161results about How to "Suppress noise interference" patented technology

The invention discloses a power equipment fault prediction method based on support vector regression, and aims to improve the prediction accuracy. The method comprises the following steps: firstly, collecting multi-dimensional operation parameter data of power equipment, and constructing an original monitoring data set; a data set is segmented through a sliding time window, wavelet transform is used for noise reduction, and then a normalization method is applied to generate a standardized data set. Time domain and frequency domain features are further extracted, a feature vector space is constructed, an optimal feature subset is screened through a mutual information method, and a training data set is generated; performing preliminary training on a support vector regression model by using the training data set to generate candidate support vectors, and screening a support vector set with the maximum contribution through an optimization algorithm; a grid search method is adopted to adjust kernel function parameters and penalty factors, and a support vector set is further optimized. And evaluating the performance of the model through cross validation, applying the optimized support vector regression model to power equipment fault prediction, and generating a fault occurrence probability prediction value.

The invention relates to a remote sensing image change detection method of a difference enhancement network based on edge guidance, and the constructed difference enhancement change detection network mainly comprises a backbone network, an edge enhancement module, a double-branch difference enhancement module, a multi-scale difference feature fusion module and a decoder. Edge information of the changed object is refined through an edge enhancement module, so that feature information with clear boundaries can be extracted for the changed object; a dual-branch difference enhancement module effectively learns the difference between the foreground and the background. The module captures global and local information of a dual-time image through space and channel branches, emphasizes a change area, inhibits noise interference, and enhances long-range dependence of difference. According to the method, the detection performance of the small object under the high-resolution remote sensing image is improved by using a multi-scale difference feature fusion method. Experimental results show that the detection effects of the proposed network on a plurality of public data sets are superior to those of several common remote sensing image change detection methods at present.

The invention discloses a cancer image contrast enhancement method and system based on deep learning, belongs to the technical field of medical image processing, and adopts an adaptive histogram equalization algorithm for preprocessing to enhance the contrast of a local region. Texture, shape and edge features are extracted through multi-scale analysis, and feature descriptors are constructed. And based on the feature descriptor, segmenting the image by using a region growing segmentation algorithm to obtain a focus region boundary. A self-adaptive contrast enhancement technology is adopted for the segmented focus area, and the focus display effect is highlighted. In consideration of noise interference, a denoising method based on wavelet transform is introduced, noise is suppressed in a high-frequency sub-band, and details are reserved in a low-frequency sub-band. Through the organic combination of the above technical schemes, the contrast and definition of the cancer image can be effectively enhanced, the focus area is highlighted, and noise interference is inhibited at the same time.

The invention relates to the technical field of air traffic control and voice recognition, in particular to an air traffic control instruction end-to-end voice recognition method under a high-noise condition, which comprises the following steps: preprocessing to-be-recognized voice, and extracting to obtain original voice frequency characteristics; carrying out noise reduction intensity control, frequency band gain control and fundamental tone tracking processing on the original voice frequency characteristics by adopting a self-adaptive attention noise reduction module to obtain noise-reduced voice; preprocessing the noise-reduced voice, and extracting to obtain noise-reduced voice frequency characteristics; a voice recognition module is adopted to carry out shared encoder coding, connection time sequence classification decoder decoding and attention-based encoder-decoder decoding on the noise reduction voice frequency characteristics, and an output voice recognition transcription text is obtained; according to the invention, the voice recognition precision of the air traffic control instruction under the high-noise condition can be improved.

The invention provides a multi-band and multi-scale end-to-end pest detection method and electronic equipment, and belongs to the technical field of computer vision. The method comprises the following steps: in an encoder, firstly performing feature fusion and dynamic multi-scale feature enhancement DMFE operation on multi-scale features from a high-dimensional direction to a low-dimensional direction to realize full fusion of high-dimensional features and low-dimensional features so as to enhance the multi-scale feature extraction capability of a model; meanwhile, in an encoder, wavelet transform is utilized to perform frequency band decomposition on features after preliminary fusion, and dynamic cross-scale convolution attention noise reduction DCCD calculation is performed for high-frequency sub-bands with relatively concentrated noise content, so that the multi-scale feature extraction capability of the model is further improved while noise interference is suppressed. According to the method, the background noise interference resistance and the multi-scale target detection capability of the pest detection model are effectively improved, so that the detection precision and generalization capability of the model are improved.

