37results about How to "Strong representation ability" patented technology

The invention provides a pig physical appearance scoring method and device, and the method comprises the steps: carrying out the filtering, registration and segmentation of the point cloud data of a to-be-detected pig, and obtaining the boundary segmentation data and body size data of a first part of the to-be-detected pig; processing the image data based on a target detection model to obtain an identification result of the second part of the to-be-detected pig; scoring the first part according to the boundary segmentation data, and scoring the body length and the body height of the to-be-measured pig according to the body size data to obtain a first scoring result; scoring the second part according to the identification result to obtain a second scoring result; and obtaining a target scoring result according to the first scoring result and the second scoring result. According to the method, the cloud processing technology is combined, multi-dimensional physical features with higher characterization capability are extracted based on a multi-head self-attention and deformable network target detection model, and the accuracy of pig physical feature evaluation is improved.

The invention provides a method for converting a multi-stage non-iodine contrast CT image into a coronary artery CTA image, and the method comprises the steps: obtaining an NCCT volume and a CTA volume which are matched, carrying out the data preprocessing, and constructing a training data set; respectively constructing an NCCT encoder-decoder network and a CTA encoder-decoder network, and carrying out pre-training based on the training data set; a mark reconstruction network is constructed and trained, a masked low-resolution CTA mark is reconstructed under the condition of an NCCT token obtained by a trained encoder, and then a low-resolution CTA volume is reconstructed by a trained decoder; a super-resolution network is constructed and trained, and a high-resolution CTA image is generated from the low-resolution CTA image; a high resolution CTA volume is generated from the low resolution NCCT volume using the trained multi-stage model. The method has the beneficial effects that the problem of posterior crash is avoided, the decoding speed during reasoning is higher, and the training process is more stable; in the low-resolution stage, the model can learn multi-scale context information, and in the high-resolution stage, the model focuses on recovering details, more fine structures are reserved in final output, and clinical requirements can be met.

The embodiment of the invention provides a driver fatigue monitoring method, which relates to a driver fatigue technology, and comprises the following steps: aiming at each frame of an upper body image of a driver, judging an eye state and a mouth state of the frame of image; extracting high-level features of a plurality of RO I areas of each frame of image, and fusing the high-level features to obtain combined features; performing feature feedback prediction to obtain an initial fatigue state; determining an initial eye state and an initial mouth state; and carrying out weighted integration on the states to obtain a final fatigue estimation result. According to the technical scheme, firstly, a plurality of RO I regions are introduced to obtain combined features with higher characterization capability, in addition, an existing sequential network structure is improved, and a more effective fatigue probability prediction value is obtained through feature feedback; and the fatigue prediction value after weighted synthesis is obtained by combining the fatigue prediction value of the key point detection and the fatigue prediction value of the sequential network, so that the fatigue prediction result is more accurate, and the problems of false report and missing report in the current fatigue detection technology can be effectively solved.

The invention discloses a non-stationary industrial process fault monitoring method based on a self-adjusting non-stationary converter network. The method comprises the following steps: firstly, dividing a variable into a stationary variable and a non-stationary variable by adopting enhanced Dickey-Fuller test; and a sequence stabilization module is designed for non-stationary variables, and the non-stationary variables are converted into stationary variables, so that the reconstruction capability of the model is improved. Meanwhile, a stationary attention removing module is provided to compensate non-stationary information lost by stationary operation, and it is ensured that the model can retain and make full use of inherent non-stationary information when processing non-stationary data. Secondly, a two-stage training strategy based on weight self-adjustment is provided, in the first stage, input window data are reconstructed, the reconstruction deviation of the first stage is used as priori knowledge of attention weight updating of an attention module in the second stage, the model is made to pay more attention to a time sequence with the large deviation, and therefore the convergence speed of the model is increased; and the accuracy and reliability of fault monitoring are improved.

