52results about How to "Increase sample size" patented technology

The invention relates to the technical field of machine learning, in particular to a supervised time sequence water level data generation method and system and a storage medium. The method comprises the following steps: acquiring multi-source hydrological monitoring data; performing abnormal value identification and data preprocessing according to the hydrological monitoring data to generate standardized hydrological basic data; constructing a data generation model of the conditional generative adversarial network based on the standardized hydrological basic data; hydrological variable driven condition feature regulation and control are carried out on the data generation model, parameter optimization of the generation network is carried out, and hydrological adaptability enhanced condition input feature data are generated; and injecting the condition input characteristic data into the data generation model to generate initial water level time sequence data. According to the method, the rare extreme hydrological events are oversampled by using a hierarchical importance sampling and self-service method (Bootstrap), so that the performance of the generative model in rare event prediction is remarkably enhanced.

The invention relates to the technical field of artificial intelligence, and particularly provides a large language model training method, system and device and a storage medium, and the method comprises the steps: collecting task samples based on task difficulty and task diversity demands; obtaining a support set and a query set of the task samples, and performing data enhancement processing on the support set; performing internal loop training on the meta-model by using the support set, and performing external loop training on the meta-model subjected to the internal loop training by using the query set to obtain a meta-gradient; and summarizing the element gradients, and updating the global model based on the element gradient summarized data by using an adaptive optimizer. The adaptive capacity of the large language model to the task type is improved.

The invention relates to a power distribution network grounding fault line selection method based on an image level, and the method comprises the following steps: obtaining historical waveform data of each line through power grid data, building a power distribution network fault feature system, and carrying out the training to obtain a grounding fault line selection model; acquiring real-time waveform data, uploading the real-time waveform data to the grounding fault line selection model, predicting a line state, and stopping power supply and positioning a line fault position when a line fault occurs; fault waveform data during line fault classification are analyzed by using a deep learning method, and the fault type of the line is obtained; and the line fault position and the fault type are sent to a power grid system, and the power grid system makes a fault decision for the line. The fault state of each line in the power distribution network system can be monitored in real time through the grounding fault line selection model, then the fault type of the line is obtained by using the deep learning method and the random forest classifier, and finally the purpose of making a fault decision on the line rapidly, fully automatically and accurately is achieved.

The invention relates to the field of traffic monitoring, in particular to a multi-millimeter-wave radar vehicle track splicing method and system in a tunnel. The method comprises the following steps: carrying out space-time calibration on all millimeter wave radars in a tunnel; respectively acquiring a set of vehicle information of all vehicles at any same moment in the overlapping area of all adjacent millimeter wave radars; similarity of adjacent radar vehicle information; and inputting the similarity into a pre-trained BP network, and determining whether to splice the vehicle tracks or not according to a result. According to the method, the vehicle tracks between the millimeter wave radars are spliced by adopting a BP neural network method, and the problem that the splicing effect of the vehicle tracks in the tunnel is not ideal due to the fact that millimeter wave radar signals have a multipath effect in the tunnel, detection interference is large and monitoring data precision is insufficient is solved.

The invention belongs to the field of semiconductor life evaluation, relates to a data analysis technology, and is used for solving the problems that in the prior art, feature analysis cannot be performed according to life test results of different types of semiconductor devices, and links such as sample extraction and key test parameter optimization cannot be performed by combining feature analysis results. The intelligent service life evaluation system adapted to the power semiconductor device design comprises a sampling analysis module, a balance evaluation module, a tendency evaluation module and an evaluation analysis module, the sampling analysis module is in communication connection with the balance evaluation module and the tendency evaluation module, and the evaluation analysis module is in communication connection with the balance evaluation module and the tendency evaluation module; according to the method, evaluation samples of the power semiconductor device can be extracted, and evaluation modes of the power semiconductor device can be marked according to evaluation test results of different device types in an evaluation period, so that an acceleration model can be reasonably distributed in a next evaluation period.