The invention relates to the field of medical image processing and analysis, in particular to a cardiovascular image analysis system and method. The method comprises the steps of obtaining and preprocessing cardiovascular image data, converting the preprocessed cardiovascular image data into an initial feature space, optimizing cardiovascular image features in the initial feature space through recursion, and further constructing dynamic feature mapping to obtain a dynamic feature mapping value; adjusting the dynamic feature mapping value to obtain an optimized feature mapping value; a segmentation result is obtained by adopting block feature aggregation and a joint optimal iteration strategy, geometric features of the cardiovascular system are further extracted, and a vascular network graph is constructed; and detecting an abnormal region in a segmentation result, judging a lesion region, and finally generating diagnosis information. The problems that self-adaptive capturing capacity for dynamic features and local changes is lacked, analysis cannot be carried out in combination with a segmentation result in detection of a lesion area, and the segmentation result cannot comprehensively reflect the continuity of a cardiovascular structure are solved.

The invention relates to a speckle image quality evaluation and automatic correction system and method in a high-temperature environment, and belongs to the field of digital image processing. Comprising a speckle field preparation module and an evaluation and correction module. The speckle field preparation module comprises a support constraint assembly, a two-dimensional motion assembly, a speckle spraying assembly and an upper computer control assembly; the supporting and restraining assembly comprises a test piece, a back supporting plate, a side pressing block and a bottom supporting plate; the two-dimensional motion assembly comprises a motion supporting beam, a longitudinal direction motion assembly and a transverse direction motion assembly. The evaluation and correction module comprises two CCD (Charge Coupled Device) cameras, a data resolving upper computer and a high-low temperature test box; clear and reliable speckle image acquisition is realized by adopting a correction-improvement-correction automatic adjustment mode, speckle image preparation, quality evaluation and automatic correction in the high and low temperature environment are realized, accurate acquisition of data in digital speckle detection is further ensured, and deformation of a structure mechanism in the high and low temperature environment is reliably solved.

The invention relates to the technical field of remote sensing data processing and earth surface parameter inversion, and discloses an earth surface matrix parameter hyperspectral data inversion method and system based on machine learning. Comprising the following steps: constructing a multi-source heterogeneous hyperspectral data set; performing feature screening on the preprocessed hyperspectral data set based on an adaptive band selection algorithm, constructing a dynamic weight matrix by calculating mutual information entropy and inter-class distance measurement between spectral bands to realize intelligent screening of key feature bands, and combining spectral derivative conversion and spectral index calculation to generate an enhanced feature vector; and a multi-task transfer learning neural network model is constructed, and an output layer realizes multi-parameter collaborative inversion based on a multi-task learning architecture. And performing preprocessing and feature enhancement operation which is the same as that of the training data on the hyperspectral image data of the to-be-inverted region, inputting the trained neural network model, and outputting a surface matrix parameter inversion result.

The invention discloses a hyperspectral image classification method based on frequency domain denoising and element gradient correction, and the method comprises the following steps: carrying out the preprocessing of all hyperspectral image data, and dividing an overall training sample set formed by the processed hyperspectral images into a training set and a verification set; constructing a sample weighting model based on frequency domain denoising and element gradient correction; in the training process, a parameterization frequency spectrum gating sensing transformation module is utilized to map features to a frequency domain through discrete Fourier transform, a learnable frequency spectrum response function is utilized to adaptively suppress spectrum jitter noise, and finally pure features are reconstructed. And automatically constructing a high-confidence pseudo-clean verification set based on a Gaussian mixture model and time domain consistency. According to the method, a time domain momentum updating mechanism is introduced, the variance of statistical estimation is effectively smoothed, random interference caused by training fluctuation is resisted, and the accuracy of pseudo clean set construction and the convergence stability of overall model training are further improved.

The invention provides a current signal iteration Vold-Kalman filtering detection method for a rotating machinery fault, and belongs to the technical field of rotating machinery, and the method comprises the steps: carrying out the coordinate transformation of a three-phase stator current during the operation of the rotating machinery, and obtaining Id; obtaining a time-frequency spectrum of the original current, constructing a cost function, and extracting an instantaneous frequency ridge line of the fundamental wave of the original current; according to the instantaneous frequency ridge line and the component design parameters, calculating the characteristic frequency when each component breaks down; an iterative Vold-Kalman filter is constructed, the calculated fault frequency and the frequency multiplication of the fault frequency are taken as cut-off frequencies, and fault harmonics in the current signal Id are iteratively checked and extracted through a substitution check method; and carrying out time-frequency analysis on the extracted fault harmonic waves, and identifying the fault type. According to the method, a non-intrusive method is adopted, the non-stationary characteristic of the signal is considered, the calculation complexity is reduced, meanwhile, the fault harmonic extraction precision is improved, and the rotating machine fault can be effectively detected through the current signal under the variable rotating speed working condition.

