180 results about "Residual neural network" patented technology

The invention discloses a radar lifting control method and system based on meteorological monitoring, and relates to the technical field of radar lifting control, and the method comprises the steps: completing the switching of a power supply and communication after a radar is powered on, initializing a controller, collecting the data of a meteorological station, and generating a future fusion wind speed in real time through a Kalman filtering physical model and a residual neural network; future fused wind speed is converted into wind pressure for evaluation, the risk degree is judged according to the evaluation result, early warning is given out, and the controller is preheated to enter a lifting preparation state. The input stability is improved through meteorological data sliding window smoothing and feature extraction, wind speed dynamic prediction and uncertainty quantification are achieved through XGBoost prediction and residual variance estimation, the time sequence consistency and robustness are enhanced through remote API interpolation correction and adaptive extended Kalman filtering, residual correction is conducted through a neural network, the prediction precision is improved, and the prediction accuracy is improved. And a reliable decision basis is provided for radar lifting control.

The present disclosure belongs to the technical field of health status assessment of secondary circuits in power systems, and specifically relates to an image recognition method for a secondary circuit terminal based on contrastive learning and an improved CRNN. The method includes: step 1: pre-training sample data of a secondary circuit terminal block of a power system through the contrastive learning; step 2: improving a feature extraction layer of a CRNN by using a residual neural network; and step 3: introducing an ECA-Net on the basis of the step 2 to construct a recognition model for the secondary circuit terminal based on the contrastive learning and the improved CRNN. In the method of the present disclosure, an image of the secondary circuit terminal block can be accurately recognized, and the accuracy of image recognition, detection accuracy and detection efficiency can be greatly improved.

The invention discloses a visual detection test method for damage cracks of a solidified soil test piece, and particularly relates to the technical field of geotechnical engineering material detection. The method comprises the following steps: doping fluorescence labeling particles in a solidified soil test piece, and applying periodic load to induce crack formation after curing and forming; synchronously exciting fluorescent particles through an ultraviolet excitation light source, and collecting a short-wave fluorescence response image sequence; inputting the image sequence into a deep residual neural network model, identifying a nonlinear enhancement region of a fluorescence signal, and deducing a crack initiation position and a propagation path; constructing a three-dimensional dynamic model of crack evolution in combination with an image space reconstruction algorithm and time sequence comparison, extracting indexes such as a crack propagation rate, a bending angle and crack density, performing comparative analysis on the indexes and initial parameters of a material, evaluating a microstructure damage development mechanism, and outputting a visual damage evolution result; the method provided by the invention can realize high-precision, non-destructive and dynamic detection of the fine cracks in the solidified soil, and has the advantages of high identification sensitivity, strong modeling precision and good engineering adaptability.

The invention discloses a grinding state on-line monitoring method based on multilayer data fusion and a mathematical combined driving model, and the method comprises the steps: carrying out the signal collection through a multi-source sensor, carrying out the fusion according to a corresponding weight, and carrying out the noise reduction through stacking self-coding. When signals are extracted, frequency domain multi-dimensional features are fed into the residual neural network and the bidirectional double-layer LSTM fusion model after being subjected to Pearson correlation screening, and single-working-condition grinding state recognition is achieved. 10 types of discrete grinding states are defined, and a physical driving type empirical formula for working condition and monitoring signals and working condition and state characterization is established based on working condition parameters such as grinding depth and main shaft rotating speed. And generating a training data set by using an empirical formula, and training a cross-working-condition monitoring model through a transfer learning mechanism to realize cross-working-condition identification. According to the method, online accurate recognition of the grinding state is achieved, the limitation of traditional single-working-condition monitoring is broken through, and the problem of repeated data collection training is solved through the physical law of an empirical formula.

The invention discloses a porous asphalt concrete gap blockage identification method based on pavement noise signals, and belongs to the technical field of asphalt concrete blockage identification, and the method comprises the following steps: S1, collecting tire-pavement contact original noise signal audio data, and carrying out the preprocessing; s2, carrying out pavement gap blockage degree sensitive correlation analysis, and extracting multi-dimensional acoustic features; s3, dividing a single-layer drainage asphalt pavement data set and a double-layer drainage asphalt pavement data set, and giving a blockage level label to obtain a marked data set; s4, constructing a deep residual neural network model, and performing training and verification to obtain a gap blockage degree identification model; and S5, identifying porous asphalt concrete gap blockage. According to the method, the accurate classification of the gap blockage degree is realized through the blockage recognition model based on the acoustic characteristics, the optimal cleaning and maintenance opportunity can be determined, a scientific decision basis is provided for the maintenance of the porous asphalt concrete pavement, and the pertinence and economy of the maintenance are improved.

