49results about How to "Improve processing precision" patented technology

The invention discloses a homeland space planning data intelligent analysis method based on big data, and relates to the technical field of homeland space planning, and the method comprises the steps: collecting multi-source data, carrying out the quality screening, and forming a standardized data set. And setting a space element extraction rule based on a specific planning target, and mining key space elements from the standardized data. Constructing a big data analysis model, inputting key elements to calculate and obtain a spatial development tendency index, and generating a preliminary planning scheme according to the spatial development tendency index; and performing compliance verification on the scheme, generating evaluation data, and adjusting the scheme according to the evaluation data. And finally, performing space efficiency deduction calculation on the adjusted scheme, optimizing the scheme according to deduction result data, and outputting a final planning scheme. According to the method, the accuracy of data mining is improved through target-driven element extraction, and the perspectiveness and scientificity of a planning scheme are enhanced through dynamic efficiency simulation.

The invention discloses a scene perception information fusion decision-making method for industrial field multi-source heterogeneous data, which comprises an edge layer, a cloud layer and an interaction layer which form a three-level collaborative architecture, and belongs to the technical field of industrial field data processing. The method comprises the specific steps that S1, an edge layer collects industrial field multi-source heterogeneous original data and carries out self-adaptive preprocessing to obtain cleaned multi-source feature data; s2, the edge layer processes the multi-source feature data through a lightweight scene perception model, and outputs a preliminary scene judgment result; s3, the cloud layer constructs and updates a causal relationship graph based on industrial business knowledge and incremental data of the edge layer, processes data uploaded by the edge layer through a causal reasoning fusion model, outputs a decision result and issues the decision result through the interaction layer; and S4, the cloud layer optimizes parameters of the causal reasoning fusion model and the lightweight scene perception model based on a decision execution result fed back by the edge layer, and issues the optimized parameters to the edge layer through the interaction layer to realize dynamic iteration.

The invention discloses a laying hen lossless data processing method based on improved XGBoost, and relates to the technical field of data processing, discrete feature numeralization is realized through adaptive tag coding, abnormal value detection and non-linear mapping correction are combined, a robustness purification feature set is constructed, and a layer lossless data processing method based on the improved XGBoost is obtained. The problems that a traditional method is sensitive to sensor noisy points and lacks an effective data cleaning mechanism are solved, and prediction precision reduction caused by extreme sample interference is avoided. And performing nonlinear expansion on the purification features by using a polynomial kernel, and accurately capturing a coupling relationship between the features. The global optimal hyper-parameter optimization of the XGBoost integrated regression model is driven by high-dimensional features, so that the parameter optimization efficiency is improved, local optimum is avoided, the method adapts to a multi-source heterogeneous and high-noise nonlinear data scene of laying hen breeding, the processing precision and generalization ability of the model are remarkably improved, reliable data support is provided for accurate management and quality monitoring of laying hen breeding, and the method is suitable for large-scale popularization and application. And efficient landing of the intelligent breeding technology is promoted.

The application provides a kind of wind power equipment multi-factor coupling data processing method, device and equipment, it is related to wind power generation technical field.The method comprises: collecting the original data set of wind power equipment;Based on the preset differentiating stratified pretreatment strategy, the original data set is carried out noise reduction processing, and the standard data set is obtained.According to the original data set, the correlation matrix of multi-factor coupling is constructed;Correlation matrix represents the coupling correlation degree between each data in original data set.According to the preset fusion model, the attention weight of each data in the original data set is updated, and each data is weighted and fused based on the updated attention weight, and the comprehensive feature set is output.According to the comprehensive feature set, determine the preliminary analysis data;According to correlation matrix, the preliminary analysis data is corrected, and the target analysis data is generated.The method is used to improve the data processing accuracy under the effect of multi-factor coupling interference.

The application discloses a data clustering method and device based on density peaks, a medium, and direct subordination is used to describe the relative density of each data point, wherein the data point is a picture or a word feature in a data set, a mixed density value combining the relative density and the absolute density is designed to identify a clustering center, and the effective subordination is defined through similarity reachability, main distribution of non-center points is performed, and the label of a remaining point is determined in combination with the k-layer upper distribution of different clusters, so that the data points of the same category in the data set can be effectively gathered in one category, and the classification of the data set is efficiently and accurately completed.

