304 results about "Model network" patented technology

Disclosed is an industrial question answering model training method based on reinforcement learning and knowledge base matching, comprising the following steps: S1, collecting professional knowledge questions and answers in an industrial field to construct an industrial knowledge base, training a reward model, carrying out, for industrial knowledge questions and answers, matching comparison on outputs of an industrial question answering model and content of the industrial knowledge base, and obtaining reward values on the basis of similarities; S2, sorting the reward values, and using a sorting loss function to train and update parameters of a reward model network; and S3, carrying out industrial question answering model training, incorporating a penalty term for the reward values, and using a reinforcement learning algorithm to train the industrial question answering model multiple times to obtain an optimal strategy. According to the industrial question answering model training method based on reinforcement learning and knowledge base matching of the present invention, the reinforcement learning algorithm is used, and iterative training is carried out multiple times, thereby helping the industrial question answering model to learn and understand industrial professional knowledge and improving the question answering accuracy of the industrial question answering model.

The invention relates to the technical field of image processing, in particular to a dual-branch coding desert segmentation model network structure based on structure state space duality, which adopts multi-dimensional dynamic convolution to replace traditional convolution in the initial stage of an encoder, introduces a mamba2 module based on the structure state space duality into the backbone design of the encoder, and improves the robustness of the encoder. The efficiency and adaptability of the model are remarkably improved, a double-branch parallel design is adopted, one branch uses cavity convolution to extract multi-scale context information, the other branch reinforces feature expression through a mamba2 module, the model is connected in series with a space attention module and a channel attention module between an encoder and a decoder, and the algorithm is more accurate. The method has the advantages that the method is simple and easy to implement, interference of irrelevant information on segmentation results is suppressed, a deformable large kernel attention module is introduced to the tail end of a decoder, and global and local modeling capability of the model in processing desert complex boundary regions is effectively improved by combining flexibility of deformable convolution and global receptive field characteristics of large kernel convolution.

A natural gas pipeline multi-working-condition fault diagnosis method and system based on Bayesian adversarial attack and single-source domain transfer, relating to the technical field of mechanical fault detection and diagnosis. The core of the method is using the transfer learning technology to solve the problem of insufficient generalization ability of existing deep reasoning models when processing pipeline fault diagnosis tasks under different working conditions. The method mainly comprises the following steps: constructing an attack sample generator on the basis of a Bayesian network, wherein the attack sample generator is used for generating, by adding delicately designed tiny disturbance into an input sample, an attack sample that can cause an reasoning model to make an incorrect decision, so as to mine and analyze a defect of the reasoning model; constructing a domain discriminator on the basis of the Bayesian network, wherein the domain discriminator is used for assist in generating a high-concealment attack sample by means of adversarial learning between the domain discriminator and the generator, that is, there is almost no visible difference between the high-concealment attack sample and an original sample; and constructing a classifier on the basis of the Bayesian network, and by expanding the distance between the attack sample and an original decision boundary of the reasoning model, constraining the posterior distribution of network parameters of the reasoning model to be adjusted towards a higher score of the attack sample, thereby enhancing the adaptability and robustness of the model when facing disturbance in different domains. By means of the steps, the present invention effectively solves the problem of missing reporting and false reporting risk improvement caused by poor generalization ability of traditional deep learning models under different working conditions.

The invention provides a geophysical gravity and magnetic anomaly source model construction and rapid forward and reverse modeling method, and relates to the technical field of intelligent three-dimensional geological modeling and simulation. According to geological information of a research area, source body modeling parameters of a gravity and magnetic anomalous field are set for the research area, and a three-dimensional model matrix M capable of describing a plurality of underground anomalous bodies is obtained; constructing a fast forward modeling network, establishing nonlinear mapping from the three-dimensional model matrix to the simulated abnormal data through the fast forward modeling network, and calculating forward modeling response of the three-dimensional model matrix; and constructing an inversion network based on a concurrent module CFTBlock and combining a CNN and a Transform, and establishing nonlinear mapping from the gravity and magnetic abnormal data to a three-dimensional model matrix. The method has the advantages of fast and accurate forward modeling, high-resolution inversion and the like, and is suitable for better interpretation of actually measured gravity and magnetic data.