The invention provides a cross-enterprise product data standardization classification method and a management software system thereof, and belongs to the technical field of product data management, standardization management, artificial intelligence and natural language process.The classification method comprises the steps that a product database and a front-end interface are designed; inputting product data, and verifying and checking the product data; determining a data set for training, and constructing a text classification model of the long-short term memory network; classifying the product data based on a trained long-short term memory network text classification model; the management software system comprises a project information module, a contract information module, a data filling module and a standard downloading module. According to the method, the problem that classification texts processed by a traditional classification algorithm are mostly discrete and are poor in mutual relevance is considered, the global memory result of each layer is supplemented by means of a long-short-term memory network architecture, a training data set is expanded in combination with random recombination, product data management services with high universality can be provided, and the product data quality is guaranteed.

The invention discloses a method for predicting soil ammonium nitrogen pollution space in a rare earth mining area. Based on a spatial layering heterogeneity theory and a Bayesian theory, a spatial interpolation technology is combined, and an innovative layering Bayesian three-dimensional curved surface mean value model is provided. According to the model, layered modeling is introduced for the situation that pollutants in the rare earth mining area have high heterogeneity in the vertical direction, that is, the pollutants are subjected to layered treatment according to different depths, non-homogeneous distribution of the pollutants in different depth layers is captured, and the distribution difference of the pollutants in different stratums can be reflected more accurately.

The invention relates to a clinical index-based biological age evaluation system, a construction method and application, and belongs to the field of biological aging research. Firstly, clinical indexes of a subject are collected, a biological age estimation model is constructed based on the clinical indexes, and the biological aging speed is further calculated. The method can be used for identifying an individual which accelerates senescence or delays senescence, and risk prediction and crowd risk stratification of death and senescence related diseases are realized. According to the method, the nonlinear relationship between clinical indexes and the aging degree is considered, so that the biological age estimated by the method has a better prediction effect on the risk of death and new aging related diseases. The biological age and the biological aging speed are quantified based on the clinical indexes, measurement of the clinical indexes can be completed only through conventional physical examination and one-time blood sampling, the method has the advantages of being low in cost and convenient to use, the biological aging speed can be evaluated in time after physical examination, and health guidance opinions are provided for aging acceleration individuals.

The embodiments of the present application provide a model training, sample generation method, electronic device and storage medium. In the model training method provided in the embodiments of the present application, a large number of training samples can be obtained based on image synthesis processing of a small number of target image samples and background image samples. In this embodiment, different target image samples can be synthesized with different background image samples, which increases the flexibility of obtaining samples, effectively expands the sample capacity of training samples, and is conducive to improving the generalization performance of the image segmentation model.

The application provides a child skull fracture automatic detection method and device based on generated data augmentation. The method comprises the following steps: acquiring a first child brain CT image set, and performing pixel labeling and preprocessing to obtain fracture data, suture data and normal data; obtaining a trained fracture lesion generation model and a suture generation model according to the fracture data, the suture data and the normal data; randomly sampling Gaussian noise for multiple times, inputting the Gaussian noise into the fracture lesion generation model and the suture generation model to obtain generated fracture CT patch data and generated suture CT patch data; training a candidate fracture detection model according to the fracture data, the suture data, the normal data, the generated fracture CT patch data and the generated suture CT patch data, and outputting a fracture detection model; and acquiring a second child brain CT image, and detecting the second child brain CT image according to the fracture detection model. The fracture lesion data and the suture data are augmented, so that the resolution of fractures and sutures can be improved.

The present application relates to the field of image recognition, and discloses a method for identifying cartridge case traces based on a neural network, the method comprising: obtaining a cartridge case bottom image, taking the groove edge of the cartridge case bottom as the boundary, determining a central region and an outer ring region in the cartridge case bottom image, and determining a central region feature corresponding to the central region and an outer ring feature corresponding to the outer ring region; matching multiple similar outer ring features in a database according to the outer ring feature, each similar outer ring feature corresponding to an outer ring of a similar cartridge case bottom image; obtaining a similar central region corresponding to the cartridge case primer region of each similar cartridge case bottom image, and determining a similar central region feature corresponding to each similar central region, and determining multiple target central regions according to the similarity between the central region feature and each similar central region feature. The present application can achieve fast and accurate matching of cartridge case traces when the image clarity is high and the sample size is large.