The present application relates to the technical field of simulation sandbox, and particularly relates to a simulation sandbox data base construction method and device, wherein the method constructs compatible domain object models in the localization BIM forward design stage; according to unstructured drawings, localization BIM models and other data types, SHP data, geometric feature data and structured topology data are extracted or generated; simulation element data is obtained through field-level direct mapping, SHP data is converted twice and coordinates are unified, and then geometric feature completion, line segment merging and topology reconstruction are performed, object construction and labeling are completed based on the simulation element data, and simulation sandbox data base is obtained. The present application solves the problems of low construction efficiency, poor compatibility and insufficient geometric accuracy, realizes standardized data processing and efficient construction, can be directly loaded into a simulation sandbox, is suitable for airport and other scenes, and improves simulation accuracy and reliability.

The invention relates to a high-precision speed reducer gear key groove machining method. The method comprises the following steps that S100, heat treatment deformation is predicted; s200, groove broaching before heating; s300, heat treatment is conducted; s400, detecting after heating; s500, finish milling is conducted after heating; through a heat treatment deformation prediction model, after-heat deviation data are predicted in advance at the front end of the process, heat treatment parameters are optimized, so that heat deformation is controllable, a compensation reserved deformation compensation amount is provided for pre-heat groove broaching, a combined process is formed, the pre-heat groove broaching can be matched with a heat deformation prediction result to carry out deformation compensation, and the heat treatment quality is improved. A workpiece has a targeted compensation structure before entering heat treatment, and then the cutting amount is corrected through a cutting parameter optimization model in combination with pre-heat compensation and post-heat detection, so that the problem of thermal deformation uncontrollability in the key groove machining process is converted into predictable, compensable and correctable process variables; the technical bottlenecks that the symmetry degree is difficult to guarantee, the matching precision fluctuation is large and the machining efficiency is low in traditional gear key groove after-heating hard machining are broken through.

The utility model relates to the technical field of cylinder cover surface treatment, and disclose a kind of cylinder cover surface treatment frock, including base, the top of base is fixedly provided with two fixed plates, a fixed plate is fixedly provided with controller, another fixed plate is fixedly provided with observation glass window, the top of base is fixedly provided with positioning plate, the side of positioning plate is fixedly provided with first electric push rod, the side of first electric push rod is fixedly provided with clamping plate, the top of fixed plate is fixedly provided with processing frame, camera is fixedly provided on the processing frame, electric slide rail is fixedly provided in the width direction of processing frame.The cylinder cover surface treatment frock, by the cooperation of first electric push rod and clamping plate, can be accurately positioned and stably clamped according to different sizes of cylinder cover, avoid shaking or displacement in the processing process, ensure the uniformity and consistency of surface treatment effect, improve product quality.

The invention provides a body-side all-in-one machine chip module design system based on multi-scale feature embedding fusion, which belongs to the technical field of chip design and comprises a chip data acquisition module for acquiring electronic component information of a chip and determining proper electronic components and circuit topological structures according to requirements so as to realize specific functions required by the chip; determining two or more indexes of the chip as optimization targets; creating and updating an optimal file; establishing an optimal individual file, and updating an optimal individual in the current file based on a multi-model random sorting and Pareto domination mode; in each ball game, a chip is configured to perform big data analysis, so that the sports game score processing efficiency and precision can be effectively improved, the problems of low efficiency and high cost of a multi-chip integrated layout design in the prior art can be improved, and the method has the advantages of scheme optimization reliability, efficiency and cost consideration.

The invention relates to the technical field of data processing, in particular to an image material processing method and system. The method comprises the following steps: acquiring a to-be-processed image material, and performing semantic analysis on the image material by utilizing a preset visual language model to obtain a text prompt corresponding to each element; inputting the text prompt and the image material into a preset target detection model to obtain a detection result; for a hierarchy corresponding to each text prompt or bounding box, determining an interaction matching threshold value of the hierarchy; screening the bounding box according to a comparison result of the confidence of the bounding box and a corresponding interaction matching threshold to obtain a target bounding box; and inputting the target bounding box and the corresponding text prompt into a preset segmentation model, and generating a segmentation mask corresponding to each element. According to the method provided by the invention, automatic positioning, separation, filling and repairing of elements of different levels in the image material can be realized, and the efficiency and precision of image material processing are improved.

