1272 results about "Feature transformation" patented technology

The invention relates to a display screen backlight brightness adjusting method and system based on a Mini QLED technology. The method comprises the following steps: dividing a display image signal into a brightness signal, a chrominance signal and a saturation signal, and generating image feature data; performing sub-pixel-level boundary detection to obtain a plurality of initial backlight sub-regions, and calculating region brightness distribution data; executing region clustering operation and region merging to obtain a plurality of target backlight sub-regions and calculating backlight partition mapping data; carrying out nonlinear feature transformation, and calculating backlight unit control data of the Mini LED backlight unit; outputting the backlight compensation amount of each target backlight sub-region by adopting a deep Q network; and performing real-time feedback correction, calculating a deviation value between an actual display parameter and a target parameter of each target backlight sub-region, and outputting a backlight optimization control signal. According to the invention, the backlight brightness is accurately adjusted, and the consistency of the display effect is improved.

The present invention provides a multimodal data closed-loop management method and system based on government affairs, which relates to the field of data processing technology. The method comprises: acquiring multimodal data from multiple information channels; performing modal classification, and extracting features of text information, voice information and image information respectively; performing feature conversion and normalization processing, and generating a cross-modal unified feature vector set; identifying cross-channel complaint information of the same event; matching with preset complaint category labels, and identifying complaint categories; determining responsible departments, and generating task assignment instructions; acquiring the processing status corresponding to the task assignment instructions, constructing task progress records, and generating early warning information when the task times out or the processing progress is abnormal; generating feedback information, and returning the feedback information through a communication channel corresponding to the complainant, and updating the processing status corresponding to the task assignment instructions at the same time. The present invention improves the autonomy and accuracy of multimodal data closed-loop management.

The invention discloses an infrared ship detection method based on an improved RT-DETR algorithm, and aims to improve the detection precision and efficiency of a ship target in an infrared image. By introducing a multi-layer dynamic feature conversion module (MDST), the diversity of feature extraction is enhanced; effective fusion of multi-scale features is realized by using a high-level screening feature fusion pyramid network (HS-FPN); an original RT-DETR backbone network is replaced by a PPA and C2f structure, so that the problem of information loss in a down-sampling process is solved; and the CA module is used for enhancing the feature selection and fusion effect. Experimental results show that the provided method significantly improves the detection performance on an infrared marine ship data set, is suitable for resource-constrained equipment such as unmanned aerial vehicles and smart phones, and has a wide application prospect.

The invention provides a physiological signal self-supervision representation learning method and system based on time-frequency reconstruction, and belongs to the technical field of signal processing and mode recognition. Comprising the following steps: acquiring original multi-channel physiological signal data and converting the data into a time-frequency diagram; carrying out segmentation / blocking, random masking and feature conversion on the physiological signal and the time-frequency diagram thereof to obtain a primary time domain representation and a primary frequency domain representation; coding and fusing the preliminary time domain representation and the frequency domain representation in a time dimension and a channel dimension to obtain a time-frequency mixed representation; pre-training an encoder by adopting a reconstruction learning method based on the time-frequency mixed representation; a pre-trained encoder is used for obtaining multi-channel time-frequency representation of to-be-processed original multi-channel physiological signal data, the multi-channel time-frequency representation is mixed to obtain time-frequency mixed representation, and the time-frequency mixed representation serves as learned physiological signal self-supervision representation and is used for a physiological disease classification task. The method has higher generalization ability and adaptability, and an effective solution is provided for a physiological signal classification task.

A method of performing Cross-Domain Spatial matching (CDSM) for monocular object detection in 3D free space surrounding a vehicle includes receiving into an image processing network, from a camera of the vehicle, at least one input image, determining from the input image, a set of 2D image features of potential targets in surrounding the vehicle, transforming the set of 2D image features into a set of 3D image features of the potential targets by aligning a lateral axis and a vertical axis of the 2D image features with, respectively, a tensor height axis and a tensor width axis of a 3D birds eye view grid, applying a CDSM aggregation to the 3D image features to extrapolate depth information, and generating a set of aggregated 3D features to the potential targets, and detecting, based on the aggregated 3D features, one or more objects associated with the potential targets.