The invention relates to a structured light illumination confocal super-resolution measurement system and method, and relates to the technical field of optical measurement. The system comprises a digital micromirror device (DMD), a sample platform, an electric displacement platform, a microscopic imaging assembly, a complementary metal oxide semiconductor (CMOS) imaging assembly and computer equipment, and the DMD is used for adjusting the point light source array and the structured light field. According to the DMD-based confocal super-resolution measurement method, a structured light field and a signal-noise separation technology are combined, and high-precision and high-resolution three-dimensional surface reconstruction can be realized. By using the flexibility of the DMD and the advantages of the super-resolution technology, the spatial resolution and speed of confocal imaging are effectively improved, so that the measurement process is more accurate. In addition, low-rank decomposition and self-supervised deep learning methods are fused, the image reconstruction process is optimized, noise interference is suppressed, and the overall performance of the system is further improved.

The invention discloses an agricultural machinery intelligent scheduling method and system based on yield prediction, and belongs to the technical field of intelligent agricultural machinery, and the method comprises the steps of data collection, data preprocessing, crop yield prediction, agricultural machinery intelligent scheduling and comprehensive agricultural machinery scheduling. According to the method, crop yield prediction is carried out by adopting the time convolution feature enhanced gated circulation network, so that the method has better long and short term time sequence dependence modeling capability, key features can be adaptively enhanced, the stability and accuracy of crop yield prediction are improved, and the reliability of subsequent agricultural machinery intelligent scheduling is improved; the discrete particle swarm optimization algorithm is adopted to perform agricultural machinery intelligent scheduling, so that the method has relatively strong global search capability, can adapt to different agricultural production environments and scheduling requirements, has significant advantages in large-scale agricultural production, and improves agricultural machinery scheduling efficiency; comprehensive agricultural machinery intelligent scheduling based on yield prediction is adopted, crop yield difference is fully considered, agricultural machinery operation is reasonably arranged, and resource waste is effectively avoided.

The invention discloses a temperature control system of an immersion type freezing device and relates to the technical field of immersion type freezing temperature control. Temperature control in the dipping and freezing process is achieved, it is ensured that large ice crystals are reduced in the food freezing process, damage to the interior of food is reduced, and meanwhile energy consumption and cost are reduced. The system comprises a temperature acquisition module, a liquid pump driving module, an upper computer, a data transmission module, a control module and a refrigeration equipment control module, the upper computer performs data interaction with the control module through the data transmission module; the temperature acquisition module acquires the temperature of the frozen object and the external environment, converts the acquired temperature signal into an electric signal and sends the electric signal to the control module; the control module controls the power of the liquid pump driving module through a cascade control strategy and a PID (Proportion Integration Differentiation) control algorithm according to the temperature signal of the frozen object and the influence value of the external environment on the freezing liquid; and starting and stopping of the refrigeration equipment control module are controlled according to the real-time temperature of the frozen object. The method is suitable for the field of dipping type freezing temperature control.

The invention discloses a sound event positioning and detecting method applied to three-dimensional scene reconstruction. According to the method, an identification network comprising a channel-space extrusion feature extraction module, a multi-scale pooling fusion module, a Transform network module and a full connection layer is constructed, preprocessed audio data is input into the identification network, and network parameters are optimized by using a mean square error loss function. Wherein a residual network module with an attention mechanism of the channel-space extrusion feature extraction module effectively filters noise and reverberation of a surrounding environment, and extracts a frequency band distribution feature corresponding to a sound event at a current moment; the multi-scale pooling fusion module enhances the feature representation capability, ensures that some important information in the audio is not lost, and can effectively promote the network to analyze the acoustic information of the environment where the network is located. The method provides the ability to identify and locate sound events from audio in the field of three-dimensional scene reconstruction.

The invention relates to the technical field of intelligent manufacturing, and particularly discloses a noise robust multivariable time sequence anomaly detection method for the field of intelligent manufacturing, which comprises the following steps: acquiring multivariable time sequence data; mapping the input data to a low-dimensional potential space, and restoring the input data to an original feature space to obtain reconstructed data; calculating a reconstruction error matrix of the original data and the reconstruction data, calculating a gradient matrix of the loss function relative to the input data, averaging and normalizing the gradient matrix along the time dimension to obtain a dimension importance score, and fusing the dimension importance score and the reconstruction error matrix to generate a weighted reconstruction error matrix; distinguishing the original data from the reconstructed data through a discriminator, and alternately optimizing parameters of the generator and the discriminator; and calculating an abnormal score based on the weighted reconstruction error matrix, introducing an abnormal diffusion mechanism to ensure the time sequence consistency, and judging that the sample with the abnormal score exceeding a threshold value is abnormal.