The application discloses a kind of medicine material winnowing machine intelligent control method and system, the method includes obtaining the mass distribution data of medicinal material, density characteristic value, profile data and texture feature, obtain medicinal material characteristic data set;According to characteristic data set, classification identification medicinal material, generate physical characteristic label, obtain classified medicinal material;According to classified medicinal material, query preset medicinal material winnowing parameter mapping database, obtain the equipment control parameter scheme required by winnowing;According to classified medicinal material, adjust fan speed, obtain the fan speed matched with medicinal material;According to density characteristic value, adjust air duct angle and airflow distribution plate, obtain actual air duct angle;Monitoring medicinal material separation effect, if it is detected that medicinal material separation trajectory deviates from preset range, then correct air duct angle and fan speed, update optimal air duct angle and fan speed to winnowing parameter mapping database, obtain final equipment control parameter scheme.The method can accurately identify medicinal material characteristics and dynamically match winnowing parameters.

The application discloses a kind of water area ecology monitoring method, device, equipment and storage medium.Firstly, the ecological hydrological data of target water area is collected by real-time remote sensing technology, including water table color, suspended solids concentration and surrounding vegetation coverage image, radiation calibration processing is carried out on vegetation coverage image to obtain radiation brightness value and apparent reflectivity, and regional growth algorithm is used to identify vegetation area image.Combined with apparent reflectivity, atmospheric correction is carried out to obtain water surface temperature.At the same time, the predicted water level height is obtained by inputting suspended solids concentration into water level prediction model.If real-time data (such as predicted water level height, real-time water table color, real-time suspended solids concentration, vegetation color or vegetation density) is inconsistent with target data, ecological early warning is triggered.The application automatically collects and processes real-time remote sensing data of target water area, efficiently extracts key ecological indicators using algorithms and integrates multi-dimensional data, improves monitoring efficiency and accuracy, and optimizes ecological early warning effect.

This invention relates to a pre-stack noise suppression method for seismic data, belonging to the field of seismic exploration technology. The method includes the following steps: 1) After decompiling and defining the observation system of the acquired seismic data, static correction, spherical diffusion compensation, and surface wave suppression are performed sequentially; 2) The first arrival time of the surface wave suppressed seismic data is corrected to a relatively uniform time, and then random noise and residual noise from the surface wave suppression process are suppressed; 3) The first arrival time of the noise-suppressed seismic data is reverse-corrected to restore the first arrival time to its state before correction. This method can effectively suppress shallow noise interference in near-offset data while better preserving the detailed structure of the seismic data, improving the accuracy of seismic processing; and avoids other noise interference introduced by general processing methods that apply strong noise suppression to achieve the purpose of suppressing shallow noise in near-offset data, thus improving the amplitude fidelity of the processing.

The application discloses a kind of medical intermediate production wastewater coupled processing equipment and method, it is related to wastewater treatment technical field.The equipment of the application includes wastewater tank, processing tank, carbon box and pressure assembly, carbon box is clamped in the top of processing tank, and pressure assembly carries out pressure control to carbon box.Wastewater tank is communicated with processing tank by liquid pipe, liquid pump, and processing tank is internally provided with transfer chamber, bottom chamber, microbial area and drainage chamber, and transfer chamber is configured with component sensing module for detecting the toxicity of wastewater.Processing tank is connected with multiple groups of valve and liquid pump, to realize wastewater delivery, residual wastewater backflow and carbon box quick replacement.The method first introduces wastewater into carbon box detoxification, and after detecting that toxicity reaches standard, it is sent into microbial area degradation, and if it does not reach standard, then replace carbon box and re-detoxification.The application realizes the coupling processing of adsorption detoxification and biodegradation, avoids microbial poisoning, and improves wastewater treatment stability and efficiency.

The utility model relates to a novel etching liquid daubing device in the technical field of etching daubing, which comprises a pushing mechanism and a daubing mechanism, the pushing mechanism comprises a bearing module and a driving module used for pushing the bearing module to move horizontally, the daubing mechanism comprises a material storage tank and a flexible daubing strip arranged at the lower end of the material storage tank, and the flexible daubing strip is arranged in the material storage tank. A plurality of liquid drainage holes are uniformly formed in the positions, corresponding to the flexible smearing strip, of the lower end of the storage tank, and a plurality of arc smearing parts are arranged at the bottom of the flexible smearing strip side by side. According to the utility model, by optimizing the structural design of the flexible smearing strip and adopting the arrangement of the plurality of arc smearing parts which are arranged side by side, the etching liquid can be uniformly distributed, the etching effect is effectively improved, a continuous and uniform liquid film is formed by the etching liquid in the smearing process due to the arrangement of the plurality of arc smearing parts, and the etching quality is improved. And the phenomena of liquid drop splashing and non-uniformity in a traditional spraying mode are avoided, the processing precision and consistency of the surface of the substrate are further improved, and the stability of the product quality is ensured.

