568 results about "Neutral network" patented technology

A system for generating contrast-enhanced magnetic resonance images of a subject includes an input configured to receive at least one non-contrast enhanced image of the subject, and a contrast-enhanced magnetic resonance (MR) image synthesis neural network coupled to the input and configured to generate a contrast-enhanced magnetic resonance image of the subject based on the at least one non-contrast enhanced image of the subject. The contrast-enhanced MR image synthesis neural network is trained using a set of training data comprising at least quantitative data.

This invention belongs to the field of fuel cell health status estimation technology, and specifically provides a method for estimating the health status of proton exchange membrane fuel cells based on EIS prediction. The method includes the following steps: S1: Collecting proton exchange membrane fuel cell operating data and impedance spectrum data; S2: Preprocessing the operating data; S3: Establishing an EIS prediction model based on a time-series neural network; S4: Verifying the effectiveness of the impedance spectrum; S5: Decomposing the effective impedance spectrum using the DRT method; S6: Extracting polarization resistance characteristics corresponding to different polarization processes based on the relaxation time distribution function; S7: Establishing a mapping model between polarization resistance characteristics and health status; S8: Online health status estimation. This invention solves the problems of high difficulty in obtaining electrochemical impedance spectroscopy data and high testing costs in practical applications, improving the engineering application feasibility of this method in online or quasi-online health status monitoring of proton exchange membrane fuel cells.

The application discloses an electromagnetic interference identification method and device, equipment and a readable storage medium, and relates to the technical field of signal processing. The in-phase and quadrature data of a to-be-identified radar signal in multiple distance banks is acquired, and the maximum power variation coefficient, the skewness coefficient, the kurtosis coefficient and the maximum sharpness of the to-be-identified radar signal are calculated according to the in-phase and quadrature data. The above characteristic parameters are input into a pre-trained electromagnetic interference identification model, the electromagnetic interference identification model adopts a full connection neural network structure, the identification results of the to-be-identified radar signal in the multiple distance banks are obtained through an output layer, and the identification results are used for distinguishing meteorological echo useful signals from electromagnetic interference signals. The pollution area of the to-be-identified radar signal is determined according to the identification results of the to-be-identified radar signal in the multiple distance banks, the pollution area is recorded, and an alarm prompt is sent to technical personnel, so that the missed judgment of electromagnetic interference and the misjudgment of useful signals can be avoided, and the identification accuracy of various electromagnetic interferences is improved.

The application relates to the technical field of optical fiber communication monitoring, and discloses a power communication line monitoring method and system based on optical fiber multi-source data fusion. The method comprises the following steps: calculating the weight of each data source through a confidence evaluation neural network according to the historical stability and the error code rate variation of each data source; calculating a communication service quality correction factor and correcting to obtain a line state fusion characteristic value according to the error code rate and the optical signal-to-noise ratio after weighting and fusing each weight and normalized data characteristics; when the weight ratio of optical signal data weight and communication performance data weight exceeds a set threshold, extracting the optical fiber disturbance parameters and the optical power sequence of adjacent monitoring positions to calculate a compensation disturbance value through optical power weighted least square fitting; and calculating a theoretical optical power variation according to the compensation disturbance value and the optical fiber loss coefficient and taking the deviation between the theoretical optical power variation and the measured value as a compensation effectiveness criterion. The application improves the accuracy and reliability of optical fiber communication line monitoring.

The application discloses a blood sugar management applet with home support function. A data acquisition interaction module is connected with an external blood glucose meter device through a Bluetooth communication protocol to collect blood sugar data of a user in real time; a home collaborative management module realizes safe sharing of blood sugar data of the user among family members, and the family members can check relevant blood sugar data, diet and exercise records through the applet end; and an intelligent analysis and reminding module is based on an LSTM neural network model, performs trend analysis on historical blood sugar data after a risk early warning trigger, generates personalized health reminding information automatically in combination with preset rules, and sends the information to the user and associated family members in a multi-modal form. The application effectively improves the blood sugar management efficiency of patients, reduces the risk of data leakage, and can significantly reduce the fasting blood sugar value and the 2-hour postprandial blood sugar value of patients compared with the prior art, reduces the cost of manual intervention, and has significant clinical application value and social and economic benefits.