The application belongs to the field of integrated circuits, and specifically provides a low-power consumption enhanced reading type 11T CNFET SRAM cell circuit to solve the problems of slow speed, low stability, high power consumption and the like in the prior art; the application comprises N-type carbon nanometer field effect transistors MN1-MN9 and P-type carbon nanometer field effect transistors MP1-2, on the basis of a conventional 6T SRAM cell circuit, a dynamic load structure composed of transistors MN7, MN8 and MN9, and a high-resistance differential readout circuit composed of transistors MN3, MN4, MN5 and MN6; the adoption of P-type transistors realizes a bistable latch structure to reduce holding and writing power consumption, eliminate path competition phenomena caused by data flipping during writing, and improve writing speed; during reading, a resistive load tube is turned on to eliminate charge accumulation and ensure data storage stability; the high-resistance differential readout circuit reduces voltage disturbance of a storage node caused by bit line load; meanwhile, the differential writing / reading circuit can suppress common-mode noise and external interference on the bit line, reduce the probability of error sampling, and enhance overall anti-interference capability.

The application relates to the field of fault detection, and discloses a device fault detection method based on small sample learning and acoustic feature migration, a device and a medium, the method comprising the following steps: acquiring an acoustic signal of a target device; extracting a fusion feature vector; generating device-independent features based on a DANN network; generating a synthesized fault acoustic signal by using a Mel-CGAN; mixing the synthesized fault acoustic signal with an actual acoustic signal, and then training the DANN; and obtaining a fault detection result by using the trained DANN. The application can significantly reduce the dependence of fault diagnosis on labeled data, and can reduce the accuracy in a cross-device scene. The application can be widely applied in the fields of wind power and intelligent manufacturing.

The invention discloses an electronic device thermal characteristic evaluation method based on a Wiener filter. The method specifically comprises the steps that the junction temperature of an electronic device to be tested is obtained through thermal testing, a thermal impedance curve of the electronic device to be tested is obtained through calculation according to the junction temperature, and a thermal impedance element conversion derivative curve is obtained through element conversion processing and derivation processing; carrying out deconvolution on the thermal impedance transmutated derivative curve to obtain a time constant spectrum of the electronic device to be detected; obtaining a thermal characteristic evaluation result of the electronic device according to the time constant spectrum; in the deconvolution process, a Wiener filter is applied to the frequency domain form of a thermal impedance transmutated derivative curve, and a thermal characteristic representation function is introduced. Compared with two methods which are widely applied at the present stage, the modulation method is simpler, the accuracy is more close to the essence, and the obtained structure function can accurately reflect the thermal characteristics of each layer of the tested device.

The application provides a steam delivery pipeline leakage detection system, which comprises a metal conducting rod fixedly connected to the outer wall of a steam delivery pipeline, a sensor connected to the metal conducting rod, a signal analyzer electrically connected to the sensor, and a DCS control center electrically connected to the signal analyzer, and the sensor is used for collecting the vibration sound signal of the metal conducting rod. The steam delivery pipeline leakage detection system of the application can accurately and timely determine the leakage and the position of the leakage point by using the metal conducting rod connected to the steam delivery pipeline to transmit the vibration sound generated by the intense friction between the leakage medium and the pipeline wall, and the sound signal is not obviously attenuated in the transmission process due to the solid medium transmission, the transmission distance is long, and the sound signal is not easily interfered by other signals, the false positive rate of the leakage detection is effectively reduced, the operation and maintenance cost of the steam delivery pipeline is significantly reduced, and the industrial safety is improved.

The invention provides a cable joint partial discharge fiber bragg grating sound wave intelligent sensing and detection method, and relates to the technical field of cable partial discharge detection, and the method comprises the steps: obtaining multi-channel time-domain spectral response data of a detected cable joint through a fiber bragg grating sensing array, and demodulating the data to obtain a multi-dimensional feature vector group; decomposing the waveform signal into narrowband intrinsic mode components, extracting energy density through Hilbert transform and executing singular value decomposition, and screening dominant mode components to reconstruct the waveform signal to obtain stress wave time sequence characteristics; calculating a waveform high-order cross-correlation tensor as a hyperedge weight, extracting a time-frequency amplitude feature vector as an initial node feature, constructing a space-time hypergraph, and generating a node embedding vector through graph convolution; and initializing a Gaussian noise field based on a node embedding vector and setting condition information, executing reverse sampling denoising to generate a three-dimensional space probability density field to identify an initial positioning coordinate, and iteratively correcting the positioning coordinate and a wave velocity value until a residual error converges.