The embodiment of the invention discloses a voice data processing method and device. According to the method, voice activity detection is adopted at the equipment terminal to screen out the first audio data stream only containing effective voice segments, and the invalid audio transmission quantity is reduced. According to the method, the problem that the voice frames processed by an equipment terminal are possibly discontinuous due to limited computing power of the equipment terminal is solved, and in order to solve the problem, the method adopts a mode of calculating the mute duration at a server side to determine the truncation point between effective voice segments, so that the completeness of sentences in audio data is ensured, and the processing precision of subsequent services is improved. Through the method, the contradiction between the transmission bandwidth and the processing precision can be effectively balanced. The voice data processing device provided by the embodiment of the invention also has the above beneficial effects.

The utility model relates to a microplate reader provided with an automatic sample adding inactivation module. The microplate reader comprises a rack, a pipetting module; a detection module; the inactivation module comprises an inactivation cavity, a multi-channel sample adding head and a rotating component; a plurality of independent inactivation units are arranged in the inactivation cavity, and each inactivation unit comprises a heating sheet, a cooling sheet and a temperature control sensor; the multi-channel sample adding head is connected with the top of the inactivation cavity through a lifting rod, the multi-channel sample adding head comprises a plurality of sample adding channels, each sample adding channel is provided with a micro pump and a flow controller, and a nozzle is arranged at the bottom of each sample adding channel; the rotating component comprises a rotating shaft and a plurality of fixed arms, the rotating shaft penetrates through the top of the inactivation cavity and is in transmission connection with the motor, and a sample holder is arranged at the tail end of each fixed arm. According to the microplate reader provided with the automatic sample adding and inactivating module, automatic sample adding, precise inactivating and multi-stage processing of samples can be achieved, the inactivating efficiency and the automation degree are remarkably improved, and meanwhile the universality and compatibility of equipment are enhanced.

The present application relates to the field of engineering surveying and mapping technology, in particular to an engineering surveying and mapping data intelligent acquisition, analysis and processing system and method, mainly applied to data acquisition, analysis and processing in the process of engineering construction; the system covers image acquisition, data classification, database storage, data processing, data analysis and decision feedback modules; the image acquisition module obtains an engineering surveying and mapping image feature set, the data classification module receives and classifies and stores the data; the database stores the classified data, the data processing module divides and calculates the importance of the image feature nodes; the data analysis module identifies abnormal data, and the decision and data feedback module receives the identification result and adjusts the data threshold; the system significantly improves the surveying and mapping precision through feature node processing and importance calculation; at the same time, the multi-level comparison mechanism improves the accuracy of abnormal detection, and provides efficient and accurate data support for engineering construction.

A computational offloading and resource allocation method for vehicle-road cooperative systems includes: establishing a communication model and a machine learning (ML) task model based on the vehicle-road cooperative system framework; establishing a computational model based on the communication model and the ML task model; modeling the computational offloading and resource allocation problem into a Markov decision model based on the computational model; introducing a multi-action output policy network into the SAC algorithm based on the Markov decision model; allowing the agent to continuously interact with the environment and obtain rewards, ultimately converging to a computational offloading and resource allocation strategy that maximizes the total reward; executing the ML task according to the optimal computational offloading and resource allocation strategy, and transmitting the processing results to an MEC server; constructing a digital twin vehicle on the MEC server; and then using the digital twin vehicle to predict the vehicle's behavior to assist in autonomous driving. This invention maximizes the processing accuracy of the ML task while reducing the total processing latency and energy consumption.