The invention is suitable for the technical field of short-term power load prediction and intelligent scheduling, and provides a short-term power load prediction method based on improved sparrow search algorithm (ISSA) optimization. The method comprises the following steps: constructing a multi-scene prediction task according to the time resolution and regional seasonal characteristics of a load; local features are extracted in combination with a convolutional neural network (CNN), time sequence dependence is modeled by a long short-term memory (LSTM) network, and an attention mechanism is introduced to strengthen key features; meanwhile, an ISSA is adopted to optimize a model network structure and hyper-parameters, the number of layers, the learning rate and the batch size of the CNN and the LSTM are adjusted in a self-adaptive mode, and the search efficiency and convergence performance of the ISSA are improved through Latin hypercube sampling, cosine annealing, dynamic spiral search and a Levy flight strategy. Simulation results show that the method can maintain high prediction precision under different time resolutions and regional and seasonal conditions, the generalization ability and cross-scene adaptability of the model are enhanced, and a stable and efficient load prediction scheme is provided for power grid dispatching optimization.

The invention relates to the field of computer vision target detection, and provides a chip packaging defect detection method applied to edge equipment based on YOLOv11m, and the method comprises the steps: obtaining a chip packaging defect data set, carrying out the data enhancement of the chip packaging defect data set, and dividing the chip packaging defect data set into a training set, a test set and a verification set; a model network structure is innovated, a backbone network adopts a Starnet network and is combined with C2CGA, SimAM and other modules, feature extraction and key feature capture are enhanced, a neck network introduces a GSConv technology and an improved module based on the GSConv technology, a feature pyramid structure is optimized, a detection head fuses multi-scale features by means of an innovative LWNBDet detection head, and efficient target prediction is achieved. Importing the data set into a detection model for training to obtain an improved model; according to the improved model, the detection speed is remarkably improved while the detection precision is maintained, the size and the calculation amount of the model are reduced, and the feasibility of deployment of edge equipment is improved.

The invention relates to the technical field of geophysical exploration, in particular to a multichannel deep learning magnetotelluric inversion method based on physical information constraint. The method comprises the following steps: generating a synthetic data set containing a geoelectric model and forward modeling response thereof, and adding a noise simulation actual observation condition; constructing a hybrid network architecture combining Transform and U-Net, taking apparent resistivity and impedance phase as dual-channel input, extracting global features by using an encoder, gradually recovering spatial resolution through a decoder, and outputting an underground resistivity model; network training adopts a composite loss function fusing model loss and data loss, and an inversion process is constrained by introducing a magnetotelluric forward modeling physical rule, so that a result is ensured to fit observation data and conform to a physical mechanism; after training is completed, preprocessed actual measurement data are input into the model, and a resistivity image can be directly obtained. The method is used for geological structure identification and reservoir interpretation, and the inversion precision and reliability are effectively improved.

The invention discloses a remote sensing image semantic segmentation method based on geometric perception diffusion guidance, and is applied to the technical field of remote sensing image semantic segmentation. Comprising a training stage and a testing stage, in the training stage, original remote sensing images, nDSM corresponding to each original remote sensing image and real semantic segmentation images are selected to form a training sample set, and a segmentation everything model based on geometric perception diffusion guidance is constructed and trained; comprising an enhanced visual converter encoder, a diffusion prompt module, a segmented everything image prompt encoder, a segmented everything image mask decoder and a prompt level supervision strategy. In the test stage, various channel components of a to-be-detected remote sensing image are input into the trained model, and the model network outputs a remote sensing image semantic segmentation prediction map corresponding to an original remote sensing image. According to the method, multi-modal remote sensing data can be effectively fused, a multi-scale space structure is captured, and full-automatic semantic segmentation is realized, so that the segmentation efficiency and accuracy are remarkably improved.

The invention provides an edge server dynamic activation method and system based on deep reinforcement learning, and the method comprises the following steps: S1, building system models, including building a network model, an energy consumption model, a communication and service delay model and a state switching cost model, the network model including RES and MES; s2, optimizing a network model by calculating the sum of the cost of the energy consumption model, the communication and service delay model and the state switching cost model; s3, constructing a Markov decision process model, and improving RES stability; s4, predicting a traffic load; s5, calculating a baseline value; s6, training the network model through a centralized intelligent scheduling algorithm CDDS to obtain a trained strategy network model; s7, dividing the trained strategy network model through a federated distributed intelligent scheduling algorithm FDDS, and training the model based on federated learning to obtain a DDPG model; and S8, the DDPG model is deployed to each RES.