The application provides a frequency standard measurement method, device, computer equipment and storage medium, the method comprises: frequency source output measurement signal is carried out frequency division processing, obtains low frequency pulse signal; using high-precision time interval counter to low frequency pulse signal is carried out delay measurement and time interval counting, and is output in the form of time stamp; by setting a sliding window, using sliding average algorithm to time stamp is carried out sliding average filtering processing, all time stamps are carried out sliding difference to obtain time difference, time difference is grouped according to order and is taken average in group to obtain average time difference; by increasing the length of the sliding window, the number of samples is increased to reduce the statistical error; using average time difference, calculate frequency measurement value, relative frequency deviation and allan variance. The method of the application increases the sliding window, and the frequency measurement result obtained by the algorithm of grouping and averaging the time difference has higher measurement precision and stability.

The application relates to the technical field of deep learning, in particular to a traffic scene image small target detection method and device, equipment and a storage medium. The method comprises the following steps: acquiring a traffic scene image and a detection result of a small target in the traffic scene image, generating a training data set according to the traffic scene image and the detection result of the small target in the traffic scene image; constructing a convolutional neural network based on a feature extraction module, a feature fusion module, a sample distribution module and a positioning classification module, and optimizing the sample distribution module and the positioning classification module of the convolutional neural network to obtain an optimized convolutional neural network; training the optimized convolutional neural network by using the training data set, updating network parameters of the optimized convolutional neural network based on gradient back propagation, and stopping the training until a preset training stop condition is met; and detecting small targets in the traffic scene image by using the trained convolutional neural network. Thus, the problems of small target sample imbalance and poor positioning capability in the related art are solved.

The invention discloses a visual detection method for the welding quality of a lithium battery pole, and the method comprises the steps: S1, obtaining a lithium battery pole welding image and a pixel-level marking image corresponding to each welding image, and forming a sample set; step S2, carrying out preprocessing on the training sample set train; s3, building a visual detection model; s4, training the visual detection model, storing a training weight file of the visual detection model, and generating the visual detection model; s5, inputting a to-be-tested lithium battery pole welding image into the trained visual detection model to obtain a segmentation result image, and processing the segmentation result image to obtain a semantic segmentation region result image; and S6, performing post-processing on the semantic segmentation region result map, processing and segmenting internal mixing and undersize pixel filtering, and obtaining a final semantic segmentation type region result map. According to the method, the defects of low detection precision, insufficient generalization adaptive capacity, low detection speed and multiple required samples in the prior art are overcome.

The application discloses a co-entropy ceramic protective coating sintering temperature and aerodynamic thermal load cooperative control system and relates to the technical field of sintering temperature control. The system comprises an information acquisition module: acquiring previous sintering information, wherein the previous sintering information comprises sintering temperature and performance results, and the performance results are the performance of aerodynamic thermal load. The application considers coating material and structure information and environmental information to formulate a reference directory under different basic information in the whole process, and the sintering temperature is regulated according to the aerodynamic thermal load of the coating material structure and environmental manufacturing demand, and then the density, interface bonding and residual stress distribution of the coating are realized through the controllable sintering temperature curve at the preparation end, so that the aerodynamic thermal load of the coating in use is ensured, the sintering temperature is designed and optimized, the response characteristics of the coating to the aerodynamic thermal load in use are within the user demand, and the closed-loop design of optimizing the response of the service end by controlling the manufacturing end is embodied.

This invention discloses a training method and apparatus for a 3D meandering river recognition model, and a method for identifying meandering rivers. The method includes: acquiring several interpretation results and corresponding seismic data volumes in a target area; initializing a preset label data volume based on the interpretation results and the seismic data volume; masking the label data volume to update it; inputting each seismic data point in the seismic data volume as sample data into a pre-constructed image segmentation model for training; calculating a loss function value based on the output of the image segmentation model and the updated label data volume; adjusting the parameters of the image segmentation model inversely based on the loss function value; iteratively training until the loss function reaches a preset convergence condition, stopping the training, and obtaining a trained 3D meandering river recognition model. This model enables the identification of meandering rivers in a target area using sparse labels, improving the efficiency of seismic data interpretation.