The invention belongs to the technical field of aviation safety, relates to an accident reproduction method and system based on multi-fault coupling and QAR reverse driving, and aims to solve the problems of low accident reproduction precision and poor authenticity in the prior art. The method comprises the steps of preprocessing accident flight data to generate an initial fault feature set, and calculating an initial total coupling degree in combination with an aviation fault library; determining a scene weight correction factor based on the initial total coupling degree, performing refinement processing on the flight data, generating QAR time sequence data and a standardized fault feature set, and constructing a multi-fault coupling relation model; according to the QAR time sequence data, the simulator is reversely driven, and a fault injection trigger node is positioned; and when the simulator reproduces to a trigger node, dynamically injecting each fault and the corresponding coupling relationship into the simulator in sequence for dynamic simulation. According to the method, the QAR data refinement processing front closed loop and the multi-fault coupling three-dimensional hierarchical model are constructed, so that dynamic fault injection is realized, and the accuracy and authenticity of data processing and accident reproduction are improved.

This invention discloses a deep learning-based method for measuring the depth of seismic wells in seismic exploration. The method divides field construction videos frame by frame into two categories: insertion / removal frames and non-insertion / removal frames. Image recognition and target detection training sets are created based on these two types of video frames. The datasets are input into the image recognition neural network and the target detection neural network, respectively, for training. The neural networks are then tuned to achieve optimal fit. The construction video is input into the trained image neural network for inference. If the video frame shows drill pipe insertion / removal operations, the target detection neural network detects the drill pipe position and records the drill pipe apex change curve. Gaussian filtering is applied to the curve, and a window selection algorithm is executed on the filtered curve to count the drill pipes. Finally, the depth is measured by multiplying the number of drill pipes by the drill pipe length. This invention effectively improves the overall video processing efficiency and significantly reduces the working time for measuring the depth of seismic wells from large amounts of actual construction video.

The application discloses an image color difference correction method and device, electronic equipment and a storage medium. The image color difference correction method comprises the following steps: acquiring first color difference gray values and second color difference gray values corresponding to each pixel point in an input image, and first color difference index values and second color difference index values of the pixel points; for any pixel point in the input image, based on the first color difference index values and the second color difference index values of the pixel point, and first mapping relationship and second mapping relationship, first color difference correction weight values and second color difference correction weight values of the pixel point are obtained; based on the color difference correction weight values, the input image is subjected to color difference correction to generate a corrected output image; wherein the first mapping relationship and the second mapping relationship are obtained by training a neural network model based on a plurality of sample images and a color difference perception result data set of the plurality of sample images.

The present application relates to the technical field of data processing, and especially to a multi-dimensional reconciliation method and system based on a group meal platform, which comprises the following steps: determining the state of a sub-node according to a steady-state metric value and a correlation index, and determining a reference data selection method according to the state of the sub-node; in the single-node direct selection, the sub-node corresponding to the maximum comprehensive evaluation value is selected as the reference data; in the multi-node compensation selection, a compensation method is determined according to a steady-state mutual correlation degree and a steady-state balance coefficient to select the reference data; and a reconciliation sub-node selection method is determined according to a reference comparison coefficient and a cross-correlation coefficient of the sub-node corresponding to the to-be-reconciled data and the reference data; and the present application can improve the reconciliation processing efficiency.

The invention relates to the field of image recognition, in particular to an image segmentation microstructure recognition method based on a convolutional neural network model, and the method comprises the steps: carrying out the sampling of a turbine blade, obtaining a plurality of samples, and carrying out the image collection of each sample after material processing, so as to obtain a plurality of target images; performing noise processing on each target image, determining an image feature state according to a contrast uniformity value and a local difference value of each target image after noise processing, and determining a final binarization threshold corresponding to each target image according to a pixel gray value corresponding to an inter-class variance or a sub-region based on the image feature state; performing threshold segmentation marking on the target image after binarization processing to obtain a label image; the ResUnet model is trained through the training data set; according to the invention, adaptive adjustment is carried out on sample processing and model training according to the actual collection condition of the collected sample, and the model training effect is improved.

This invention relates to a multi-bus data fusion processing method and system based on a logging tool, belonging to the field of oil and gas drilling measurement data transmission technology. The method includes receiving logging data collected by the logging tool's measurement probe and different data packets sent by the system bus; performing preliminary processing on the received logging data, including self-identification, self-diagnosis, and self-correction; identifying and converting the different received data packets; performing real-time fusion processing on the self-identified, self-diagnosed, and self-corrected logging data and the identified and converted data packets; and packaging the real-time fused data according to different communication protocols, and sending the packaged data to the system bus and / or storage unit. This invention forms a multi-bus data fusion processing system, which is embedded in the logging tool connected to the drilling string, enabling data interaction between the logging tool and other instruments in the instrument string, and achieving parallel processing and self-diagnosis of multiple sets of data.

