204 results about "Model sample" patented technology

The invention relates to a near-infrared coal quality detection method based on multi-dimensional data fusion so as to improve the coal detection precision. According to the method, a modeling sample set containing multi-dimensional data of the producing area of a coal sample, granularity characteristics, the distance between the coal sample and a probe, air humidity and environment temperature is constructed, near infrared spectrum data of the coal sample is collected, spectrum preprocessing, feature extraction and data fusion are carried out, and finally a quantitative analysis model is constructed to predict the coal quality characteristics in a high-precision mode. According to the method, a one-dimensional convolutional neural network (1d-CNN) technology is adopted, automation and intelligence of coal quality detection are realized, and the detection efficiency is improved. In addition, accurate data of moisture, ash content and volatile components in the coal sample are obtained strictly according to a national standard experimental method, and the representativeness of the sample and the reliability of the data are improved. Noise interference is effectively reduced through the preprocessing step, and scattering influence is corrected. The method provides powerful technical support for quality control, reasonable utilization of resources and the like in the coal industry.

The invention provides a diesel engine misfire fault diagnosis method based on a GRU and a PINN. The diesel engine misfire fault diagnosis method comprises the following steps of diesel engine multi-source data acquisition, data preprocessing and model sample generation; designing a gating circulation unit, introducing a physical constraint embedding module, constructing a misfire fault diagnosis model, and fusing the misfire fault diagnosis model with a long-short term memory model and a physical neural network; performing joint training and optimization on the misfire fault diagnosis model in the step S2 by using the samples in the step S1; and deploying a misfire fault diagnosis model and performing real-time diagnosis. The method has the beneficial effects that the time sequence data-physical model dual-drive diagnosis of the misfire fault of the diesel engine is realized; the accuracy of misfire diagnosis is improved; and the false alarm rate of complex working conditions is effectively reduced.

The invention belongs to the technical field of defect detection, and relates to a PCB defect detection method based on image processing, which comprises the following steps: collecting defect-free PCB image data, and making a data set; loss functions are set for the reconstructed subnet and the segmented subnet respectively, the data set is used for training to obtain the trained reconstructed subnet and the segmented subnet, and the trained reconstructed subnet adopts a diffusion model to carry out noise removal on to-be-detected image data subjected to preprocessing and data enhancement step by step according to time steps to obtain a reconstructed image. And decomposing the reconstructed image into a low-frequency component and a high-frequency component through discrete wavelet transform, inputting the low-frequency component into the trained segmentation subnet, comparing the low-frequency component with the to-be-detected image data, and distinguishing a defect part. According to the method, the number of used samples is far smaller than that of other target detection model samples, discrete wavelet transform is utilized, high-frequency components are extracted for back diffusion denoising, the reasoning speed is increased, and the segmentation and defect detection precision is improved through the extracted low-frequency components.

The invention discloses a cable defect detection method of a binocular intelligent inspection robot, and relates to the field of image processing, and the method comprises the steps: firstly obtaining the state information of a cable, then carrying out the preprocessing and stereo matching of the image information of the cable, and then constructing a target detection and segmentation model; the target detection and segmentation model is adopted to detect and segment cable defects, and model sample data is updated in real time in an incremental learning mode; fusing the cable state information acquired by the infrared camera, the thermal imager and the wireless sensor network by using a multi-sensor data information fusion model; the cable is maintained and managed based on a detection result; according to the invention, the detection precision and accuracy can be improved, cables of different types and specifications can be handled, the adaptability and stability are ensured, and the real-time performance and accuracy of detection can be improved.

The invention relates to the technical field of data management, in particular to a pearlescent material production data management system and method based on data integration analysis. The method comprises the following steps: obtaining a process original data set of a pearlescent material production line, and carrying out standardization processing to obtain a standardized process data set; obtaining a quality detection data set of the pearlescent material; performing time interval division on the standardized process data set and the quality detection data set, and performing batch number marking to obtain a process-quality batch data pair; and based on the process-quality batch data pair, designing a modeling sample data set, and constructing a quality process coupling model. According to the invention, the intelligent, data-driven and high-quality stable control of the pearlescent material production process is comprehensively realized by constructing a closed-loop system integrating data integration, modeling prediction and regulation and control optimization.