The invention relates to an electric hand drill wear state prediction and health management system, which belongs to the technical field of intelligent fault diagnosis and predictive maintenance of industrial equipment, and comprises a data acquisition and preprocessing unit used for acquiring and processing a multi-modal physical signal to generate a standardized data frame; the multi-domain feature transformation unit is used for receiving the standardized data frame and transforming the standardized data frame into a health feature vector and a load feature vector; the dynamic health baseline construction unit is used for reconstructing and generating a dynamic health baseline through a depth generation model according to the time sequence of the health feature vector and the load feature vector; and the residual error sequence generation and statistical monitoring unit is used for calculating the distance between the health feature vector and the dynamic health baseline, generating a residual error sequence, and performing statistical processing on the residual error sequence to obtain a statistical magnitude. According to the invention, the interference of working condition change on health state assessment is eliminated, and pure and reliable data input is provided for subsequent accurate monitoring.

The invention relates to the technical field of solid state disks, and discloses a load collaborative scheduling method and system for a solid state disk, and the method comprises the steps: carrying out the data collection of a main controller of the solid state disk, and obtaining the original feature data; a Kolmogorov-Arnold neural network is input to carry out nonlinear feature transformation and linear combination operation, and a load prediction result is obtained; performing hierarchical processing to obtain a hierarchical scheduling model; constructing a multi-objective optimization function and carrying out constraint setting to obtain a constraint optimization equation set; performing convex difference function decomposition and augmented Lagrange function construction on the constraint optimization equation set, and solving through an alternating direction iteration method to obtain a scheduling decision scheme; according to the scheduling decision scheme, load balancing control is carried out on data write-in physical address allocation, data read cache scheduling and multi-channel parallel transmission, a real-time performance monitoring index is obtained, and the method realizes load collaborative scheduling of the solid state disk.

The invention discloses a data security dynamic protection method based on artificial intelligence, and relates to the field of network security, and the method comprises the steps: employing an anomaly detection algorithm to recognize malicious data streams and attack features according to a data stream log; according to the identified malicious data flow, a multi-agent reinforcement learning algorithm is adopted, and a protection strategy is adjusted through multiple agents according to the current environment state and attack characteristics; constructing a simulation environment, and training and evaluating an anomaly detection algorithm and a multi-agent reinforcement learning algorithm by generating a confrontation sample and a confrontation strategy; wherein the adversarial sample is generated by performing feature transformation on the malicious data stream and the normal data stream, and the adversarial strategy is generated by performing random disturbance on the existing attack strategy. In view of poor adaptability of a protection strategy for malicious data streams in the prior art, the malicious data streams and attack features are identified through an anomaly detection algorithm, and multi-agent reinforcement learning, collaborative optimization of flow control and a honeypot deployment strategy are adopted, so that the adaptability is improved.

The invention provides a cluster network flow prediction method based on multi-scale time feature fusion, and belongs to the technical field of computer network flow prediction. The method comprises the following steps: determining a multi-index prediction sequence based on traffic load characteristics of cluster IP instances, and constructing a high-quality time sequence data set; fourier transform and discrete wavelet transform are used for time-frequency feature analysis, and noise filtering and data dimension reduction are completed; projecting sequences of different time granularities to a unified model dimension, performing one-dimensional channel convolution merging, inputting the merged sequences into a time encoder and a cross-channel encoder, and capturing cross-scale long-term time dependence and a coupling relationship between variables; in the loss function design, time domain and frequency domain loss are fused, double-domain error calculation is carried out on a prediction result and a label through Fourier transform, and the robustness of a model to non-stationary fluctuation is enhanced; and through linear layer decoding and reverse normalization processing, the abstract feature is converted into an actual flow prediction value. According to the invention, the precision and reliability of cluster network flow prediction are significantly improved.

The invention relates to the technical field of industrial equipment state monitoring and intelligent fault diagnosis, and discloses an industrial fault diagnosis method and system based on natural language fault features and large model knowledge enhanced reasoning, and the method comprises the steps: obtaining a multi-source monitoring and state analysis result from detected equipment in an equipment operation stage; converting the features into a fault feature text of a structured natural language; carrying out vectorization coding on the fault feature text, executing similarity retrieval in a pre-constructed industrial fault knowledge base, and recalling knowledge fragments; and based on the fault feature text and the recall knowledge fragment, constructing a reasoning prompt word, inputting the reasoning prompt word into a large language model for knowledge enhanced reasoning, and generating a diagnosis result. According to the method, the problems that numerical evidence and semantic knowledge are difficult to unify, the knowledge coverage and updating cost is high, and cross-working-condition migration and conclusion consistency are insufficient in an existing intelligent diagnosis technology based on a rule base or a knowledge graph are effectively solved.