The application discloses an artificial intelligence-based high-voltage cable temperature monitoring system, relates to the technical field of high-voltage cable monitoring, and aims at solving the problems of low prediction accuracy, lack of multi-factor interaction analysis, incomplete fault evaluation and untimely early warning in the existing high-voltage cable temperature monitoring. The data acquisition module is used for collecting multi-source monitoring data of the high-voltage cable in real time, storing and sending; the data preprocessing and fusion module is used for receiving the multi-source monitoring data, preprocessing the multi-source monitoring data, and screening out an effective feature data set through correlation analysis; the intelligent temperature prediction module constructs a neural network prediction model based on the effective feature data set, realizes accurate prediction of the core temperature, surface temperature and joint temperature of the high-voltage cable, and outputs a temperature prediction result. The application has the advantages of high temperature prediction accuracy, strong real-time performance, multi-dimensional data fusion and interaction analysis, comprehensive and quantitative accurate fault risk evaluation, timely early warning, strong operation pertinence, strong system compatibility and high scalability.

Methods for quantizing, training and using a depthwise separable convolutional neural network. The network layers include one or more pointwise convolution layers, which are suitable for performing pointwise convolutions and in each case comprise a plurality of pointwise convolution layer weights, and one or more depthwise convolution layers, which are suitable for performing depthwise convolutions and in each case comprise a plurality of pointwise convolution layer weights. The quantization method includes quantizing the plurality of pointwise convolution layer weights to a plurality of quantized pointwise convolution layer weights in a first discrete range and quantizing the plurality of pointwise convolution layer weights to a plurality of quantized pointwise convolution layer weights in a second discrete range, wherein the first discrete range has a strictly lower cardinality than the second discrete range.

This invention relates to a method for evaluating the energy-saving performance of green buildings based on a backpropagation (BP) neural network, belonging to the field of green building technology evaluation. Using the Delphi method and comparative analysis, 14 evaluation indicators are selected from five dimensions—energy-saving design, environmental energy consumption, building lighting, building ventilation, and building sunshine—to construct a quantitative indicator system. Then, the analogy-heuristic (AHP) method is used to determine the indicator weights, and consistency checks are performed to eliminate subjective bias. Finally, a BP neural network evaluation model is established, using the quantitative data of the 14 indicators as input. Through training and optimization, the energy-saving level of the green building is determined. This invention can accurately determine the indicator weights using a multi-dimensional quantitative indicator system and the AHP method, and it can ensure high accuracy of the evaluation results through the strong data processing and self-optimization capabilities of the BP neural network.

The application belongs to the technical field of image processing, and discloses a rape root tumor level recognition method, device, equipment and storage medium; the method comprises the following steps: acquiring a to-be-recognized image of a rape root tumor and inputting a rape root tumor grading neural network model; performing feature extraction on the to-be-recognized image through a feature extraction layer of the rape root tumor grading neural network model to obtain a to-be-recognized feature vector, wherein the feature extraction layer comprises a basic residual module and a residual module containing an attention mechanism; performing feature recognition on the to-be-recognized feature vector through a feature recognition layer of the model to obtain a tumor level of the rape root tumor; through the rape root tumor grading neural network model, the attention mechanism is used to improve the capture of the neural network on the features at the details in the rape root image, and through the recognition of the features at the details, the ability of the neural network to recognize the tumor level features of the disease area of the rape root can be effectively improved, so that the accuracy of the rape root tumor grading can be effectively improved.

The present application belongs to the technical field of intelligent processing of electric shovels, and discloses an electric shovel intelligent excavation prompting system based on a hybrid encoder and a Transformer decoder, comprising: a data acquisition module, a sensor is arranged on the electric shovel to acquire real-time operating parameters of the electric shovel and real-time environmental information parameters of the working environment of the electric shovel; a hybrid encoder module, which dynamically constructs the relationship between nodes and edges through a graph neural network for the real-time operating parameters and the environmental information parameters, and converts the relationship into graph structure data; and removes noise information from the graph structure data through a cross-modal variational autoencoder to obtain encoded structure data; and a Transformer decoder module, which captures the correlation of the encoded structure data based on a self-attention mechanism, acquires working condition information of the electric shovel, and outputs excavation prompting information. The present application realizes intelligent processing of real-time operating parameters and environmental information parameters of the electric shovel, outputs excavation prompting information to assist the electric shovel driver in operation, and improves the efficiency and safety of electric shovel operation.