The invention discloses an OCT (Optical Coherence Tomography) retina image denoising method based on a high-frequency enhanced diffusion model, which is characterized in that a double-branch diffusion model network constructed based on frequency perception Fourier transform attention (FFTA) is used for separating different frequency domains, enhancing specific frequency domain features and fusing the specific frequency domain features into a spatial domain feature map to realize stronger frequency domain perception and processing capability; the invention relates to a time step T coding method for a module of frequency domain attention. The time step T of the diffusion model is used as a parameter to be coded into a frequency domain attention module to participate in self-attention matrix calculation, so that frequency domain feature processing is aligned with the iteration step number of the diffusion model, and different frequency domain information is pointedly processed at different time steps T; the frequency selective hopping mechanism uses pooling operation to obtain a high / low frequency characteristic pattern, so that the network has adaptive frequency domain retention capability for different local parts.

The invention relates to a satellite-ground fusion network multi-target space-time switching decision-making method. The method comprises the following steps: S1, establishing system models based on a real environment and a Starlink satellite constellation, wherein the system models comprise a terminal mobile model, a network model and a communication model; s2, defining performance indexes of service quality and experience quality, and constructing a target function; s3, modeling a connection switching decision process: modeling the connection switching decision process as a Markov decision process; s4, proposing a space-time decision method for connection switching based on the double-mask double-depth Q network: obtaining an optimal switching strategy by utilizing continuous interaction between an intelligent agent and an environment; in the exploration stage and the utilization stage, action masks are used respectively to improve the learning efficiency of the intelligent agent in the high-dimensional action space. According to the method provided by the invention, the total switching frequency in the transmission process is reduced, the average connection duration is prolonged, the average throughput is improved, the delay is reduced, and meanwhile, dense switching in the transmission process is avoided.

The invention discloses an echocardiogram video classification method, system and terminal, and the method comprises the steps: constructing a classification model network which comprises a feature extraction module, a feature enhancement module and a feature aggregation module; obtaining an echocardiogram video, obtaining a plurality of standard section views according to the echocardiogram video, performing interpolation processing and feature extraction on the plurality of standard section views through a feature extraction module, and outputting a plurality of video features; inputting the plurality of video features into a feature enhancement module for aggregation enhancement of spatial features and time sequence features, and outputting a plurality of enhanced features; and inputting the plurality of enhanced features into a feature aggregation module for frame-level feature weighted fusion to obtain a plurality of key frame features, selecting related features from the plurality of key frame features, obtaining fusion features according to the related features, and classifying the fusion features to obtain a classification result of the echocardiogram video. According to the method, the classification accuracy of the echocardiogram videos is effectively improved.

The invention relates to the technical field of weather prediction, and discloses a weather prediction method based on an improved quantum long short-term memory network, which comprises the steps of inputting weather data into a CGRU model to perform spatial feature extraction on the weather data, and then inputting an output hidden state sequence into an HAQLSTM model to perform prediction, the HAQLSTM model is an improvement of a quantum long short-term memory network model, and the quantum long short-term memory network model is an improved quantum long short-term memory network model. A parameterized variable component sub-circuit is adopted, an attention mechanism and residual connection are added, a quantum long-short-term memory network serves as a time modeler, the residual connection enhances information transmission, the attention mechanism dynamically balances the importance of different parts of input data, in the processing process, a self-attention mechanism is adopted to calculate the correlation weight of input sequence elements, and the correlation weight of the input sequence elements is calculated. Four parallel parameterized variable component sub-circuits are used for carrying out key calculation so as to improve the model performance; the problems that an existing weather prediction model network is insufficient in expression ability, and the efficiency of capturing the long-term dependency relationship in the sequence is low are solved.

The invention discloses a digital twin three-dimensional city modeling and analysis method based on deep learning, and the method comprises the following steps: S1, collecting multi-view image data, extracting camera parameters, and building a three-dimensional city coordinate system; s2, constructing a volume modeling network based on a NeRF method, and generating a three-dimensional radiation field; s3, extracting semantic features of an intermediate layer, and constructing a spatial feature field; s4, dividing a local area and forming a local feature block; s5, inputting the local feature block into a quaternary capsule coding network, and generating a quaternary capsule unit containing existence, attitude, state and attribute vectors; s6, constructing a dynamic routing mechanism, and generating a structure capsule unit; s7, constructing a structure capsule atlas and binding city operation state data; s8, performing image rendering in combination with the atlas and the radiation field to generate a three-dimensional city model; and S9, performing state recognition, flow clustering and energy consumption classification analysis on the model. According to the invention, an integrated process of city modeling and state analysis is realized.