The invention discloses a test method for simulating rock hydraulic fracturing in a true triaxial stress state, and relates to the technical field of hydraulic fracturing, the test method comprises the following steps: S1, a sample preparation stage: processing a rock block into a standard square rock sample, and processing a plurality of hydraulic fracturing holes according to a test target size; s2, a sample mounting stage; s3, an equipment inspection stage; s4, a pre-load applying stage; s5, a hydraulic pressure applying stage of the hydraulic fracturing model: injecting hydraulic fracturing fluid with different flow rates into the hydraulic fracturing holes at different flow rates by utilizing a constant-flow constant-pressure pump through the hydraulic fracturing hydraulic pressure applying pipeline, and recording data and closing the constant-flow constant-pressure pump after the square rock sample is damaged; s6, an unloading stage; and S7, ending the test. According to the invention, a single hydraulic fracturing hole or a plurality of hydraulic fracturing holes are processed on a square rock sample to simulate single well and multi-well model samples, and different hydraulic pressurization tests are realized, so that the qualitative analysis of the reservoir rock crack evolution law under the true triaxial stress condition is completed.

The invention relates to a continuous flow chemical reaction yield soft measurement method and a continuous flow chemical reaction yield soft measurement system. According to the method, real-time online prediction of the continuous flow chemical reaction product yield is realized through a deep learning technology. Firstly, soft measurement modeling parameters are selected, and after data acquisition, a soft measurement modeling sample set is created through data alignment processing and dynamic time window feature construction. The soft measurement model is based on a multi-head attention mechanism enhanced LSTM, a multi-scale feature extraction and frequency domain analysis mechanism is fused, and local dynamic features under different time resolutions are extracted through a multi-scale convolution kernel; key time steps in a long sequence and a global dependency relationship of the key time steps are effectively captured through a multi-head attention mechanism enhanced LSTM module; and the periodic fluctuation characteristics in the time series data are extracted and identified through the frequency domain characteristics. The invention provides an economical, efficient, real-time and reliable intelligent solution for yield monitoring in the continuous flow chemical reaction process.

The invention discloses an image editing method and device and a storage medium. The method comprises the steps that diffusion processing is conducted on an original image through a diffusion model; determining the total number of noise reduction iterations by using a first model trained in advance; and in each noise reduction processing process, sampling a network module of the noise prediction network in the diffusion model by using a pre-trained second model, and carrying out noise reduction processing by using the network module selected by sampling to obtain an edited target image. By applying the scheme of the embodiment of the invention, the first model determines the total number of iterations of noise reduction, and the second model samples the network module of the noise prediction network, so that the proper number of iterations and the proper network module can be adaptively selected for different images, and the noise reduction efficiency is improved on the basis of ensuring the image quality. The complexity of a long iteration process and a network structure is greatly avoided, and the generation efficiency of image editing is effectively improved.

The present disclosure provides a method and an apparatus for evaluating the quality of model samples, a storage medium and a computer device. The method includes: inputting sample data into an AI-generated content detection model to obtain a hit probability of the sample data; matching a content evaluation system based on the attribute information of the sample data; processing the sample data based on the evaluation rule in the content evaluation system to determine a test value of the sample data relative to at least one preset evaluation index; and performing a weighted calculation on the hit probability and the test value based on a target weight corresponding to the hit probability and the preset evaluation criterion, to obtain a quality score of the sample data. This method is capable of filtering data that may mislead model training, and also realizing a high-precision evaluation of the training data.

The present invention relates to the technical field of industrial device detection, and discloses a device damage intelligent monitoring method and system based on mechanism and working condition big data. The monitoring method comprises: first, setting various damage modes of a device, constructing a mechanism model sample of each damage mode on the basis of a technical standard and an engineering case, and constructing a working condition big data sample of each damage mode on the basis of field data; then, using the mechanism model sample and the working condition big data sample as data sets for model training so as to respectively construct a mechanism model and a working condition model; and finally, inputting real-time text data to be predicted and real-time working condition data to be predicted into the mechanism model and the working condition model, respectively, calculating the possibility that the real-time text data belongs to various damage modes and the possibility that the real-time working condition data belongs to various damage modes, and generating a device damage monitoring result on the basis of the two possibilities. The present invention effectively improves the utilization rate of real-time information during damage monitoring of a pressure-bearing device, so that the monitoring result is more accurate.

