verification agent module and a zero-knowledge verification engine module. According to the invention, most of certificate state verification requests are locally processed by a user by constructing a distributed verification network and cooperating with a local cache and incremental synchronization mechanism, so that the real-time dependence on a centralized online certificate state protocol responder is greatly reduced; the verification performance bottleneck caused by network delay or overhigh load of the center node is effectively overcome, and the processing efficiency of the signature verification service and the service reliability of the whole system are remarkably improved. A zero-knowledge proof verification protocol is introduced, so that when remote verification is necessary, a user does not need to expose a specific serial number of a certificate to be verified to a verification node, the leakage risk of a user service behavior track is fundamentally eradicated, and stronger privacy protection capability is provided.

The invention discloses an artificial intelligence robot control system fused with a world model architecture, and the system comprises a control main system which comprises a shared multi-mode backbone network module, a strategy head, and a world model head. A visual encoder, an ontology perception encoder, a text encoder and multi-mode fusion are integrated in the shared multi-mode backbone network module, and the strategy head is used for generating a current action instruction. According to the method, a self-supervision signal provided by a world model is used as an additional training constraint, the dependence on large-scale teaching data is reduced, and a strategy head generates actions based on representation rich in physical dynamic information, so that the device has the advantages that the decision is more stable when facing environmental noise or uncertainty; error accumulation in a long-range task is remarkably reduced, and compared with a traditional open-loop strategy model, the method has the advantage that the generalization ability of training out-of-distribution scenes is remarkably improved.

The invention relates to the technical field of image processing, and particularly discloses a valve well gas leakage detection system based on image region segmentation. The system comprises an image acquisition module, a preprocessing and enhancement module, a multi-scale feature extraction module, a semantic segmentation network module, a false leakage suppression module and a decision output module, through multi-modal image fusion, semantic segmentation and texture consistency verification, accurate identification and positioning of a leakage area are realized, false leakage interference is suppressed, and the detection reliability is improved.

The invention discloses a reinforcement learning intrusion detection method based on space-time attention and dynamic courses, and the method comprises the steps: carrying out the dynamic weight distribution of input features in space and time dimensions through a multi-level space-time attention mechanism, and enhancing the expression capability of key features; a dynamic curriculum learning strategy is adopted to realize progressive difficulty adjustment of training samples, and the learning ability of the model is gradually improved through initial difficulty estimation, dynamic threshold adjustment and curriculum sample selection; designing a composite reward mechanism to optimize a reward signal, and combining a basic classification reward and a stability reward to promote stable convergence of the strategy; a self-evolution target network module is introduced, and dynamic update management of a target network is realized through performance monitoring, emergency update triggering and frequency self-adaption. A space-time cooperative detection closed loop is constructed, and an end-to-end intrusion detection process is realized through environment interaction, strategy optimization and online detection. The method effectively improves the accuracy and robustness of intrusion detection, and adapts to a complex network attack scene.

The invention discloses a tea quality evaluation method based on spectral image bimodal fusion. The tea quality evaluation method comprises the following steps: acquiring hyperspectral data of different grades of tea; preprocessing the tea hyperspectral image data; acquiring spectrum and image modal information of the tea hyperspectral image; a spectrum-image dual-mode fusion neural network (SIFNet) model is established, and the SIFNet adopts a dual-branch network architecture. A spectrum feature extraction branch is combined with a one-dimensional convolutional neural network and a bidirectional long-short-term memory network module to extract time sequence dependence features of a spectrum mode; an image feature extraction branch introduces a two-dimensional convolutional neural network and a fuzzy logic processing module to extract accurate features and fuzzy features of an image modal, and advantage complementation of multi-modal information is realized through a feature depth fusion technology; the SIFNet model is trained; and evaluating the quality of the tea to-be-detected sample. Through deep coupling and complementary enhancement of the spectrum-image bimodal features, the tea quality grading accuracy is improved.

The invention, which relates to the technical field of phased-array antenna testing, discloses an automatic phased-array testing system comprising a vector network analysis module, an antenna horn and servo control module, a phased-array antenna module, a testing tool module and an upper computer control module. The invention also discloses a phased array automatic test method. The method comprises the following steps: preparing before testing; configuring test parameters; executing an automatic test; processing and storing test data; and finishing the test and resetting the equipment. Full-process automatic testing is achieved through the upper computer automation unit, traditional testing time is shortened, large-scale phased array batch testing is adapted, personal errors are eliminated through the testing tool, the vector network module and the servo module, data credibility is improved, operation is simplified, data are automatically stored, a report is generated, modular design is adapted to multiple types of phased arrays, and testing efficiency is improved. And the cost is reduced, the application is wide, and better use prospects are brought.