The invention discloses a computer system based on a cross-model Internet of Things equipment access efficiency calculation model, which integrates an equipment performance sub-model, a network state sub-model and a response prediction sub-model. The identification module is used for identifying a device inherent identifier of the Internet of Things device to be accessed so as to obtain a response time statistical characteristic value and a historical communication parameter set; the prediction module is used for acquiring network state parameters in real time, calling the response prediction sub-model to calculate expected response time, and optimizing the historical communication parameter set to generate an initial communication parameter set; the comparison module is used for monitoring the actual response delay and the data packet loss rate of the current access equipment and obtaining a delay deviation value; and the adjusting module is used for dynamically adjusting the bandwidth allocation strategy in the initial communication parameter set when a preset condition is met, and synchronously updating the data analysis priority to generate a target communication parameter set. According to the invention, differentiated access control of different types of Internet of Things equipment can be realized.

A method comprises receiving a request to predict a deployment configuration for at least one application, analyzing code of the at least one application to identify one or more additional applications on which the at least one application will depend, identifying a plurality of network paths between the at least one application and the one or more additional applications, and using one or more machine learning algorithms to predict execution times for the at least one application over the plurality of network paths. The predicted execution times for the at least one application over the plurality of network paths are inputted to a network graph model. The network graph model predicts the deployment configuration for the at least one application based at least in part on the predicted execution times, wherein the deployment configuration comprises a subset of the plurality of network paths.

The invention belongs to the technical field of image processing, and particularly relates to an image tampering detection method fusing noise residual errors and compression artifacts and a program product. According to the scheme, noise residual features of compression artifact features of an original image are extracted through an error level analysis technology and an airspace rich model, and then an image tampering detection model with quality adaptability is constructed by combining the two types of features. The network model firstly extracts tampering features from image compression features and steganography noise dimensions through a convolutional layer in an ELA branch and an SRM branch, and then performs multi-scale feature coding through a lightweight MobileNetV2 network; feature interaction enhancement is realized in combination with an SECA attention mechanism; and finally, generating a pixel-level binary mask representing the tampered area through a feature fusion strategy. According to the scheme, the compression characteristic and the noise characteristic of the image can be fully utilized, and the detection precision of the model on the low-quality image and the robustness in a multi-quality scene are improved.

The invention belongs to the technical field of power system scheduling, and particularly relates to a photovoltaic energy storage scheduling method and system based on Gaussian kernel adaptive dynamic programming, and the method comprises the steps: constructing a state space containing multiple optimization targets based on a photovoltaic energy storage system; performing similarity measurement on the state space based on a Gaussian kernel function, and calculating a performance index function value of a non-representative point by using the performance index function value of the representative point; generating a control strategy in a state space through an execution network based on the multiple optimization targets; adopting a model network to predict state transition after execution of the control strategy; performing Gaussian kernel weighting by using the evaluation network to update and iterate the control strategy; and scheduling the photovoltaic energy storage system by using the control strategy after iteration convergence. Compared with the prior art, the method has the advantage that the consumption of computing resources is greatly reduced.

The invention discloses a high-frequency statistical method and application thereof in artificial intelligence large model training or reasoning, and the high-frequency statistical method comprises the following steps: after an ASIC chip and a shared memory are initialized, first statistical data are obtained through self-circulation in an operating system and written into a memory partition corresponding to a first write index value; and starting a timer in the upper-layer application, obtaining the statistical data in the timing period, arranging the statistical data in the timing period, and updating the initial read index value. According to the high-frequency statistical method, accurate millisecond-level even sub-millisecond-level high-frequency statistical data can be provided, the traffic characteristics of a large model network such as the traffic characteristics of each network device in a period of time can be accurately described, accurate data support is provided for optimization of an artificial intelligence large model, and the training efficiency is improved.

The invention discloses an EMSDH-YOLO network-based detectable target detection method for a low-quality image, and relates to the technical field of target detection in industrial automation. According to the method, partial convolution in a fourth C3k2 module of a backbone network of a YOLO11 model and a C3k2 module at the tail end of a Neck network is replaced by efficient multi-scale convolution, and an EMS-ConvBlock module is constructed; adding large-kernel separable convolution into a spatial pyramid structure of a backbone network of the YOLO11 model, fusing a two-dimensional convolution solution idea, and constructing an LSK-SPPF module; the method comprises the following steps: adding the idea of Dynamic Tanh into a C2PSA module of a backbone network of a YOLO11 model, realizing adaptive calibration of attention weight by using hyperbolic tangent gating, and constructing a DyT-C2PSA module; the method comprises the following steps: adding an idea of channel shuffling and space rolling mixing into an up-sampling architecture in a YOLO11 model Neck network, and constructing an EU-SCMixer module; and a loss function of the YOLO11 model is optimized by adopting a Wise-IoU v3 method. According to the method, the detection precision of the mechanical arm on various grabbable objects can be improved when the mechanical arm carries out grabbing operation under the low-quality imaging condition.