The invention relates to the technical field of artificial intelligence and hydraulic engineering, in particular to a gate dam safety monitoring multi-element collaborative prediction and early warning method based on deep learning. The method comprises the following steps: synchronously collecting monitoring data such as displacement, osmotic pressure and temperature of a gate dam through an automatic safety monitoring facility, carrying out collaborative correction processing on the monitoring data, and converting the monitoring data into a standardized sequence sample; intercepting the standardized sequence sample into a source sequence and a target sequence, constructing a model sample library, and dividing the model sample library into a training set and a verification set; training a pre-constructed gate dam safety monitoring multi-element collaborative prediction and early warning model by using the training set, evaluating the precision of the model by using the verification set, and automatically calibrating model parameters; and utilizing the determined model to execute prediction so as to prompt early warning. According to the method, the problems of lack of fusion of global association and local dependence, short model prediction period and the like of traditional safety monitoring single-station modeling can be solved, the limitation of single-element threshold alarm is broken through, and the multi-element cooperative monitoring and early warning precision and efficiency are improved.

The invention discloses a medical image clustering method based on correlation entropy and an adaptive bipartite graph, and the method comprises the steps: collecting image data of medical imaging equipment, and forming a sample data matrix; selecting anchor point samples from the sample data matrix, and establishing an original sample low-dimensional representation-anchor point low-dimensional representation graph for capturing a local geometric structure of an embedded space; introducing an orthogonal basis matrix; carrying out joint modeling on the sample reconstruction error, the structure retentivity, the adaptive graph constraint and the discriminative features by adopting a joint objective function; iteratively updating the joint objective function; and inputting the finally obtained low-dimensional representation of the original sample as an embedded representation into a clustering device of K-means, and completing final clustering label distribution. According to the method, more stable, efficient and interpretable unsupervised clustering is realized, the clustering precision and generalization ability are improved, and the method is particularly suitable for medical image data analysis tasks of high-dimensional and complex structures.

The invention discloses an activity identification method based on conditional antagonism data enhancement, which comprises three parts of a conditional generative adversarial network, a kinematic feature constraint module and a prototype sample generator, a system hardware structure is designed based on an embedded heterogeneous computing platform, a main processor adopts an ARM Cortex-A72 core operation condition generative adversarial network, and the model sample generator is designed based on an embedded heterogeneous computing platform. The coprocessor adopts NPU to accelerate kinematics parameter calculation, and the sensor interface module is connected with a six-axis IMU sensor through an SPI bus to collect data of an accelerometer and a gyroscope in real time. According to the activity recognition method based on conditional antagonism data enhancement, enhancement data with diversity, authenticity and semantic accuracy can be provided for a human body activity recognition model in a scene of insufficient training data, so that the overall recognition accuracy of the model on a UCI HAR data set is improved by 14.7%, the F1 value on the aspect of few sample categories is averagely improved by 21.3%, and the recognition accuracy of the model on the UCI HAR data set is greatly improved. And an effective technical solution is provided for data scarce scenes such as medical monitoring and motion analysis.

The invention relates to the technical field of data storage, and discloses a storage management method and system based on data encryption, and the method comprises the steps: obtaining a model sample set, building a feature thermodynamic diagram based on the model sample set, determining a key feature cluster based on the feature thermodynamic diagram, carrying out the decoupling of the key feature cluster according to an auto-encoder, determining a model data set, and storing the model data set in a database. To-be-stored classification data are determined based on a multi-head self-attention model, each piece of to-be-stored classification data is layered based on analytic hierarchy process, a judgment matrix is constructed, and corresponding encryption strategies are adopted for all the to-be-stored classification data based on a data comprehensive score to determine encrypted data. The temporarily stored encrypted data is moved to the target sub-memory according to the data verification result, the backup mode of the target sub-memory is determined based on the storage occupancy rate, whether the backup mode is adjusted or not is judged based on the data retrieval condition, and the stability and reliability of data storage management are ensured.