The invention discloses a reliability analysis method and system for a power system, and relates to the technical field of intelligent operation and maintenance of the power system, and the method comprises the steps: collecting data, and constructing a dynamic space-time atlas network of the power system, which comprises time dimension information and space dimension information; feature extraction is carried out based on a spatial-temporal feature learning framework constructed by a dual-channel graph attention network and a gated causal convolution network, and spatial-temporal fusion features containing complete spatial-temporal information are generated by using gated causal convolution and a time attention capture time sequence evolution mode; and converting the space-time fusion features into a local risk state and a global system state through a state feature mapping network, calculating a local component risk index and a system overall reliability index, generating a collaborative decision strategy, and realizing reliability analysis of the power system. According to the method, the intelligent operation and maintenance level of the power system is remarkably improved, the shutdown time and the maintenance cost of the system are reduced, and meanwhile, the system performance is continuously improved through preventive maintenance and an active optimization strategy.

A computer-program product, computer-implemented method, and computer-implemented system includes obtaining a raw dataset; executing an outlier filtration process based on obtaining the raw dataset; training a model using a refined outlier-reduced dataset; and predicting, via the trained model, a value of the target entity at a future time.

The invention discloses an environmental protection facility state monitoring and diagnosing method and system based on vibration and infrared detection, and the method comprises the following steps: collecting vibration and temperature data of environmental protection facility equipment, removing noise and environment interference through a filtering algorithm, and carrying out the standardization processing; extracting features of the data through a space-time convolutional network; performing deep fusion on the features by using an adaptive feature transformation network; performing abnormal mode detection; analyzing potential reasons of the abnormal mode by using a causal inference model, and generating a fault diagnosis report; a real-time processing strategy is generated, and the operation state of the environmental protection facility equipment is controlled in real time; and the fault diagnosis result and the instant processing strategy are uploaded to a cloud management platform, a notification is sent to related maintenance personnel, and remote monitoring and collaborative maintenance are realized. According to the method, the space-time convolutional network and the causal inference model are adopted, the faults of the environmental protection facility equipment are effectively identified, a proper emergency processing strategy is generated according to the real-time monitoring data, and the accuracy and response efficiency of equipment fault diagnosis are improved.

The invention discloses an injection molding process fault diagnosis model training method and system based on a large language model and a fault diagnosis method. The model training method comprises the following steps: collecting and cleaning process parameters under the fault working condition of the injection molding machine, converting the process parameters into a natural language text, combining the natural language text with a fault label to construct a textualized data set, and dividing the textualized data set into a training set and a verification set according to a proportion; and in combination with the text data dimension and the fault category number, loading the pre-trained large language model and configuring a diagnosis model structure in a quantitative mode. And inputting the training set into a model to extract semantic features, processing the semantic features by a feature conversion module to generate a high-order feature vector, and inputting a classification head to output a fault category probability. And back propagation is carried out by using a loss function, and model parameters are efficiently and finely tuned in combination with low-rank adaptation and a layered freezing strategy. And repeating training until the performance reaches the standard, and outputting a final diagnosis model. The method is efficient in training, and can effectively reduce the maintenance and use cost of the model.

The invention relates to the technical field of speech processing, can be applied to business scenes of medical health, financial science and technology and the like, and discloses a speech enhancement method based on multi-scale feature learning, which comprises the following steps: framing an input audio signal, extracting a Mel-frequency spectrum feature, extracting a frequency domain feature by using a multi-scale convolutional neural network, and carrying out multi-scale feature learning on the frequency domain feature; carrying out coding dimension reduction on the image; noise is suppressed through a deep residual network, enhanced audio features are generated, a non-autoregression generative model is adopted for feature conversion, and finally a generative adversarial network is used for reconstructing a target voice waveform. Voice frequency domain features are extracted through the multi-scale convolutional neural network, and the feature expression ability of different frequency bands is improved; noise suppression is carried out through a deep residual network, and the purity of the voice signals is enhanced; feature conversion is optimized through a non-autoregression generation model, and the modeling efficiency of speech enhancement is improved; the target voice waveform is reconstructed through the generative adversarial network, and the naturalness and definition of the generated voice are improved.