The invention provides a measurement-based quantum calculation method for a physical driving variable component sub-character solicitation solver, and the method comprises the steps: S1, constructing a variational structure according to a Hamiltonian expression of a to-be-solved quantum multi-body system; s2, converting the variational structure into a graph state and a measurement mode corresponding to a measurement type quantum calculation scheme; s3, initializing an input state; s4, according to the designed measurement sequence, performing a series of single particle measurement on the graph state from the input state to obtain the output state of the variable component sub-line; and S5, continuously iteratively optimizing the variational parameters until the ground state energy of the target Hamiltonian is solved. According to the method, a variational structure is constructed from Hamiltonian of a problem to be solved, the advantages of single-particle measurement operation in measurement type quantum calculation are utilized, implementation of a multi-quantum-bit revolving door is simplified by introducing a small number of auxiliary quantum bits, and therefore quantum resource overhead of a quantum circuit is greatly reduced.

The invention discloses an automatic driving vision multitask sensing method and system based on aerial view features, and belongs to the technical field of automatic driving. Comprising the following steps: acquiring images from a look-around camera of the automatic driving system and extracting features; a multi-task distillation module is adopted to cooperatively predict dense depth features, semantic features and content features, and a dense depth prediction result and a dense semantic prediction result are generated; splicing the content feature and the dense semantic prediction result, and generating a semantic enhanced aerial view feature by using the dense depth prediction result as a depth weighted value of the content feature; optimizing aerial view features; predicting the distribution probability of occupied grid categories, fusing the probability distribution of foreground categories and optimized aerial view features; and generating a target detection prediction result according to the fused features. According to the method, the relevance between the target detection task and the occupied grid prediction task is fully mined, the model performance is improved, deployment redundancy caused by multiple independent tasks is reduced, and therefore the training and operation and maintenance cost is reduced.

The invention provides an LSTM network model for enhancing short-term memory, and belongs to the technical field of network models.The LSTM network model comprises a data input module, a data calculation unit, a feature extraction module and an output module, the data input module is connected with the data input end of the data calculation unit, and the data input module is used for inputting external processing data; the output end of the data calculation unit is connected with the feature extraction module and the output module, the output end of the feature extraction module is connected with the input end of the data calculation unit, the data calculation unit is used for memorizing and forgetting information, and the feature extraction module extracts feature values from vectors output by the data calculation unit and outputs the feature values to the output module. And then the data is circulated back to the data calculation unit, so that the processed data can be memorized again, and the data processing efficiency is improved.

The invention provides a multispectral image classification method based on an adaptive feature fusion residual network, and belongs to the technical field of computer vision and remote sensing image processing. The method comprises the following steps: firstly, preprocessing an original multispectral image; then, a novel adaptive feature fusion residual network model is constructed, the core of the model comprises a dual-path feature extraction backbone network, and spectral features and spatial context features are extracted respectively; a self-adaptive feature fusion module is introduced, multi-level features from double paths are dynamically learned and fused through a channel attention mechanism, and optimal integration of spectral information and spatial information is achieved; and finally, completing pixel-level or image-level classification by using the fused advanced features. According to the method, the problems of insufficient utilization of spectrum-space joint features of multispectral data and weak model generalization ability of a traditional method are effectively solved, and the precision and robustness of multispectral image classification are remarkably improved.

The invention discloses a lithium battery state estimation method and system based on a PNGV model and a DEKF algorithm, and the method comprises the steps: firstly, carrying out the systematic offline test of a battery, and obtaining an open-circuit voltage-state of charge curve of the battery and the equivalent capacitance of a second-order PNGV equivalent circuit model; secondly, discretizing a continuous state equation of the model into a difference equation through bilinear transformation, and identifying other model parameters of the battery in different charge states by applying a recursive least square method; and finally, constructing a state filter and a parameter filter by adopting a double-extended Kalman filtering algorithm, and realizing collaborative estimation of the charge state and the health state through data interaction and feedback of the two filters. According to the method, through a progressive process of parameter identification and cooperative state estimation, the problems of time-varying characteristics and state coupling of battery model parameters are effectively solved, and the joint estimation precision and robustness of the state of charge and the state of health of the lithium battery under complex working conditions are remarkably improved.

The application provides a Raman spectrum urine iodine concentration detection method based on machine learning, and relates to the technical field of Raman spectrum urine iodine concentration detection.In the application, first, standardized spectrum data is acquired;second, multi-scale wavelet decomposition is performed to extract energy features, mean features and standard deviation features of different decomposition levels, and multi-scale statistical features are formed;based on the characteristic vibration wave band of urine iodine molecules, physical prior features of the corresponding wave number interval are extracted;then, the multi-scale statistical features and the physical prior features are fused to form a fusion feature vector;finally, the fusion feature vector is taken as the input of a pre-trained machine learning model to acquire a urine iodine concentration prediction value.The method extracts multi-scale statistical features by performing multi-scale wavelet decomposition on the Raman spectrum, combines the physical prior features of the characteristic vibration wave band of urine iodine, and constructs a fusion feature vector with stronger representation ability, so that the accuracy and stability of urine iodine concentration detection are significantly improved by using machine learning technology.