The present invention relates to a system and method for evaluating and automatically correcting a speckle image quality under a high temperature environment, and belongs to the field of digital image processing; the system and method comprise a speckle field preparation module and an evaluation and correction module; the speckle field preparation module comprises a support constraint component, a two-dimensional motion component, a speckle spraying component, and a host computer control component; the support constraint component comprises a test piece, a back support plate, a side clamping block, and a bottom support plate; the two-dimensional motion component comprises a motion support beam, a longitudinal motion component, and a transverse motion component; the evaluation and correction module comprises two CCD cameras, a data solving host computer, and a high and low temperature test chamber; a correction-improvement-correction automatic adjustment mode is adopted to realize clear and reliable speckle image acquisition, and the present invention realizes speckle image preparation, quality evaluation, and automatic correction under high and low temperature environments, further ensures accurate data collection in digital speckle detection, and realizes reliable solution of deformation of structural mechanisms under high and low temperature environments.

The invention discloses a wind speed measurement method and system based on a bistatic laser wind finding radar with separate transceiving, the wind speed in the atmosphere is accurately measured by transmitting a laser beam with a specific wavelength and using the Doppler frequency shift principle, the system integrates a plurality of modules, and the system is simple in structure and convenient to operate. Comprising an environmental parameter acquisition module, a Doppler frequency shift correction module, a signal processing module and an intelligent layout optimization module. By collecting environmental parameters in real time, the system dynamically adjusts Doppler frequency shift correction factors to compensate errors caused by environmental changes; the multi-stage signal processing module improves the signal-to-noise ratio and accuracy of wind speed measurement through multi-scale feature extraction, depth separable convolution and adaptive noise suppression technologies; the intelligent layout optimization module dynamically adjusts the position configuration of the base station based on a depth Q network algorithm, and ensures the measurement precision and stability in a complex environment; the method significantly improves the precision and reliability of wind speed measurement, and is suitable for the fields of wind power plant optimization layout, meteorological monitoring, aerospace flight plans and the like.

This application belongs to the field of visual compensation technology and provides a visual compensation method under starlight conditions. Through pre-sampling and grayscale variance statistics, it achieves the determination of starlight compensation intervals and the dynamic division of target and background regions in the imaging plane. It adopts a sampling method with different exposure time series for the target and background regions to obtain multiple frames of original sampled data and pixel integration time. Based on the pixel integration time, it constructs a spatially variable gain matrix and completes inter-frame registration and gain normalization processing to separate signal and noise components in the image. The signal component is used as a sparse sampling stream, and the pixel variance distribution of the noise component is used as a hyperparameter of the variational inference algorithm. Image reconstruction is completed through probability density gradient estimation, making the variational inference process match the actual noise distribution characteristics under starlight conditions, thereby improving the quality and reliability of visual imaging under starlight conditions.

This invention relates to the field of energy forecasting technology, and in particular to a method and system for short-term energy load forecasting based on an improved Transformer using SE-Block. The method includes reversible normalization preprocessing of acquired multivariate load sequence data; feature extraction and fusion of the preprocessed data using improved cross-scale interactive patching, including multi-scale feature extraction, cross-scale interactive alignment, residual correction, and dynamic fusion; and feature filtering of the fused features based on a channel attention mechanism, including feature response based on improved SE-Block and nonlinear interaction of context vectors. This invention addresses the non-stationarity of actual load caused by meteorological conditions and user behavior. By automatically eliminating noise interference among multiple variables, the model accurately depicts the fluctuation details of the load curve, demonstrating its robustness in multivariate load forecasting for integrated energy systems.

The invention discloses a total variation regularization depth expansion network method for hyperspectral image unmixing. The method comprises the following steps: inputting hyperspectral image data and an end member spectrum library; constructing a hyperspectral data unmixing model fusing a sparse constraint term and a total variation regularization term and an optimization objective function of the hyperspectral data unmixing model; converting the optimization objective function into an equivalent augmented Lagrangian form, and performing iterative solution on each variable; expanding an iteration process into a deep learnable network structure with a plurality of layers, and parameterizing hyper-parameters and a plurality of approximate operators in iteration into learnable network parameters; constructing a comprehensive loss function, and continuously optimizing network parameters by using back propagation and gradient descent methods; according to the method, total variation regularization prior is fused into a sparse unmixing model, the sparse unmixing model is expanded into a learnable end-to-end neural network through iterative solution, and the method has excellent unmixing precision and robustness for noise-containing data.