The invention relates to the technical field of image analysis and processing, in particular to a high-precision detection method based on image analysis, which comprises the following steps: acquiring an original image to be detected, performing preprocessing operation on original image data, including normalization processing, de-noising processing and brightness adjustment, performing multi-scale feature conversion on the preprocessed image to generate feature matrixes of different scales, calculating the correlation degree between the feature matrixes through a correlation self-adaptive fusion module, and generating a fusion feature map; calculating a significance score of each pixel point in the fused feature map, generating an abnormal score matrix, and determining a significant abnormal region in combination with the deviation matrix; original image information is decomposed into different scales through a multi-scale feature conversion module, local and global abnormal features of an image are captured by using comprehensive information of feature matrixes of different scales, information redundancy among the features of the scales is eliminated, and the identification degree and the detection accuracy of an abnormal region are improved.

The invention relates to the technical field of wireless communication, in particular to a wireless communication anti-interference enhancement system based on multi-mode signal fusion. Comprising an intelligent signal transceiving module, a signal sensing and preprocessing module, an interference feature coding and classification module, an interference suppression decision module, an interference suppression module, a signal fusion module, an interference suppression performance evaluation module, an interference feature library and an interference suppression strategy library, the interference feature coding and classification module adopts a three-layer progressive architecture of multi-dimensional manifold feature extraction, feature transformation and coding inspired by a group theory, and deep learning classification based on chaotic dynamics, extracts interference features from three dimensions of time domain, frequency domain and nonlinearity, and realizes accurate interference identification; the signal fusion module adopts a multi-level fusion strategy, and dynamically adjusts the fusion weight according to the signal quality; the interference suppression module selects an optimal suppression strategy according to the interference type, and executes frequency domain cancellation, space domain cancellation and time domain cancellation; and performance evaluation feedback forms a closed-loop optimization mechanism.

The invention discloses a panoramic 3D occupancy prediction method based on 3D Gaussian sputtering. The method comprises the following steps: performing 2D feature extraction on a multi-view RBG image by using a 2D image feature extraction network; converting the 2D image features into 3D features through a deep prediction network; extracting the 3D features by using a 3D feature extraction network; performing 3D Gaussian sputtering on the 3D features, and then rendering the 3D features to a final 2D image; a loss function is set, and a 2D true value is used for supervision; and constructing a dense 2D semantic and deep truth value data set. According to the invention, effective supervision of 3D occupancy prediction is realized, and the training speed and prediction accuracy are improved.

The invention discloses a cross-modal target detection method based on learnable Fourier transform, and mainly solves the problem of insufficient fusion of a visible light image and an infrared image in a complex scene due to inter-domain difference in the prior art. According to the implementation scheme, the method comprises the steps that bimodal features are extracted through a double-flow CSPDarknet53 network; a target position guiding module is utilized to enhance target area representation and suppress background interference; the features are converted to a frequency domain, and amplitude texture information of the visible light image and phase contour information of the infrared image are adaptively enhanced through a learnable frequency domain feature enhancement module; suppressing noise through global filtering and then inversely transforming back to a spatial domain; and finally, outputting a target detection result of the multi-modal image by the detection head. According to the method, frequency domain physical characteristics are fully utilized, full complementation and adaptive fusion of cross-modal features are realized, the detection precision and robustness of vehicles, pedestrians and other targets under low-illumination and complex backgrounds are remarkably improved, meanwhile, high calculation efficiency is kept, and the method can be applied to the fields of automatic driving, intelligent monitoring and the like.

The present invention relates to the technical field of expression recognition, and in particular, to an expression recognition method and system based on multi-scale features and spatial attention. The method includes: performing feature extraction on acquired facial image data by using an HNFER neural network model to obtain an original input feature map; performing pooling and concatenation on extracted features based on a CoordAtt attention mechanism to obtain a feature map; performing deep convolution processing on the feature map to obtain an attention map, and then performing element-by-element multiplication to obtain a final feature map; and performing feature transformation and normalization on the final feature map to obtain an expression category probability and output the expression category probability. In the present invention, by integrating scale perception and spatial attention technologies, the model can recognize and classify different emotional states more accurately and maintain high performance even under complex environmental conditions.

The invention provides an intelligent classification method for Beidou positioning data, and relates to the technical field of Beidou positioning data classification, and the method comprises the steps: obtaining the Beidou positioning data of a monitoring target; performing standardization processing, abnormal value processing and time sequence alignment processing on the acquired Beidou positioning data to generate a preprocessed time sequence data matrix; constructing a classification model; performing feature extraction, including time feature extraction and spatial feature extraction, on the preprocessed time sequence data matrix by using a spatial-temporal feature extraction module to obtain a time feature vector and a spatial feature vector respectively; carrying out fusion processing on the time feature vector and the space feature vector by adopting a spatial-temporal feature fusion module, and constructing a spatial-temporal joint feature; and inputting the spatio-temporal joint features into a classification module, and obtaining a classification result through feature conversion, attention weight calculation and feature aggregation. According to the invention, accurate identification and classification of the motion state of the monitored target are realized by establishing a data processing and classification framework of the system.