This invention discloses a method and system for terrain accessibility assessment of unmanned ground vehicles in unstructured environments, belonging to the field of terrain accessibility analysis technology. The method includes: acquiring image data and LiDAR point cloud data using data acquisition equipment, and extracting features from the image data and LiDAR point cloud data using a pre-trained model to obtain image feature maps and point cloud feature maps; the data acquisition equipment includes a camera and a LiDAR; fusing the image feature maps and point cloud feature maps to obtain multimodal feature maps; spatially transforming the multimodal feature maps to obtain transformed multimodal feature maps; classifying and calculating the transformed multimodal feature maps to obtain an elevation map and a accessibility map, using the accessibility map as the final output of the terrain accessibility assessment. This invention is applicable to unstructured scenarios with complex terrain types and frequent dynamic environmental changes.

The invention discloses an audio generation model training method and device and electronic equipment, and belongs to the technical field of artificial intelligence. The method comprises the steps of encoding a sample audio through a neural network model to obtain at least two first sample features, wherein the ith first sample feature represents feature information of each audio segment obtained after the sample audio is segmented according to the ith time scale; generating a reconstructed audio based on the at least two first sample features through a neural network model; and training a neural network model through the sample audio and the reconstructed audio to obtain an audio generation model, the audio generation model being used for generating a second audio based on the first audio. The different first sample features reflect the information of the sample audio on different time scales, so that the representation capability of each first sample feature is stronger, based on this, the reconstructed audio with higher accuracy can be generated according to each first sample feature, and the audio generation model with higher accuracy can be obtained through training based on the reconstructed audio.

The present application discloses an information prediction method based on artificial intelligence, related equipment and storage medium, which are suitable for information prediction services implemented based on deep learning networks. After obtaining business data from an information prediction request, the method uses a target neural network model to extract features from the business data, so as to perform information prediction based on the extracted business data features and obtain information prediction results. The target neural network model is obtained by performing parameter value optimization processing and compression processing after parameter optimization processing on a reference neural network model. The parameter value optimization processing at least includes the optimization of weight parameters, bias parameters and mask parameters associated with each hidden layer neuron, and the mask parameters are used to indicate the importance of the hidden layer neurons. The compression processing refers to deleting the hidden layer neurons corresponding to the mask parameters whose parameter values ​​in the hidden layer of the reference neural network model are lower than the threshold, thereby improving information prediction efficiency and reducing business deployment costs.

The application discloses an electric power transaction data mining system based on artificial intelligence, which comprises a time sequence construction module, an association mining module, a hidden space coding module and an abnormal positioning module.The time sequence construction module acquires electric power transaction historical data and constructs a transaction state time sequence.The association mining module inputs the transaction state time sequence into an adaptive graph convolution network based on an attention mechanism, extracts dynamic association features between electric power transaction participants, and generates a preliminary transaction behavior feature graph.The hidden space coding module inputs the preliminary transaction behavior feature graph into a variational autoencoder based on adversarial training, maps the preliminary transaction behavior feature graph to a hidden space through an encoder, and obtains hidden variables by sampling.The abnormal positioning module inputs the hidden variables into a decoder to reconstruct a reconstruction error matrix, identifies electric power transaction time periods and participants with abnormal transaction modes according to the reconstruction error matrix, and takes the electric power transaction time periods and participants as initial anchor points for data mining.The system belongs to the technical field of electric power transaction data mining and can effectively mine dynamic association features and locate abnormal transactions.