Embodiments of the present disclosure relate to a general image prior based on Langevin diffusion. An original representation (image, 3D model, audio) is optimized to resemble a data distribution learned by a trained diffusion model. The original representation may be incomplete and is completed by the optimization process. In an embodiment, the diffusion model may be trained to use an additional conditioning input, such as a text prompt. The diffusion model receives a noisy latent as input and generates a denoised output, such as an image. Generally, the diffusion model samples the learned data distribution to produce the output. The conditioning input provides additional constraint that causes the output to be “nudged” towards the learned data distribution. An example synthesis problem is to generate a panorama image that is much larger compared with images used to train the diffusion model.

The invention relates to the technical field of livestock measurement, in particular to a beef cattle body size intelligent measurement device and method, in the beef cattle body size intelligent measurement device and method, hoof joint point pixel coordinates are converted and angle differences are calculated in combination with a lexicographical sequence block reordering algorithm, identity binding codes are accurately generated, uniqueness and stability of beef cattle individual recognition are ensured, and the accuracy of beef cattle body size measurement is improved. The method effectively reduces the identification confusion risk when multiple beef cattle are measured at the same time, generates a simulation sample through local pixel block rendering and generative adversarial network, screens a sample according with a threshold according to authenticity scores, improves the sample richness, optimizes the coverage of a model training set, extracts body size fields and position coding values, and calculates correlation coefficients between fields. Low-correlation field pairs are removed, only high-correlation fields are reserved for cross combination, a modeling sample pool is constructed, the representativeness of samples and variable combination diversity are enhanced, and the fitting precision and popularization capability of a body size estimation model are improved.

The invention provides a laying hen feed raw material sample screening method and system based on multi-source variability, and relates to the technical field of data processing analysis, and the method comprises the steps: obtaining multi-source attribute data and conventional component content data of raw materials, mapping the attribute data into tensor modal dimensions through multi-source variability tensor construction processing, and obtaining a multi-source variation tensor model; the component data is used as a characteristic component, and a multi-source variability tensor is obtained through decoupling and compression. Variability spectrum decomposition processing is carried out, local rank spectrum decomposition is carried out along producing areas, time and component dimensions, and a variability spectrum vector set is obtained; and identifying a candidate modeling sample set through multi-scale extremum and sparsity screening. And evaluating the contribution degree and sensitivity of the sample to a standard ileum amino acid digestibility prediction equation through a leave-one-out method and sensitivity analysis, and screening out an optimal modeling sample. According to the method, the multi-source variation information of the raw materials can be integrated, the variation spectrum is comprehensively covered with the minimum sample size, and the precision and generalization ability of the prediction model are remarkably improved.

The embodiment of the invention provides a user interface defect image generation method, a model training method, a user interface defect detection method and device, electronic equipment, a storage medium and a program product, and relates to the technical field of image processing. The method comprises the steps that segmentation information, element information and copywriting information of a screenshot of a user interface are obtained, the segmentation information is used for indicating a segmentation area corresponding to each pixel in the screenshot, different segmentation areas represent different control elements, and the element information indicates a reference area, in the screenshot, of each control element in the user interface; the copywriting information indicates a copywriting area of each copywriting in the screenshot; and respectively processing the screenshots through the construction mode of the at least one user interface defect image and the target information required by the corresponding construction mode to obtain the user interface defect image of the at least one defect type. User interface defect images of various defect types can be generated without limiting the number, and the problem that the number of model samples is small is solved.

The present invention discloses a method and device for restoring three-dimensional posture information. In the above method, a first frontal image of a current human body facing a camera device is obtained; multiple joints in the first frontal image are identified, and the first length information of the rigid bones corresponding to each of the multiple joints in the first frontal image is determined; a second frontal image of the current human body facing the camera device and photographed after the human body moves and changes its posture is obtained; multiple joints in the second frontal image are identified, and the second length information corresponding to the rigid bones corresponding to each of the multiple joints after the movement is determined; and the posture information of the current human body in the third dimension orthogonal to the imaging plane is restored according to the projection coordinate position information and the second length information of the rigid bones corresponding to each of the multiple joints on the image imaging plane. The acquisition cost of model samples and the computing power requirements for training models are reduced, and the amount of calculation is greatly reduced.