The invention relates to a turbine starter appearance defect detection method based on multi-scale feature extraction, and the method comprises the steps: collecting appearance defect data of a turbine starter, and defining defect types to obtain an appearance defect feature map; the image is input into a data enhancement module, parameters are adjusted according to image types, and a multi-attribute defect feature map is output; constructing a feature extraction module, extracting defect data features in the multi-attribute defect feature map, converting the defect data features into multilayer feature mapping, and generating a multi-scale defect feature map; inputting the multi-scale defect feature map into an enhanced feature extraction network, obtaining and aggregating high-dimensional multi-scale defect features, and generating a multi-scale fusion defect feature map; constructing a defect detection module, and detecting bounding box prediction and classification results of multi-scale defects; and carrying out training and weight updating on the model by adopting similarity and union set intersection loss to obtain a final defect detection model. According to the invention, accurate positioning and classification of defects of various targets with different sizes on the appearance of the turbine starter are realized.

The invention belongs to the technical field of agricultural machinery track classification, and particularly relates to an agricultural machinery track data classification method and device, electronic equipment and a storage medium. The method comprises the following steps: carrying out preprocessing and feature conversion operation on original agricultural machine GNSS trajectory data to obtain continuous and uniform time sequence trajectory data; the original agricultural machine GNSS trajectory data comprises a plurality of discrete trajectory points, and the category of the trajectory points comprises a field trajectory and a road trajectory; a trajectory classification model comprising a feature extraction module and a classification head is constructed, the time sequence trajectory data serve as training samples to train the trajectory classification model, the feature extraction module is trained in a comparative learning mode, and the classification head is trained through cross entropy loss; and deploying the trained trajectory classification model, and classifying agricultural machinery trajectory data by using the trained trajectory classification model. According to the invention, a contrast learning framework fused with the dynamic negative sample queue is designed, the classification accuracy and generalization ability are improved, and the method is suitable for track classification tasks in various complex scenes.

The invention discloses a data knowledge base management method and device for the egg industry and a medium, and relates to the technical field of intelligent breeding informatization, and the method comprises the steps: inputting a multi-modal fusion data set into a single deep learning framework, carrying out the multi-scale feature coupling of each modal feature through a self-adaptive nonlinear interaction function, and carrying out the multi-scale feature coupling; the feature weight is dynamically adjusted in combination with an enhanced feedback mechanism, and a multi-dimensional health assessment index set is generated by utilizing end-to-end joint feature extraction and nonlinear reasoning; performing entity recognition and relation extraction on the multi-dimensional health assessment index set, constructing an egg industry knowledge graph in combination with the breeding environment and the individual feature information, and performing structure optimization and dynamic evolution to obtain a dynamic evolution knowledge graph; and performing rule reasoning and data analysis on the dynamic evolution knowledge graph to generate a health management decision. Fine modeling and high-quality unified expression of multi-source data are achieved through piecewise nonlinear feature transformation and weighted fusion, and the accuracy of health state analysis input is ensured.

The invention provides a karst area pile foundation intelligent positioning and precise construction method which comprises the steps that a karst development characteristic data set of a target engineering area is obtained, and the karst development characteristic data set comprises a three-dimensional geological exploration data set and a karst attribute data set; and preprocessing the data set to obtain a karst preprocessed data set. And based on a feature extraction model, performing feature extraction on the preprocessed three-dimensional geological exploration data set to generate a three-dimensional geological feature map. And identifying the map through a karst cavity identification model to generate a karst cavity distribution thermodynamic diagram. And karst cave feature conversion processing is conducted on the karst preprocessing data set, and a pile foundation construction feature data set is obtained. And according to the karst cavity distribution thermodynamic diagram, constraint processing is conducted on the pile foundation construction characteristic data set, and a pile foundation constraint boundary diagram is generated. And outputting a pile foundation end hole positioning coordinate set according to a pile foundation positioning optimization algorithm and the pile foundation constraint boundary diagram. The construction safety and accuracy of the pile foundation in the karst area can be improved, and the construction risk and cost are reduced.