The invention relates to the technical field of reliability evaluation and life prediction, in particular to a traveling wave tube life evaluation method and device fused with multi-source degradation data and a medium. In consideration of a large number of accelerated tests of a cathode carried out on the ground, test data accumulated in an accelerated degradation test process of applying current sensitive stress are sorted and analyzed, and a hierarchical Bayesian model is utilized, and a hierarchical Bayesian fusion framework is adopted to obtain a hierarchical Bayesian model; the performance degradation data of the cathode under different acceleration current conditions are fused, namely the performance degradation data under different test conditions are fused, the service life data and the sample capacity are expanded, the influence of the cathode emission current on the service life of the cathode at a certain working temperature is explored, and the service life of the cathode of the traveling wave tube is evaluated.

The application discloses a regional access identity permission authentication method based on a block chain, relates to the technical field of block chain identity authentication, and comprises the following steps: collecting real-time biological feature data and device fingerprint data of a user terminal, integrating the data to form a to-be-authenticated data packet, and submitting the data packet to a block chain smart contract; the smart contract calls a permission mapping table on a chain to match a corresponding on-chain identity identifier, starts a cross-chain verification protocol after identity verification, and simultaneously calls a plurality of off-chain identity sources to call a bound historical biological feature template; real-time biological data is compared with the plurality of historical templates to generate a comparison score matrix, a comprehensive trust score is calculated in combination with the device fingerprint data, and when the score meets an access threshold, an effective access token is issued to the user terminal. The method realizes multi-channel biological data collaborative verification, integrates device-specific information to complete qualification assessment, and optimizes the verification logic and judgment dimension of identity permission authentication in a regional access scene.

The invention discloses a hydroelectric overhaul tool sorting method based on a SqueezeNet convolutional neural network model. The method is suitable for various tools such as wrenches, screwdrivers and sleeves. The method comprises the following steps: firstly, acquiring a maintenance tool image in a backlight environment, and removing background interference through an image preprocessing technology to generate a clear target binary image; in order to solve the problem of insufficient samples, data enhancement means such as expansion, corrosion and rotation are adopted to expand a training set. And then, deep learning training is carried out by using a SqueezeNet model, parameters are adjusted in combination with a cross entropy loss function and a gradient descent optimization algorithm, and finally, a high-precision recognition model is obtained. In addition, hand-eye calibration of the visual system and the mechanical arm is completed through a nine-point calibration method, and accurate conversion from image coordinates to mechanical arm action coordinates is ensured. In actual operation, stacked tools are separated through the flexible vibration disc, then the visual system collects images of the single tools, the types and the positions of the single tools are recognized through the model, and the mechanical arm is guided to conduct accurate grabbing.

The utility model discloses a kind of equipment for seaweed quality detection, a kind of equipment for seaweed quality detection, including lye prestorage tank, seaweed processing tank body, lye recovery tank;Wherein, lye prestorage tank is communicated with seaweed processing tank body by lye pipeline;Seaweed processing tank is communicated with lye recovery tank.The beneficial effect of the utility model is: compared with conventional seaweed detection method, conventional seaweed detection method can only handle 1 batch of seaweed at a time, each batch of seaweed is detected separately, and artificial efficiency is low.New seaweed detection equipment and method can handle 16 batches of seaweed at a time, greatly improve artificial efficiency.

The invention belongs to the technical field of nuclear radiation detection, and particularly relates to an ESEM-oriented radon solid track image counting method and system, and the method and system are composed of a solid track detector, an environment scanning electron microscope, a high-resolution camera, an image analysis processing system, a fixed frame and a communication module, and achieve the automatic image preprocessing and image edge detection. The method comprises the steps of image binarization threshold segmentation, morphological processing, connected component analysis, feature extraction, body track automatic counting, result output and the like, automatic collection and processing of a solid track image of a radon environment are completed, a marked image is obtained, the marked image comprises the marked solid track position and the serial number of a connected component, the marked image can be directly used for observing a sample, and the detection accuracy is improved. The requirements on chemical fixed trace and vacuum observation environment are reduced, the influence of charging effect and high vacuum environment on the sample is avoided, and real-time detection can be carried out.