The invention discloses a three-stage joint envelope embedded stack auto-encoder, which comprises a three-stage joint envelope module, an embedded stack auto-encoder module and a three-layer space integration mechanism module, and is characterized in that the three-stage joint envelope module is internally provided with a manifold sample pair envelope module, a dimensionality reduction keeping type clustering module and a three-stage consistency keeping module; the manifold sample pair enveloping module is used for acquiring inherent structures of adjacent samples with similar manifolds in a sample enveloping space; the maintenance dimension reduction type clustering module is used for projecting samples to a low-dimensional space; the inter-three-stage consistency maintaining module is used for maintaining the consistency of sample distribution; after an input sample enters the three-stage combined envelope module, the input sample is processed and a three-layer sample space is generated; the built-in stack automatic encoder module carries out built-in stack automatic encoder modeling in parallel on the three layers of sample spaces, and finally the three layers of space integration mechanism modules carry out fusion and obtain output samples. The method has the advantages that the clustering state space is expanded, and the sample characterization capacity is improved.

The present application discloses an image classification method based on multimodal information fusion, which relates to the field of image processing technology. The image classification method uses a multi-scale feature extraction module in a feature extraction network to perform multi-scale feature extraction on the original image of each modality. The multi-feature fusion pyramid is improved on the basis of the FPN framework combined with a hole convolution layer, which can effectively extract the context information of the original image. The multi-feature association module further enhances the multi-scale feature association, so that the feature extraction network can obtain a single-modal feature map with a stronger representation ability of the original image. After multi-modal information fusion is performed on two single-modal feature maps to obtain a fused feature map, in addition to using the fused feature map, the original single-modal feature map is retained and classified together using a classification network. The method can fully mine the image information within and between modalities, thereby effectively improving the accuracy and robustness of image classification.

The invention discloses a polymer formula analysis method and device based on a graph Q learning network, and the method comprises the steps: building a Markov decision process of formula generation through defining a state space, an action space, a state transition probability and a reward function; constructing a graph Q learning network model based on the Markov decision process; training the graph Q learning network model through a double-Q learning algorithm and a strategy function to obtain a trained graph Q learning network model; and performing state transition and action selection processing through the trained graph Q learning network model to generate a polymer formula. According to the method, the problems of low learning speed, high computing resource consumption, difficulty in accurately matching multi-target performance and the like in the prior art are solved.

The application relates to the technical field of unmanned aerial vehicle scheduling and intelligent optimization, and discloses a multi-unmanned aerial vehicle cooperative task allocation method and device for complex task scenarios, which has the technical scheme as follows: global multi-source data acquisition and space-time heterogeneous hypergraph representation, complex task causal emergence decoupling and counterfactual deduction optimization, hierarchical federated heterogeneous matching degree accurate representation, distributed external generalization causal element reinforcement learning dynamic allocation decision, distributed conflict resolution and digital twin closed loop iterative optimization; by adopting frontier technologies such as space-time heterogeneous hypergraph representation, causal emergence reasoning, hierarchical federated learning, causal element reinforcement learning and digital twin closed loop optimization, a causal driving full-link algorithm architecture is constructed to break through the inherent limitations of traditional correlation-based data-driven algorithms, so that efficient cooperative operation of a large-scale heterogeneous unmanned aerial vehicle cluster in a strong coupling task and an extreme dynamic environment can be realized.

The present invention relates to a method for detecting anomalies in time series data based on a multi-head attention model, and belongs to the technical field of anomaly detection in computer databases. The present invention builds a network framework for detecting anomalies in time series data based on a multi-head attention model, which can process multi-source time series data, has robustness and good generalization ability, and can achieve faster network training and more accurate abnormal sample detection. A frequency domain analysis module is introduced to make full use of the frequency domain information of time series data. By performing frequency domain analysis on the input sequence and the output sequence, the model can obtain effective sequence information. By measuring the distance between the two frequency domain analysis results, the network can be supervised to generate a more accurate output sequence, thereby obtaining a more accurate abnormal judgment threshold and achieving more accurate anomaly detection. The present invention has obvious advantages in the feature extraction capability of time series data and the detection accuracy of abnormal samples.

The embodiment of the invention discloses a radar radiation source individual identification method and device for airport low-altitude security and protection. A specific embodiment of the method comprises the following steps: generating a radar signal fragment group set; performing self-supervised pre-training on the radiation source feature extraction network to update model parameters; constructing a radar radiation source individual identification model; performing joint optimization on the radar radiation source individual identification model through the known radiation source signal group set to generate a known radiation source center feature set; generating a known radiation source boundary parameter set according to the radar radiation source individual identification model and the known radiation source center feature set; and according to the known radiation source boundary parameter set and the radar radiation source individual identification model, performing radiation source individual identification on the real-time radar signal to generate a radar radiation source identification result. According to the embodiment, the generalization ability of the recognition model can be improved under the small sample condition, so that the individual recognition accuracy of the radiation source is improved, and the low-altitude security and protection safety of an airport is improved.