The application provides a temperature control system and an air conditioner; the temperature control system comprises a control unit, a power supply unit, an ambient temperature detection unit and an evaporator coil temperature detection unit; the ambient temperature detection unit comprises a ring temperature sensor assembly and a first inductor, the evaporator coil temperature detection unit comprises an inner coil sensor assembly and a second inductor, the control unit is used for acquiring the ambient temperature collected by the ring temperature sensor assembly and the evaporator coil temperature collected by the inner coil sensor assembly, and controlling the temperature of the air conditioner according to the ambient temperature and the evaporator coil temperature; the system structure not only realizes the temperature control of the air conditioner, but also plays a role in inhibiting common mode and differential mode noise interference by arranging the inductors in the ambient temperature detection unit and the evaporator coil temperature detection unit respectively, improves the EMC test qualified rate of the air conditioner, and has a simple structure and good practical value.

The invention provides a non-contact pupil state monitoring and physical sign data acquisition system. The non-contact pupil state monitoring and physical sign data acquisition system comprises a high-resolution non-contact camera device which is installed within a proper distance and used for capturing high-definition eye images, especially pupil shapes and dynamic changes, when a user opens eyes, the camera device can automatically recognize and track pupil positions of the user, and the camera device is used for monitoring the pupil states of the user. Even if the head of the user slightly moves, the pupil image can still be accurately captured. According to the system, through the high-resolution non-contact camera device and the multi-band infrared light source module, accurate monitoring of the pupil state of the user is achieved, even if the head of the user slightly moves, the camera device can still automatically track the pupil position, and the accuracy of data is ensured.

The application discloses a kind of based on acoustic wave FMCW's respiratory frequency detection method, belong to acoustic wave signal processing technical field;Method is: utilize mobile phone loudspeaker to generate a segment frequency continuous change acoustic wave signal, form continuous acoustic wave detection sequence;Mobile phone microphone receives the acoustic wave signal that is reflected back synchronously;To the acoustic wave signal that is reflected back received filtering processing;With reflected signal, the distance change curve of thoracic cavity with breathing movement is generated by autocorrelation operation to transmission signal;Distance change curve is carried out smoothing processing;The main frequency frequency is obtained by fast fourier transform, and respiratory frequency is calculated.The present application effectively suppresses environmental noise interference by using the advantages of FMCW technology, analyzes the autocorrelation coefficient of reflected signal, accurately extracts the thoracic movement information related to respiration, and realizes the accurate detection of human respiratory frequency;Using autocorrelation function is easier to find peak, more easily tracked nature, greatly increase the convenience of user.

This invention provides an enzyme turnover rate prediction method based on a dual-route hybrid expert mechanism, belonging to the field of bioinformatics technology. It solves the problems of low data utilization, insufficient information mining, and poor robustness in existing prediction methods due to missing values ​​and temperatures. The technical solution includes the following steps: S1: Constructing a unified standard enzyme turnover rate dataset; S2: Extracting multimodal embedding features; S3: Constructing intra- / inter-modal encoders; S4: Combining modalities into an expert hybrid module; S5: Designing an attention fusion mechanism. This invention can achieve high-precision prediction under complex in vitro environmental conditions.

This application provides a multimodal ultrasound intelligent identification method and precision diagnosis system for thyroid lesions. The method determines image data blocks with multimodal spatial registration and grayscale-contrast temporal alignment. Based on the grayscale ultrasound image sequence in the image data blocks, a first depth feature map of the target user's thyroid region is determined. A second depth feature map of the target user's thyroid region is determined based on the elastography map in the image data blocks. The hemodynamic spatial distribution characteristics of the parametric map in the image data blocks are captured to determine a third depth feature map of the target user's thyroid region. Cross-modal fusion is performed on the first, second, and third depth feature maps, and then the lesion segmentation mask, benign / malignant probability value, and risk classification of the target user's thyroid region are inferred and output in parallel. Using the scheme of this application, high-precision spatiotemporal alignment and cross-modal depth feature fusion of multimodal ultrasound images can be achieved to complete end-to-end intelligent identification of thyroid lesions.