The invention discloses an initial low-altitude economic air traffic take-off and landing point layout method based on mobile communication big data. The method comprises the following steps: acquiring mobile phone signaling data within a time period selected by a mobile communication operator in a region; carrying out travel segmentation, selecting travel stop points according to a stop time threshold value of the user track data, obtaining OD point pairs, carrying out track segmentation according to the OD point pairs, obtaining travel travel mode identification, obtaining POI data of traffic in the target area, matching the POI data with a travel track, and identifying a travel mode; pOI data of a key area in the area are obtained, and travel related to the key area is judged; generating a low-altitude air traffic travel OD point library; and identifying air traffic take-off and landing point layout schemes under different volumes. According to the method, travel purpose division is realized, demand identification can be carried out on position points, and the method plays an important role in low-altitude economic air traffic take-off and landing point site selection planning.

The present invention discloses a method for calculating the phase angle of variable valve timing, comprising the following steps: judging whether the camshaft is interrupted; judging whether it is the second tooth of the camshaft; calculating the edge phase angle of the second tooth of the camshaft; judging whether the crankshaft is interrupted; crankshaft counting; judging whether the Hall is interrupted; Hall step counting; and calculating the step counting phase angle according to the characteristics of the reducer. The above technical solution not only greatly increases the number of samples and improves the calculation accuracy of the variable valve timing phase angle, but also has good real-time performance, is conducive to the realization of the phase closed-loop algorithm, and is more conducive to the stability of the phase.

The application discloses a kind of data-driven model establishing method and device, the method includes: according to the relevant parameter data of multi-energy park, the minimum cost objective function of multi-energy park and corresponding constraint condition are established;According to the deep reinforcement learning algorithm, the minimum cost objective function of multi-energy park is solved, and running state data are generated;According to the mapping relationship between the running state data and load data after the establishment of Gaussian process regression algorithm, the running state data is converted into new load data according to the mapping relationship;According to new load data, update load database, and establish the data-driven model of multi-energy park according to load database. The precision of the data-driven model of the multi-energy complementary park generated by the embodiment of the application can be effectively improved.

This invention discloses a well-seismic combined seismic data overlay method, relating to the field of geophysical exploration processing and interpretation technology in oil and gas field exploration. The method includes the following steps: S1, acquiring acoustic and density logging data, calculating the wave impedance curve and converting it to reflection coefficients, selecting a wavelet close to the seismic data, convolving it with the reflection coefficients to form a conventional synthetic record, and performing well-seismic calibration; S2, determining the target wavelet for overlay, and convolving it with the reflection coefficients to form a broadband synthetic record; S3, combining well-side seismic data, conventional synthetic records, and broadband synthetic records to construct a sample set, and balancing the number of samples; S4, expanding the sample set through mutation; S5, establishing a nonlinear relationship for overlay based on the sample set; S6, applying the nonlinear relationship to conventional seismic data to obtain high-resolution seismic data. This invention utilizes the advantages of wide bandwidth of well logging data and seismic frequency bands to enhance high-frequency seismic data and broaden the seismic frequency band, thereby improving seismic resolution.

The invention provides a method for detecting pit defects of a polished wafer of a silicon carbide substrate. The method comprises the steps of feeding, mechanical conveying, edge searching, OCR recognition, conveying to a detection table, first-time reflection light source detection and transmission light source detection, second-time reflection light source detection and data analysis. The first light source reflection detection comprises the following steps: irradiating the upper surface of the silicon carbide substrate polished wafer by using a reflection light source, and collecting a reflection light signal and a scattering light signal; the transmission light source detection comprises the following steps of: enabling transmission light to enter from the lower surface of the polished surface of the silicon carbide substrate through a polarizer and exit from the upper surface of the polished surface through a polarization analyzer, and collecting a transmission light signal; and second-time reflected light source detection: shielding the lower surface of the silicon carbide substrate polished wafer so that the light cannot penetrate through the lower surface, then irradiating the upper surface of the silicon carbide substrate polished wafer by using the reflected light source, and collecting a reflected light signal and a scattered light signal. The detection method provided by the invention can improve the precision, efficiency and reliability of pit defect detection.