The present application provides a kind of pulse neural network conversion method and device for underwater sonar autonomous detection, the method comprises the following steps: step S1: modify YOLOv3-tiny model based on deep neural network model (DNN), obtain the YOLOv3-tiny model after modification;Step S2: load underwater sonar image data set, train the YOLOv3-tiny model after modification;Step S3: for underwater target detection, the trained YOLOv3-tiny model is converted into YOLOv3-SNN model based on pulse neural network (SNN), and the YOLOv3-SNN model includes input layer, hidden layer and output layer, wherein input layer uses real number coding, hidden layer uses the two-state pulse neural coding of active state and resting state, and output layer uses membrane voltage decoding;Step S4: using the YOLOv3-SNN model based on pulse neural network (SNN), is directly used for underwater sonar small target detection task.

The application provides a device and working method for AI chip system architecture verification, including a control center for logic control, timing coordination and data interaction of each module. A pulse writing module generates a writing signal with a special pulse form to realize state programming and conduct adjustment of storage or computing devices. A programming selection module accepts the signal of the pulse writing module, supports the structure requirements of various different memory devices of the measured unit, and performs different selective writing according to different array structures. A decoding module accepts the signal of the programming selection module and the feedback signal of the measured unit to be read, and is used for address space mapping, excitation line management and signal on-off control. The measured unit accepts the writing signal of the decoding module and feeds back the reading signal to the decoding module, and is used for storing neural network weights, logic states or continuous conduct grades. The application provides key test support and decision basis for AI chip system architecture research.

This invention provides an end-to-end voice encryption method, apparatus, and Bluetooth headset suitable for multi-hop lossy channels. The method includes front-end noise reduction, framing, and segmentation of the acquired voice at the transmitting end. Based on the session mode, a group public key or a point-to-point private key is automatically selected as the master key from a pre-configured key set. Then, multiple time-domain segments within each frame are scrambled, subjected to segment-by-segment Fast Fourier Transform, and frequency-domain encryption based on subkeys, according to the master key. The encrypted voice is then generated through inverse transformation and overlapping weighted smoothing concatenation, and transmitted through a multi-hop voice communication channel containing multi-level lossy encoding and decoding. At the receiving end, the reverse process is performed: framing and segmentation, transformation, inverse frequency-domain encryption, and inverse time-domain scrambling. Combined with U-shaped neural network noise reduction at both the front-end and back-end, the voice is restored to intelligible speech. This method enables flexible key management and high-quality secure voice transmission in scenarios where group calls and point-to-point private calls coexist and undergo multiple lossy compression operations.

This invention discloses an intelligent risk assessment method and system for dam-break flood propagation in water network nodes, comprising: collecting basic data of water network engineering and dam-break scenario parameters to generate a standardized dataset; constructing a directed topological structure model of water network nodes-channels that integrates engineering semantic constraints and hydraulic propagation direction constraints, and generating a water network adjacency matrix; using a neural network model based on water network structural constraints and the physical knowledge of mass and momentum conservation to predict the propagation process of dam-break floods in the water network; converting the flood propagation prediction results into flood impact indices, calculating the risk levels of engineering facilities and population units based on the flood impact indices, and outputting risk assessment results and decision support information. This invention provides decision support information for water network scheduling and emergency management.

This invention discloses a liver image recognition system and method based on graph neural networks, belonging to the field of medical image processing and intelligent recognition technology. It includes a liver image receiving module for receiving liver image data and performing slice processing, grayscale standardization, and spatial location information association processing on the liver image data to generate standardized liver image data. In this invention, by performing slice processing, grayscale standardization, and spatial location information association processing on the liver image data, a unified correspondence is formed between multiple liver image slices in three dimensions: slice number, grayscale expression, and spatial location. Results of liver parenchyma regions, candidate abnormal regions, and boundary transition regions are extracted. The adjacency relationships, grayscale variation relationships, and boundary continuity relationships of adjacent regions within the same slice are uniformly organized, thereby improving the data orderliness, regional discrimination ability, and completeness of the expression of relationships within a single slice of multi-slice liver images.