The invention discloses a lightweight underwater sonar target detection model based on Transform architecture, which comprises a backbone network, a lightweight hybrid encoder and a decoder, and is characterized in that the backbone network inputs the features of the last three stages of an input image into the encoder; the encoder converts the multi-scale features into a series of image features; the decoder fuses the loss functions WCIoU and NWDLoss while capturing the category and position information of the object, and generates the category and bounding box by iteratively optimizing the object query. According to the method, the backbone network adopts the depth separable convolution TConv to reduce the calculation complexity, and the feature processing and characterization capability of the backbone network is remarkably improved by fusing the index moving average and the gating attention mechanism; a query selection mechanism based on WCIoU perception is combined with an NWD loss function, collaborative optimization of classification information and position information is achieved, and the detection capacity for small targets is particularly enhanced.

The invention relates to the technical field of geological survey, in particular to a potential landslide hidden danger identification method based on rock mass deterioration characteristics. The technical problems that in the prior art, in landslide monitoring, the monitoring range is limited, data collection is not comprehensive, and early warning timeliness is insufficient are solved; according to the technical scheme, the potential landslide hidden danger identification method based on the rock mass degradation characteristics comprises the steps that feature extraction is conducted on a preprocessed image through a convolutional neural network, the rock mass degradation characteristics are identified, and the identified degradation characteristics are converted into numerical value characteristics; compared with the problems of limited monitoring range, incomplete data collection and insufficient early warning timeliness in landslide monitoring in the prior art, the intelligent machine vision technology is used for identifying the rock mass degradation characteristics, and the sensor network is deployed in combination with the Internet of Things technology to monitor the rock mass state in real time; the cloud computing platform is used for data analysis and processing, and finally intelligent early warning of the landslide hidden danger is achieved.

The invention discloses a multi-modal time sequence hybrid query system and method. The hybrid query method comprises the steps of collecting multi-modal time sequence data, performing normalization, denoising and feature extraction on a numerical time sequence, performing cleaning, conversion and classification feature extraction on text data, and performing enhancement and feature extraction on picture data; the statistical features of the numerical time series are converted into vectors, similarities of different modes are calculated, comprehensive similarity measurement is constructed, and a neighbor graph index and a multi-layer index structure are constructed based on the measurement; the method comprises the following steps: analyzing a mixed query statement input by a user, extracting a multi-modal keyword and a natural language description, and converting the natural language description into a multi-modal feature vector; according to the method, a comprehensive similarity measurement and multi-layer index structure is constructed, so that related data can be quickly positioned and matched in the query process, the problems of low multi-modal search efficiency and poor accuracy are solved, the combination of a time sequence and multi-modal query is realized, and the query efficiency is improved.

The invention relates to the technical field of image optimization, and discloses an image lossless fusion method and system based on depth feature extraction, and the method comprises the following steps: collecting multi-modal input data from different sensors, carrying out the preprocessing of each modal in the multi-modal input data, and carrying out the preprocessing of each modal in the multi-modal input data; the multi-modal input data comprises a visible light image, an infrared image and a depth map, a multi-layer pulse neural network is used for simulating pulse emission characteristics of biological neurons, time sequence feature extraction is performed on the multi-modal input data, and feature representations of different levels are captured. According to the method, deep feature extraction is carried out through the SNN algorithm to carry out image fusion, the problems of feature mutual exclusion and modal conflict when infrared and visible light images are processed through a traditional method can be effectively solved, a concrete image can be converted into an abstract feature, and then the abstract feature is converted into an end-to-end conversion process of the fused image. And the real-time and fidelity requirements in a dynamic scene are met.

According to the motor anomaly detection model training method and the motor anomaly detection method provided by the invention, a plurality of audio samples and category labels thereof during motor operation are acquired, a plurality of disturbance audios are generated through speed disturbance for each audio sample, the disturbance audios are input into a frequency domain feature extraction sub-model and a time domain feature extraction sub-model, and the frequency domain feature extraction sub-model and the time domain feature extraction sub-model are extracted. Performing linear transformation and weighted summation on the enhanced wavelet time-frequency spectrum and the enhanced time-domain features through a full connection layer to generate fusion enhanced features, performing feature transformation on the fusion enhanced features through a deep neural network, inputting the fusion enhanced features into a classifier to obtain a classification probability, and calculating triple loss according to a category label and the classification probability; and updating parameters of the frequency domain feature extraction sub-model, the time domain feature extraction sub-model, the full connection layer, the deep neural network and the classifier through loss back propagation and a stochastic gradient descent algorithm to obtain a trained motor anomaly detection model. Therefore, the accuracy of motor state recognition of the model in a high-noise environment is improved.