The utility model discloses a tool for analyzing quality characteristics of a metal spraying electrode, and relates to the technical field of capacitor detection tools, the tool for analyzing the quality characteristics of the metal spraying electrode comprises a rotating disc mechanism, the rotating disc mechanism comprises a rotating disc and a driving motor, and the driving motor drives the rotating disc to rotate; the number of the fixing seats is three, the fixing seats are circumferentially and uniformly distributed at the top of the rotating disc, and a plurality of fixing columns are arranged on the fixing seats in the radial direction at intervals; the electrode modules are designed to simulate the shape of a core, the multiple electrode modules are arranged on the multiple fixing columns respectively, and a material groove is formed in the top of each electrode module; and the spray gun is located at the top of the rotary table mechanism and sprays back and forth in the radial direction of the rotary table. According to the tool for analyzing the quality characteristics of the metal spraying electrode, the sample amount is increased, the sampling coverage area is increased, and the uniformity of sample blocks is improved.

The invention provides a code retrieval data generation method and device and a code retrieval method and device, and relates to the technical field of computers.The method comprises the steps that basic code data and a plurality of preset retrieval scenes are obtained; respectively combining the basic code data with each retrieval scene to generate a plurality of pieces of first code retrieval data; and rewriting the first code retrieval data in at least one mode of replacing the basic code data, replacing the retrieval scene and replacing the language type to generate second code retrieval data. According to the embodiment, combination and rewriting are carried out on the basis of the preset retrieval scene and the basic code data, and the high-quality code retrieval data capable of being used for model training is obtained. Specifically, code rewriting is carried out on the basis of various modes of basic code data, retrieval scenes and language types, code retrieval data can be effectively expanded, and the number of samples of the code retrieval data is greatly increased.

The invention relates to the field of explosive performance evaluation equipment, and discloses a thermal explosive heat resistance evaluation device, system and evaluation method. The thermal explosive heat resistance evaluation device comprises a constant-temperature drying box, the interior of the constant-temperature drying box is provided with a containing cavity, and the constant-temperature drying box is used for responding to a received environmental parameter adjusting instruction, adjusting environmental parameters in the containing cavity and controlling the duration of the adjustment; wherein the environmental parameters comprise temperature and / or humidity; the heating medium bearing container is arranged in the constant-temperature drying box and is used for bearing a heating medium; the sample container is arranged in the heating medium bearing container, the heating medium is uniformly filled outside the sample container, and the sample container is used for bearing explosives and powders; and the temperature acquisition module is in contact with the sample container and is used for acquiring the temperature of the sample container. The actual use condition of the explosive can be simulated, and meanwhile, the test device is simple, convenient to construct and relatively high in efficiency.

This invention relates to the field of in vitro diagnostics, and particularly to a combined diagnostic kit for prostate diseases. The invention provides reagents including pH, peroxidase, elastase, lecithin bodies, citric acid, zinc, and acid phosphatase under mild reaction conditions, enabling rapid differentiation of inflammatory and non-inflammatory prostatitis, benign prostatic hyperplasia, and non-prostatic genital tract infections using a single test card. Through extensive experimentation, this invention has developed a one-step reaction scheme suitable for the dry chemical determination of acid phosphatase and lecithin bodies in seminal plasma. This method eliminates the need for a stop solution, achieving a one-step reaction under mild conditions that will not cause damage to operators or fully automated instruments. The samples used are seminal plasma or semen, allowing for large sample volumes and easy acquisition, significantly improving diagnostic efficiency for doctors and patient comfort.