The application provides a colon cancer diagnosis method based on an adaptive molecular LVQ neural network, and comprises the following steps: determining the functions of an input layer, a competition layer and a linear output layer of the LVQ neural network; designing basic function modules of the molecular LVQ neural network according to DNA strand displacement reactions; adding a supervised learning module on the basis of the basic function modules; determining the DNA strand structures and small fulcrum structures of the auxiliary substances and reactants in the chemical reaction process; cascading the basic function modules and the supervised learning module into the adaptive molecular LVQ neural network; constructing a data set from TCGA and GTEx databases; inputting the data set into the adaptive molecular LVQ neural network, performing benign or malignant diagnosis, and triggering weight dynamic updating according to the diagnosis error, so as to complete adaptive optimization of the adaptive molecular LVQ neural network. The application realizes the supervised learning and adaptive optimization at the molecular level, and successfully classifies and identifies the benign and malignant colon cancer samples.

The present application belongs to the intelligent traffic technical field, specifically is a kind of trajectory restoration system based on urban traffic road network.The system of the present application includes traffic trajectory data acquisition module, GPS trajectory feature extraction module, GPS trajectory restoration learning module;Wherein traffic trajectory data acquisition module collects the latitude and longitude data of the vehicle driving on urban road in driving process, GPS trajectory feature extraction module extracts space-time feature to the low sampling rate trajectory point input neural network, and GPS trajectory restoration learning module inputs the output of last stage into new neural network model to predict the latitude and longitude information of missing sampling point in input low sampling rate trajectory.The present application can greatly improve the utilization of low sampling rate trajectory data;At the same time, combined with the topological structure of urban road network, the restored GPS trajectory is more in line with the real trajectory, effectively improves the utilization efficiency of trajectory data.

The application provides a soil erosion risk early warning method and device based on dust data, which comprises the following steps: performing grid processing and screening on a region to be analyzed to obtain an effective grid region; obtaining NDVI data and soil moisture data of the effective grid region; determining a slope length factor and a slope factor of each effective grid region; determining net dust data in the region to be analyzed by eliminating the influence of wind direction; determining a dust emission value of each effective grid region based on the NDVI data, the soil moisture data and the net dust data; calculating the NDVI data, the soil moisture data, the slope length factor, the slope factor and the dust emission value by using a neural network algorithm to determine a soil erosion intensity prediction value; combining the soil erosion intensity prediction value to determine a soil erosion value of each effective grid region; determining a soil erosion risk grid based on the current soil erosion value and historical soil erosion data, and performing risk early warning on the soil erosion risk grid.

The present application relates to the field of livestock and poultry intelligent breeding and environmental control, in particular to a sheep house environment adaptive regulation system for Nanjiang yellow sheep breeding, comprising a data acquisition device, an environment control device, a feeding execution device and a controller; the controller receives real-time environmental data of a target environment and real-time physiological behavior data of a target breeding object, generates a physiological stress load value through a time integral algorithm, and inputs the physiological stress load value and the real-time physiological behavior data into a neural network model containing a time sequence processing structure to output an effective feeding capacity, a nutrient absorption efficiency and an expected feed conversion rate; when the effective feeding capacity is lower than a standard feeding capacity, a target intervention combination strategy composed of an environment compensation weight and a feeding adjustment weight is generated with the condition of maximum expected feed conversion rate, and then a feeding end control instruction and an environment control end control instruction are generated and closed-loop feedback adjusted to realize self-adaptive closed-loop control of the sheep house environment and the Nanjiang yellow sheep.

This invention discloses a digitally-based intelligent wastewater discharge method, comprising the following steps: acquiring a historical dataset, which includes historical wastewater data and historical optimal opening / closing size data of wastewater valves; training a feedforward neural network using the historical wastewater data as input and the historical optimal opening / closing size data of wastewater valves as output to obtain a trained wastewater monitoring model; inputting real-time monitoring data into the wastewater monitoring model and outputting optimal opening / closing size data of wastewater valves; and dynamically adjusting the opening / closing size of the wastewater valves according to the optimal opening / closing size data and adjustment commands. This invention dynamically adjusts the opening / closing size of wastewater valves using a digital method, ensuring wastewater is discharged in optimal condition, avoiding the resource waste and poor treatment effect caused by traditional fixed wastewater valve installations, thereby improving the overall efficiency of wastewater treatment.