The invention discloses a dexterous mechanical arm control method based on uncertainty perception fused with long and short term reward strategy gradient, which comprises the following steps: acquiring mechanical arm state information in real time, and preprocessing to obtain a unified state vector; according to the state vector, obtaining a prediction mean value and prediction uncertainty of a next state through a forward model network; according to the prediction uncertainty, dynamically generating a long and short-term reward fusion weight through an attention network; according to the fusion weight, fusing the long-term reward strategy gradient and the short-term reward strategy gradient, and updating strategy network parameters; in a strategy network updating process, according to a forward model prediction error and a real state error, shielding a short-term reward through a reliability judgment module when the error exceeds a threshold value; and according to the action instruction output by the updated strategy network, the mechanical arm is driven to execute motion. According to the method, the control precision, the convergence speed and the motion smoothness of the mechanical arm in a complex dynamic environment can be improved.

The invention discloses a current transformer error prediction method based on improved adaptive composite mode decomposition. The method comprises the following steps: preprocessing current transformer measurement data based on improved fully adaptive noise ensemble empirical mode decomposition (ICEEMDAN); performing feature selection and reconstruction on the complex components decomposed by the ICEEMDAN based on a multi-scale permutation entropy and a probability density function; further decomposing the reconstructed complex component based on variational mode decomposition (VMD); and constructing a depth prediction model based on CNN-BiGRU, optimizing sub-model network parameters by using a target optimization method, and finally superposing sub-results to obtain a final prediction result of the current transformer. According to the method, ICEEMDAN is comprehensively utilized for signal decomposition, MPE and PDF are utilized for feature selection and reconstruction, VMD is utilized for further decomposition of complex components, and CNN-BiGRU-MHA deep learning architecture is utilized for feature learning and prediction, so that nonlinear and non-stationary signals can be effectively processed, deep-level features can be extracted, a complex dynamic relationship in time sequence data can be captured, and the time sequence data can be obtained. And the prediction accuracy and the generalization ability of the model are improved.

The invention discloses a structured channel pruning target recognition lightweight method, which is used for lightweight reconstruction of a target detection algorithm model, and comprises the following steps executed by computer equipment: S1, in a network model training process, taking a scaling factor of a batch normalization layer as an importance degree standard for measuring a convolution channel, performing linear transformation on each output feature image pixel of the convolutional layer to accelerate model network convergence; s2, calculating a cutting threshold according to the maximum value of the scale factor of the batch normalization layer, and deleting redundant weight connections of which the model performance influence is lower than the cutting threshold through a channel structure pruning compression operation so as to reduce the model parameter quantity and the calculation quantity; and S3, through network re-training fine tuning, model lightweight is realized. According to the method, coding and other operations are not needed, software and hardware implementation and deployment are facilitated, the detection precision can be almost not affected, the comprehensive performance of the model is improved, and the effect of model lightweight is achieved.

The invention discloses an insulator defect detection method based on a deep neural network and saliency, and the method comprises the steps: 1, carrying out the preprocessing of an input inspection power transmission line image, including image denoising preprocessing and image standardization preprocessing; step 2, visual saliency feature construction is carried out on the preprocessed image; 3, fusing the visible light image with the saliency features; step 4, constructing a convolutional neural deep network to perform feature extraction from bottom to top; step 5, insulator defect classification and identification; step 6, insulator defect positioning regression; the problems that in the prior art, the insulator defect detection effect is limited by feature extraction operator precision, and universality is poor are solved; and a deep learning method is used to carry out mass data training, automatic detection of insulator defects is realized, and the precision of the method is greatly influenced by training data and a model network architecture.

A terminal according to an aspect of the present disclosure includes a receiving section that receives network (NW)-trained model information related to a model trained by a network (NW-trained model) and reference model information related to a reference model that is a base of the NW-trained model, and a control section that judges the NW-trained model, based on the NW-trained model information and the reference model information. According to an aspect of the present disclosure, preferable overhead reduction / channel estimation / resource use can be achieved.