The invention discloses a density attribute earthquake prediction method. The method comprises the following steps: forward modeling a longitudinal wave velocity, a transverse wave velocity and a density curve of saturated fluid rock; continuing forward modeling to generate a large number of longitudinal wave velocity, transverse wave velocity and density curves of different pseudo wells by using the bulk modulus and density parameters of the rock skeleton in the research area, and generating and synthesizing a pre-stack seismic gather under the guidance of a Zoeppritz equation; convolutional neural network training is carried out by using the longitudinal wave velocities, the transverse wave velocities and the density curves of the different pseudo wells and the synthetic pre-stack seismic gathers, and a trained convolutional neural network model is obtained; and applying the convolutional neural network model to an actual well and trace gather, correcting the trained convolutional neural network model, and performing density attribute prediction by using the corrected convolutional neural network model and applying the actual trace gather. Density attribute prediction is carried out through a deep learning method based on rock physical model sample expansion, and the problem that in the prior art, conventional elastic parameters cannot identify effective reservoirs under the medium-deep layer condition is solved.

The embodiment of the invention relates to a model sample generation method and device, equipment and a storage medium. The method comprises the steps that based on input keywords, a plurality of question and answer pairs in a natural language are sequentially generated through a first machine learning model, and a second question and answer pair in the multiple question and answer pairs is generated at least based on replies in a first question and answer pair generated before the second question and answer pair; integrating the plurality of question and answer pairs by using a second machine learning model, and generating a multi-hop question aiming at the question and answer model and a reference answer aiming at the multi-hop question; and in response to determining that the multi-hop question and the reference reply meet the first quality requirement, generating a test sample and / or a training sample of the question and answer model.

The application provides a training method and device of an intent recognition model, and an intent recognition method and device. The method comprises: obtaining a training sample set; a sample in the training sample set comprises a question sentence labeled with an intent label; a pre-training model is trained using the sample in the training sample set; an initial output vector corresponding to the sample is obtained; an initial back propagation gradient of the pre-training model is determined according to the initial output vector; a preset number of disturbances are added to the initial output vector based on the initial back propagation gradient to obtain a target back propagation gradient; and a model parameter of the pre-training model is updated according to the target back propagation gradient to obtain an intent recognition model. The application uses the gradient of the back propagation in the pre-training model to perform adversarial disturbance on the output vector corresponding to the model sample, i.e. the word embedding layer vector. This adversarial training method can improve the robustness of the model.

The invention discloses a numerical control system-oriented defect visual classification detection method and system, and the method comprises the steps: obtaining an original defect sample covering a numerical control system part, carrying out the multi-modal image collection, carrying out the weighted fusion of visible light, near-infrared light and ultraviolet light images in an original sample library through a multispectral fusion algorithm based on the original defect sample, and carrying out the detection of the defect visual classification. The method comprises the following steps: generating an original defect sample, generating an enhanced sample, acquiring three-dimensional geometric information of the surface of a part by utilizing a 3D laser scanning device, transmitting and receiving ultrasonic waves through an ultrasonic detection module, detecting defects in the part, generating a virtual sample based on the original defect sample, the enhanced sample and a 3D model sample, and combining the original defect sample, the enhanced sample, the 3D model sample and the virtual sample to obtain a three-dimensional image. And training a defect classification model based on the mixed sample library. Defect evolution under different working conditions can be simulated, virtual defect samples conforming to real physical laws are generated, the defect state coverage range is wide, and defect data which are difficult to obtain by real samples are supplemented.

The invention provides an agent training method and device, electronic equipment and a storage medium. The method comprises the steps that a plurality of opponent learning models are generated according to pre-labeled sample data, the sample data comprise game data of a plurality of roles at different grades, each opponent learning model comprises at least one opponent strategy, and each opponent strategy corresponds to one grade; a to-be-learned course is generated according to the grade corresponding to each opponent strategy, the to-be-learned course comprises a plurality of course sets with the grades arranged in sequence, and each course set comprises a plurality of opponent strategies with the same grade; and training based on the to-be-learned course to obtain a target agent. According to the method, the intensities of different opponent strategies can be effectively distinguished, meanwhile, different roles of different grades can be considered, the diversity of the opponent strategies is guaranteed, the confrontation with human players is improved, and the actual game environment of the players is better met.