The application discloses a device for judging the decompression effect of intraoperative trigeminal neuralgia, and relates to the technical field of medical monitoring devices, comprising a forceps body, a collection probe is arranged at the tip of the forceps body, a wiring box is arranged on the side of the forceps body away from the collection probe, a distance adjusting mechanism and an auxiliary center positioning mechanism are arranged between the wiring box and the forceps body, the forceps body with adjustable spacing and the distance adjusting mechanism are arranged, the flexible adjustment of the spacing of the collection probe is achieved, the overall signal of the ganglion segment can be detected by increasing the spacing, the smaller nerve area can be focused by reducing the spacing, the signal difference under different spacings is compared, the compression point is accurately positioned, blind decompression is avoided, the auxiliary center positioning mechanism is arranged, the center point of the forceps body and the corresponding position of the collection probe are quickly positioned, the medical staff accurately aims at the trigeminal nerve to be detected, the positioning deviation is avoided, the collection signal is not inaccurate, the trigeminal neuralgia clustering neural network algorithm is combined, a visual distribution view is generated, and the accuracy of decompression effect judgment and the operation efficiency are improved.

The present disclosure provides a pathology sample virtual staining apparatus and method that can easily acquire virtual staining sample images for unlabeled samples as well as decolorized samples, by including a sample image acquisition module that irradiates near-infrared light toward a placed sample to acquire a near-infrared-based phase image of the sample, and a virtual staining module that receives the near-infrared-based phase image and performs neural network computation to acquire a virtual staining image similar to an image of a stained sample in which the sample is stained.

The present application relates to the technical field of medical apparatus and instruments, in particular to a ventilation trigger recognition method and system of an anesthesia machine and the anesthesia machine. The method comprises the following steps: step 1: collecting airway pressure waveform data and flow waveform data of the anesthesia machine; step 2: converting the collected airway pressure waveform data and flow waveform data into matrix data; step 3: inputting the matrix data into a trained convolutional neural network, and the convolutional neural network outputs a probability value based on the input matrix data; step 4: comparing the probability value with a threshold value, and determining whether the ventilation trigger condition is met; if the trigger condition is met, the anesthesia machine starts ventilation support. The present application can more accurately determine the trigger condition by analyzing the shape of the ventilation curve through the convolutional neural network, so as to realize more accurate and safe anesthesia ventilation control.

This application provides a method for screening inhibitors targeting CYP7A1 and their use in the preparation of anti-liver cancer drugs, relating to the field of tumor targeted drug design technology. The method for screening inhibitors targeting CYP7A1 protein includes: using the crystal structure of human cholesterol 7α-hydroxylase as a receptor model, precise docking screening is performed using software to screen drug-like molecules with molecular weights of 250-500 Da and lipid-water partition coefficients of -1 to 5 from a compound database; a graph neural network active learning model is used to predict and rank the protein-small molecule binding affinity of compounds in the database; the ΔG between candidate molecules and CYP7A1 is calculated using molecular mechanics / generalized Born surface area methods, with molecules having ΔG ≤ -40 kcal / mol considered potential inhibitors. The inhibitors of this application do not rely on the enzymatic activity of CYP7A1 to achieve anti-tumor effects and do not significantly affect cholesterol and bile acid metabolism levels.

Embodiments of the present disclosure provide a solution for video processing. A method for video processing is proposed. In the method, a conversion between a current video unit of a video and a bitstream of the video is performed. Performing the conversion comprises: applying a down-sampling filter to the current video unit to obtain an internal video unit; applying at least one neural network (NN) -based filter to the internal video unit to obtain a filtered video unit; and applying an up-sampling filter to the filtered video unit to obtain a reconstructed video unit of the current video unit.

This invention relates to the field of image processing technology, and provides an image deblurring method based on direction-aware Transformer. The method involves inputting the original image into a polar coordinate embedding module within an encoder-decoder neural network; extracting shallow features from the original image through the polar coordinate embedding module; extracting the relative positions of pixels in the original image in polar coordinates through a polar coordinate attention module within the encoder-decoder neural network; and obtaining polar coordinate attention module features by performing attention calculations on the relative positional relationships between pixels; finally, inputting these polar coordinate attention module features into an image reconstruction module to generate a high-quality, clear reconstructed image. This invention extracts shallow features from the image through the polar coordinate embedding module, aggregating features in each direction. The polar coordinate attention module, during attention calculations, can better utilize the extracted shallow features, learning features in each direction to achieve better image restoration.