The invention provides a fault diagnosis method and system for few-sample incremental equipment in an open set environment. The method comprises the following steps: pre-training a basic model comprising a feature extractor and a cosine similarity classifier based on a basic vibration sample; training a complementary model based on the enhanced sample and the pseudo-increment sample; the complementary model comprises a feature extractor added with a CBAM module and a square Euclidean distance classifier; inputting a basic vibration sample into the basic model and the complementary model at the same time, and fusing the dual-model output probability; a trained dual-model network is obtained through minimizing a loss function; inputting real incremental data into the trained dual-model network, freezing basic model parameters, and finely adjusting the last two convolutional layers and the classifier of the complementary model; and fusing the fine-tuned dual-model output probabilities to generate a final fault classification result. According to the method, intelligent fault diagnosis under the condition of continuously introducing a new type of data set can be realized, the applicable condition is more practical, the robustness is high, and the accuracy is high.

The invention discloses a kidney stone detection system based on an SCC-YOLOX model, and belongs to the technical field of kidney stone intelligent detection.The system comprises a backbone network (Backbone), a neck network (Neck), a detection head (Head), an SE attention module, a CA attention module, a CBAM attention module and an error correction logic judgment module.The kidney stone detection system based on the SCC-YOLOX model is based on a YOLOX model network structure and combines various attention mechanisms, and the kidney stone detection accuracy is improved. The method aims at improving feature extraction capability and detection performance, effectively enhances fusion of feature expression capability and multi-scale semantic information by introducing SE, CA and CBAM attention modules, remarkably improves accuracy and efficiency of kidney stone detection, and reduces model parameter quantity, reduces calculation complexity and improves operation efficiency while keeping integrity of detection information. The built-in error correction logic judgment module realizes further investigation by identifying a specific scene with misclassification so as to improve the performance of the model and reduce the risk of kidney stone misdiagnosis.

The embodiment of the invention discloses a neural network operation acceleration method and device. The method is suitable for a large model network based on a decoder, and comprises the following steps: obtaining a first output result of a matrix multiplication array; the first output result comprises a plurality of matrix multiplication results; obtaining a second output result from the plurality of matrix multiplication results; wherein the second output result is the maximum value of each row of the matrix multiplication result; and performing softmax subsequent calculation on the first output result and the second output result, and outputting a target calculation result. Compared with the prior art, the method has the advantages that the maximum value calculation unit is added in the nonlinear calculation unit of the matrix multiplication array, and the original maximum value calculation in softmax calculation is advanced to the nonlinear calculation unit in the matrix multiplication calculation array. Compared with an existing solution, the cost of one-time data loading in subsequent softmax calculation is reduced, and the precision of an original algorithm is not changed.

The invention discloses a biomass gasification product distribution prediction method based on a hard constraint physical information neural network. The method comprises the following steps: collecting and preprocessing gasification experiment data; constructing a multi-layer artificial neural network model, and converting prior monotonicity knowledge into an inequality constraint combination only related to network parameters by adopting a hard constraint learning mode; model training: aiming at regression loss and regularization loss of experimental data, adopting a constrained particle swarm optimization algorithm to perform network parameter optimization on the model under an inequality constraint combination to obtain a hard constraint physical information neural network model; and predicting biomass gasification input data by adopting a hard constraint physical information neural network model to obtain a biomass gasification product distribution condition. According to the method, the monotonicity knowledge between the biomass gasification products and the key input parameters is embedded into the neural network model, the problem that biomass gasification experiment samples are insufficient can be effectively solved, and a more accurate and more interpretable biomass gasification product distribution result is obtained.

The invention discloses a landslide hidden danger associated element identification method and device based on multi-modal fusion and pruning, and relates to the technical field of geological disaster identification. The landslide hidden danger associated element identification method based on multi-modal fusion and pruning comprises the following steps: acquiring initial data, wherein the initial data comprises aviation optical atlas data, InSAR deformation data and topographic data; carrying out waveband level superposition on the initial data to generate a multi-channel fusion data matrix; performing channel pruning, network reconstruction and fine tuning learning processing on the first lightweight model network based on the multi-channel fusion data matrix to obtain a target lightweight model network; and obtaining to-be-identified data, inputting the to-be-identified data into the target lightweight model network, and outputting landslide hidden danger associated element data. According to the invention, multi-modal data fusion and lightweight pruning technologies can be combined, and the identification precision and real-time processing capability of landslide hidden danger associated elements are improved.