This application discloses an audio processing method, apparatus, electronic device, and medium, belonging to the field of artificial intelligence technology. The audio processing method includes: acquiring a first time period of a first audio signal appearing in each of N frequency bands of a first audio signal, where N is a positive integer; for each of the first time periods, processing the first audio signal within a Fourier window corresponding to the frequency band of the first time period to obtain first audio signal feature information corresponding to each of the first time periods, with different Fourier windows corresponding to different frequency bands; inputting the first audio signal feature information corresponding to each of the first time periods into a generator network model for ultra-high-definition resolution processing to obtain a second audio signal corresponding to each of the first time periods; and obtaining a second audio signal based on the second audio signal.

This invention provides a membrane concentration device for seawater resource recovery, relating to the field of seawater treatment technology. It addresses the problems of high salinity in the concentrated brine produced after membrane concentration, which directly impacts the marine ecosystem; and the gradual decline in membrane performance and lifespan under prolonged operation and high salinity / pressure environments, requiring frequent replacements and increasing costs and maintenance workload. The device includes an ultrafiltration unit, a nanofiltration unit, and a multi-stage reverse osmosis membrane concentration unit. The ultrafiltration unit is used to treat seawater by removing suspended solids and colloids. The permeate outlet of the final membrane module is connected to the inlet of the first-stage membrane module via a return pipe. This invention employs a three-stage treatment process—ultrafiltration, nanofiltration, and CFRO—to progressively improve accuracy: ultrafiltration removes large particulate impurities, nanofiltration removes divalent ions to prevent scaling, and CFRO achieves high-concentration salt concentration and resource separation. Low-concentration permeate is returned to avoid waste. Furthermore, it can transform high-salinity wastewater into industrial raw materials, reducing environmental burden and industrial costs.

The invention discloses a seismic data processing method and device based on a StorSeismic deep learning normal form, electronic equipment and a storage medium. The method comprises the following steps: constructing a training set and a test set based on an actual seismic data set and a synthetic seismic data set; a seismic data processing model is constructed based on a Transform network; based on the training set, adopting a StorSeismic deep learning normal form to train a seismic data processing model; testing the trained seismic data processing model based on the test set; and processing actual seismic data based on the tested seismic data processing model. The method can solve the problems that a traditional Transform network lacks full utilization of specific geometric characteristics of seismic data, is high in sensitivity to high-frequency noise and adversarial attacks, and is easy to over-fit under the condition of limited data, achieves the effective processing of various seismic data processing tasks, and improves the processing efficiency and precision.

This invention relates to the field of image analysis technology, specifically a method for testing the color development quality of photochromic paint protective films. The method includes: acquiring initial state images and color-changing state images from multiple angles; determining candidate regions for color development anomalies and corresponding anomaly labels for the effective test areas of the initial state and color-changing state images, using the brightness and chromaticity changes of adjacent pixels within the effective test areas as the subject; comparing the candidate regions from multiple angles, verifying the deformation of the candidate regions based on the differences between the angles, and considering the verified regions as the output target regions; setting influence weights based on the brightness and chromaticity of the target regions under multiple color changes, and performing secondary detection on the target regions in conjunction with fading time to calibrate the image range information of the target regions; and configuring the test results of the target regions based on the image range information of the target regions. This method achieves both accuracy and processing efficiency in paint protective film analysis.

The application discloses a train wheelset fluorescent magnetic powder flaw detection image crack identification, positioning and measurement method, which comprises the following steps: partitioning a train wheelset and collecting a crack magnetic powder image of the train wheelset; pre-processing the collected crack magnetic powder image to highlight crack features; segmenting and extracting the crack features; and identifying, positioning and measuring the crack according to crack feature information. According to the shape characteristics of the crack, the application proposes a crack feature segmentation algorithm combining the crack area and the area of its circumscribed rectangle and the aspect ratio of the circumscribed rectangle of the connected domain, realizes accurate extraction of the wheelset crack features, and identifies, positions and measures the crack. Compared with other crack detection methods, the application not only realizes nondestructive detection of the surface crack of the train wheelset, but also effectively segments the features of the train wheelset crack, facilitates identification, positioning and measurement of the crack of the train wheelset, and thus realizes automatic wheelset crack magnetic powder flaw detection.