The invention discloses a paralytic nursing intelligent early warning system and method based on the Internet of Things technology. The system comprises a multi-modal physiological parameter acquisition module, an edge calculation preprocessing unit, an intelligent data transmission module, a multi-scale time sequence feature extraction module, a space-time diagram convolutional network module, a cross-modal attention fusion module, a multi-task risk prediction module and a model training and optimization module. According to the system, multi-mode data such as electrocardio, blood pressure, blood oxygen, eye movement tracks and voice are collected, preprocessed at an edge end and then transmitted to a cloud end; a multi-scale convolutional network is adopted to extract time sequence features, parameter association is modeled through space-time diagram convolution, cross-modal data fusion is realized by using an attention mechanism, and finally risk classification, anomaly detection and trend prediction are completed through a multi-task network. The early-stage, accurate and explainable early warning of the stroke risk is realized, and the early warning accuracy and clinical practicability are remarkably improved.

The invention relates to the field of power systems, in particular to a power system load frequency adaptive regulation and control system and method, and the system comprises a state observation module which collects a real-time frequency deviation signal to obtain a state vector representing the dynamic state of the system; the strategy network module is used for receiving the vector and outputting a control action of a generator power regulating variable; the system interaction module applies the control action to a controlled system or a simulation model of the controlled system to obtain a next state vector and an instant reward signal; the simulator network module predicts the frequency deviation at the next moment according to the current state vector and the control action; the strategy optimization module is used for updating strategy network parameters by using zero-order optimization based on prediction output; and the hybrid control switching module is used for comparing the absolute value of the frequency deviation with a preset threshold value and selectively outputting a control action or a PI controller signal. The method solves the problems that the prior art depends on an accurate model, reinforcement learning is divergent due to Critic network high variance training, gradient is abnormal, and pure DRL is large in steady-state error and poor in compatibility.

The invention discloses a digital twin-driven intelligent construction site resource dynamic scheduling system and method, and belongs to the technical field of building construction management, and the system comprises a four-dimensional resource topology construction module which constructs a four-dimensional topological graph of construction resources based on UWB and RFID technologies; the mixed engine scheduling module adopts a mixed integer programming engine to generate a reference scheduling scheme in an initial stage, and adopts a reinforcement learning optimization engine to perform dynamic adjustment in an execution stage; the conflict prediction network module is used for establishing a conflict prediction model based on historical data and triggering a resource redistribution mechanism; the virtual interaction visualization module displays a scheduling scheme in a virtual reality environment and supports manual intervention, depth relation modeling, double-engine collaborative optimization and active conflict prevention of construction resources are achieved, the resource utilization rate is increased by 35% or above, the construction period is shortened by 15% or above, and conflict events are reduced by 50% or above.

The invention discloses a water chilling unit small sample fault detection method and system based on transfer learning, and the method carries out the fault detection through an expansion causal convolution module, a dense neural network module and a classification layer which are connected in sequence. A sparse connection strategy is introduced into dense blocks of the fault detection model, and representative connection between far and near layers is only reserved in each dense block, so that the network complexity is reduced, the feature redundancy is reduced, and the feature multiplexing advantage is kept; meanwhile, the method utilizes a transfer learning strategy to transfer labeled data knowledge of a source domain to a target domain, and introduces a meta-learning thought in a fine tuning stage to carry out gradient updating. In addition, feature subsets with high discrimination ability are screened out from an original high-dimensional feature space in combination with importance scores and correlation analysis, so that the detection precision of model training is improved.

The invention discloses an LED lamp group control method and system based on the Internet of Things, and belongs to the field of intelligent illumination control and Internet of Things application. The system comprises a centralized management platform module, a state monitoring and control module, an intelligent linkage module and a communication network module. The centralized management platform module is used for lighting strategy configuration, operation state monitoring and data analysis display; the state monitoring and control module collects lamp operation parameters and executes a control instruction; the intelligent linkage module automatically triggers a preset lighting scene according to the environment sensing data and the system signal; the communication network module constructs a stable data transmission channel and connects all parts of the system. Through modular design and an intelligent linkage mechanism, accuracy and adaptivity of illumination control are achieved, and the management efficiency of the system is improved in combination with data analysis. The system has the characteristics of flexible deployment, high expansibility and intelligent operation and maintenance, and is suitable for large-scale illumination control scenes such as smart cities, commercial complexes and industrial factory buildings.