The present application discloses a training method, device and electronic device for an audio generation model, and belongs to the field of artificial intelligence technology. The method includes: encoding sample audio through a neural network model to obtain at least two first sample features, the i-th first sample feature represents the feature information of each audio segment obtained after segmenting the sample audio according to the i-th time scale; generating reconstructed audio based on at least two first sample features through a neural network model; training the neural network model through the sample audio and the reconstructed audio to obtain an audio generation model, and the audio generation model is used to generate a second audio based on the first audio. Since different first sample features reflect the information of the sample audio at different time scales, the representation ability of each first sample feature is stronger. Based on this, a reconstructed audio with higher accuracy can be generated according to each first sample feature, so that an audio generation model with higher accuracy can be trained based on the reconstructed audio.

The invention provides a flight control sensor electric signal abnormity detection method and system based on electric signal analysis, and the method comprises the steps: collecting the voltage and current signal data of an aircraft control sensor in real time, and carrying out the data preprocessing of the voltage and current signal data; inputting the preprocessed electric signal data into a pre-constructed nonlinear four-rotor kinetic model for analogue simulation, and outputting an electric signal data set containing fault information; based on the real-time feature vector in the electric signal data set and a pre-stored historical feature sample set, constructing a global adjacency graph containing real-time nodes and historical nodes; inputting the global adjacency graph into a pre-constructed graph convolutional network model, and extracting space-time fusion features of the real-time nodes; and carrying out classification processing on the space-time fusion features, and outputting an abnormal state category of a sensor electric signal, so as to achieve the purposes of online high-precision diagnosis and autonomous fault-tolerant control of aircraft sensor faults and guarantee safe and reliable operation of a complex equipment system.

The invention provides a sketch graph generation method, device and equipment and a readable storage medium, relates to the technical field of image processing, and aims to solve the problems of relatively large data volume and relatively low image coding efficiency of a semantic representation method in the prior art. The method comprises the following steps: carrying out edge detection on a to-be-processed image to obtain an edge image of the to-be-processed image; determining initial sketch points in the edge graph, wherein the initial sketch points comprise sketch graph points belonging to edge points in the edge graph; based on the image features of the edge image and the position information of the initial sketch points, the connection probability of an initial sketch point pair is determined, and the initial sketch point pair comprises any two initial sketch points in the initial sketch points in the edge image; and generating a target sketch based on the connection probability of the initial sketch point pairs and the position information of the initial sketch points. According to the invention, the image coding efficiency can be improved.

The invention relates to a target detection method and device, equipment, a readable storage medium and a program product. The method comprises the following steps: preprocessing an input image to obtain a preprocessed image; performing feature extraction on the preprocessed image to obtain a first group of feature pyramids; performing feature fusion on the first group of feature pyramids through different transmission channels to obtain a fused feature image; performing feature processing on the fused feature image through two branches to obtain a prediction target category and a prediction target coordinate; and determining a matching relationship between the prediction target category and the prediction target coordinate to obtain a target detection result. Therefore, redundant reasoning pixel content can be reduced, the reasoning speed in a video monitoring scene is improved, end-to-end target detection is directly completed, the detection speed is high, and the accuracy is high.

The invention discloses a lake water quality parameter inversion method based on a remote sensing image, a medium, equipment and a product, and relates to the field of water quality parameter inversion, and the method comprises the steps: obtaining a Sentinel-2 remote sensing image and water quality sampling data, re-sampling a wave band in the remote sensing image to a uniform resolution, and carrying out the time-space matching of the wave band and the water quality sampling data, selecting a dual-band combination, a three-band combination and a four-band combination related to the water quality sampling data from the matched resampling bands, and constructing a training set; a deep learning inversion model is constructed, the deep learning inversion model comprises three feature extraction sub-networks, an attention mechanism and a fusion main network, wave band combinations are input into the three feature extraction sub-networks respectively, obtained three feature vectors are subjected to weighted fusion through the attention mechanism and then input into the fusion main network, and predicted water quality parameter concentration is output; and training the model by using the training set, and performing lake water quality parameter inversion by using the trained model. According to the invention, the inversion precision of water quality parameters is improved.