The invention relates to a soil organic carbon content estimation method based on multi-feature coupling ensemble learning. The method comprises the following steps: screening a soil hyperspectral image of a target area by using a spectral index; establishing an SOC content feature library, calculating feature importance by using an LGBM model, and selecting input features based on the feature importance to construct a training data set; an SOC estimation framework model is constructed, a first layer comprises a random forest model, a multi-layer perceptron model and a K neighbor model, and a second layer adopts CatBoost as a meta learning device; and inputting the training data set into a first layer of the SOC estimation framework model, respectively generating predicted values of the SOC content, performing five-fold cross validation on each model sample to obtain a group of predicted superposed values, and training a meta-learner of a second layer to obtain a final SOC content prediction result. Compared with the prior art, the method has the advantages of high applicability, high prediction precision, excellent interpretability and the like under different space-time conditions.

The invention relates to the field of reliability optimization design of isolation systems, in particular to an acquisition method of an isolation system agent model, an optimization method of impact resistance reliability of an isolation system and an acquisition device of the isolation system agent model. The acquisition method of the isolation system agent model comprises the following steps: establishing an equivalent model of an isolation element in an impact environment; generating a low-fidelity finite element model sample set of the isolation system according to the equivalent model of the isolation element; obtaining a high-fidelity finite element model sample set of the isolation system; and constructing a multi-fidelity agent model of the isolation system according to the low-fidelity finite element model sample set and the high-fidelity finite element model sample set of the isolation system. According to the isolation system anti-agent model obtaining method provided by the invention, the calculation cost required for establishing the agent model can be reduced, so that the time required for reliability optimization design of the isolation system is shortened.

The invention discloses an uncertainty guidance collaborative learning method for text-to-image pedestrian re-identification, and relates to the technical field of text-to-image pedestrian re-identification. The invention provides an accurate corresponding relation identification module, which combines global features and local features to realize more accurate data set classification, effectively reduces the risk of wrong supervision, reduces mismatching and misidentification, realizes more accurate data classification through a multi-level sample division strategy, and improves the accuracy of data classification. According to the method, the error supervision risk is effectively reduced, mismatching and misrecognition are reduced, an uncertainty guiding alignment module is provided, modeling is conducted on sample uncertainty, uncertainty is estimated, positive sample pair contribution is enhanced, negative influences of mismatching pairs are restrained, a processing mechanism of unreliable matching pairs is optimized by estimating the sample uncertainty, and the mismatching and misrecognition accuracy is improved. And the negative influence of the mismatching pair is inhibited while the contribution of the positive sample pair is enhanced.

The invention relates to a drug effect evaluation method for fusing target molecules and time-concentration dependent cell phenotypes, which comprises the following steps: S1, preparing samples including a modeling sample and a to-be-detected sample; s2, carrying out FRET imaging and cell fluorescence imaging; s3, carrying out FRET image processing and FRET efficiency calculation; s4, cell fluorescence image processing and feature extraction; s5, calculating a phenotype characterization value; s6, calculating an FRET characterization value; s7, basic drug response value calculation; and S8, comprehensive drug effect evaluation. Target molecule information and time-concentration dependent cell phenotypic response are fused, a comprehensive drug effect evaluation model is constructed, the effect of drugs on whole cells is reflected, and whether the drugs target specified molecules or not is specifically indicated.

The invention discloses a network training method, a modal prediction method, related equipment and a medium, the network training method is used for training a geometric deep learning network, and the network training method comprises the following steps: obtaining first sample data of a target component, the first sample data comprises a sample finite element model of the target part, a sample non-visual parameter and a corresponding sample modal real result; performing prediction based on the sample finite element model and the sample non-visual parameters by using a geometric deep learning network to obtain a sample modal prediction result of the target component; and adjusting network parameters of the geometric deep learning network by using the difference between the sample modal prediction result and the sample modal real result. By training the geometric deep learning network in this way, the accuracy of predicting the modality of the target component by the geometric deep learning network can be improved.

According to the pipeline environment event recognition method and system based on the audio, when a pipeline environment event occurs, pipeline environment sound is collected, and an event audio set is generated; generating a classification label of a surrounding event type for the pipeline environment sound in the event audio set, and performing feature extraction on the pipeline environment sound by using a Mel function and a Welch spectrum analysis algorithm to obtain feature data; according to the feature data and the classification labels corresponding to the feature data, training an ECAPA-TDNN model, establishing a model sample library, and performing identification and early warning of pipeline environment events by using the model sample library; the method is higher in maintainability and stability, and is better in anti-noise capability and higher in recognition precision.