The application provides a multi-feature fusion oral bioavailability prediction method, device, equipment and medium. Four-dimensional features of a candidate drug, including a graph structure feature, a fingerprint feature, a physicochemical descriptor and a pharmacophore feature, are fused to comprehensively depict drug molecular characteristics, so that the representation capability is stronger, and the prediction accuracy is improved. In addition, the fused features are processed by a plurality of different expert processing models corresponding to a plurality of preset different result types, so that different processing strategies can be adopted for different chemical structure types, and different application scenarios can be adapted.

The invention relates to the field of fire early warning, and provides an industrial park-oriented electrical fire comprehensive monitoring and early warning method and a corresponding product. The method comprises the following steps: synchronously acquiring multi-source data of the industrial park power supply network; processing the collected multi-source data of the industrial park power supply network so as to respectively extract electrical transient characteristics and thermal hidden danger space characteristics, fusing the characteristics with the multi-gas concentration data based on an attention mechanism, and outputting a high-dimensional hidden danger characteristic vector; carrying out parallel processing on the high-dimensional hidden danger feature vectors to respectively obtain a physical inconsistency score and a data-driven hidden danger probability score; dynamically adjusting the weight of the physical inconsistency score and the weight of the data-driven hidden danger probability score according to the signal-to-noise ratio and the feature uncertainty of the real-time data through an adaptive weighted fusion mechanism, and performing fusion to generate a dynamic electrical fire risk index; and according to the threshold interval in which the dynamic electrical fire risk index is located, generating early warning information of a corresponding level.

The application relates to an aviation laser point cloud semantic segmentation method and device based on a multi-level context feature fusion network, which comprises the following steps: S1, data down-sampling preprocessing is performed on acquired target region point cloud; S2, the processed aviation laser point cloud data is input into a pre-established multi-level context feature fusion network coding layer, and each layer feature is input into a cross-layer attention module to obtain cross-layer connection features; S3, the coding features are up-sampled to the size of the point cloud to be segmented through a decoding layer, multi-level feature fusion modules are used for fusing the features of all levels, and then a Softmax function is used to obtain the final segmentation result of the aviation laser point cloud; and S4, a hybrid loss function is constructed to obtain the error between the segmentation result and the true value, and according to the error value, a random gradient descent method is used for back propagation training, model parameters are dynamically adjusted, and a stable segmentation result is obtained. The method and device can effectively enhance the association of point feature expression and spatial information, improve the information perception ability of multi-scale context features of the point cloud, and obtain a more fine semantic segmentation result.

The invention discloses a business demand prediction method and device, equipment, a storage medium and a program product, which are used for rapidly and accurately predicting business demands and enabling a demand prediction process to have interpretability. The business demand prediction method comprises the following steps: constructing a first feature relation graph of a business object based on an object feature set of the business object; performing graph attention operation on the first feature relation graph to obtain an attention coefficient between nodes and node features of the nodes in the first feature relation graph; based on the attention coefficient between the nodes, reconstructing the first feature relation graph to obtain a second feature relation graph of the business object; and predicting business demand information of the business object based on the node features of the nodes and the second feature relation graph.

The invention discloses an oil field station environment personnel intrusion face recognition method and system and electronic equipment, and the method comprises the steps: selecting feature maps outputted by a plurality of branches on a multi-stage trunk feature extraction network from face image features, merging the deep features of the feature maps to the shallow layer step by step, and obtaining multi-scale features; shielding information in the features is decoded into feature masks through a mask decoder; weighting the feature masks and features in the original feature map, and endowing useful feature information with a larger weight; connecting the obtained features, and adjusting the number of feature channels through convolution to enable the connected features to be consistent with the original image; connecting the feature masks, and enabling the channel number of the feature masks to be consistent with that of the original image through convolution adjustment; the classification loss of face recognition and the predicted shielding mode loss are weighted and then added to obtain comprehensive loss; and judging whether the person is intruded or not according to the face image of the person by using the shielded face recognition network model obtained by training. According to the invention, the recognition accuracy of the shielded face can be improved.