The invention discloses an underground pipe network construction segment supervision system based on deep learning. The method comprises the following steps: establishing a segment coordinate framework by using construction segment boundary information and mileage pile number information; constructing a super voxel graph; obtaining super voxel maps with consistent time sequences; generating a cross-modal feature vector; outputting a first abnormity identification result; updating the first dense residual neural network model to obtain a second dense residual neural network model; obtaining a third dense residual neural network model, and outputting a second uncertainty score and a third anomaly recognition result; and generating a structured anomaly report according to the third anomaly recognition result, the second uncertainty score and laws and regulations in the building information model and the geographic information model. According to the method, abnormal features of different spatial scales can be captured at the same time, meanwhile, multi-type abnormal structures are subjected to accurate grading identification, the generalization ability and the robustness adaptability to multi-scale anomalies are higher, and the probability of false alarm and missing alarm is remarkably reduced.

The invention discloses a three-dimensional vision measurement system error compensation method based on a spatial adaptive weighted RBF residual neural network, and particularly provides a spatial adaptive residual learning framework SAW-RFramework, in the first stage, global rigid body errors are eliminated based on a Kabsch algorithm, and point cloud rigid alignment is realized; in the second stage, a thin-plate spline kernel RBF interpolator is used for modeling a spatial smoothing error field, and position-related systematic deviation is captured; and in the third stage, a partition weighted residual neural network is adopted, and an improved residual compensation network is proposed for residual high-fluctuation nonlinear residual after RBF modeling: a regional importance coefficient is calculated based on a mean value and a variance of the residual in a self-adaptive three-dimensional grid, so that network training focuses on a high-error and high-fluctuation region. According to the method, the problems of insufficient global spatial modeling capability, low boundary region compensation precision, poor generalization capability and the like of an existing method can be solved; high-precision error compensation of the three-dimensional vision measurement system is realized, the measurement result reaches submillimeter precision, and the application reliability in robot control, man-machine interaction and precision manufacturing is improved.

The invention discloses a multi-sensor track management method based on spatio-temporal feature comparison, belongs to the field of multi-sensor information fusion and target tracking, and provides a track association algorithm based on spatio-temporal feature comparison by combining a bidirectional LSTM network and a residual neural network for the track association problem of multiple sensors and multiple targets. Through time feature extraction, space feature extraction, track feature comparison and a classifier module, conversion from a track association problem to a dichotomy problem is realized. And finally, through an analog simulation experiment, the convergence and generalization of the proposed network are verified. In a Monte Carlo experiment, the algorithm provided by the invention can reach the correlation accuracy of 95% or more, the trained network can quickly realize track correlation, and the track correlation effect is not sensitive to the influence of scene change.

The invention provides a self-adaptive pulse width modulation method, medium and system for a charging pile, and belongs to the technical field of charging piles. The self-adaptive pulse width modulation method comprises the following steps: acquiring voltage and current values of a battery, calculating a charge state parameter and a tilt change rate of the battery, and calculating a load coefficient and a third-order precision index by combining a load sensing algorithm; and analyzing the current waveform by using Fourier transform to obtain a harmonic distortion index, and calculating a delay compensation function value. The core technology is to call a deep residual neural network model of a double-path parallel architecture, input multi-dimensional characteristic parameters, output optimal pulse width modulation parameters and generate pulse width modulation signals to adjust charging characteristics. Meanwhile, the temperature and the internal resistance value of the battery are monitored in real time, parameter updating is triggered when the preset threshold value is exceeded, a complete closed-loop control cycle is formed, and the technical problem that the pulse width modulation parameter accurately adapts to the change of the charging state of the battery under the complex dynamic load condition is effectively solved.

The invention relates to the technical field of ship navigation behavior prediction, and discloses a ship navigation behavior prediction method and system based on interpretable artificial intelligence. Comprising the following steps: acquiring ship AIS historical data and historical marine environment data in a target sea area; based on a time scale and a space scale, fusing the ship AIS historical data and the historical marine environment data, and constructing a spatio-temporal data set; performing feature decoupling on a static variable and a time-varying variable of the spatio-temporal data set, and determining historical input data, known future input data and static input data; an interpretable space-time fusion neural network model is constructed based on a dynamic gating residual neural network, a space-time characteristic variable selection network, a Fourier analysis gating neural unit and an interpretable multi-head attention mechanism so as to accurately predict ship navigation behaviors such as a ship trajectory, an over-the-ground navigational speed and an over-the-ground course. And a navigation behavior influence factor weight visualization result is output, and the method has interpretability.

The invention discloses a six-dimensional force sensor calibration method based on an intelligent algorithm and ensemble learning, which improves calibration precision and system adaptability by combining data anomaly detection, the intelligent algorithm and the ensemble learning. The calibration method comprises the following steps: S1, building a six-dimensional force sensor calibration system; s2, loading and unloading experiments of force and torque are carried out on the six-dimensional force sensor on the standard calibration table, and multi-channel analog signal data output by the six-dimensional force sensor are obtained; s3, repeating the operation in the step S2 for a plurality of times; s4, data anomaly detection; s5, performing preliminary calibration by adopting a multiple linear regression model to obtain a preliminary decoupling matrix of the six-dimensional force sensor; s6, calculating error data after multiple linear regression calibration, and taking the error data as input characteristics of subsequent error compensation; s7, performing error compensation based on the residual neural network; and S8, determining calibration precision and outputting a final model.

The invention belongs to the technical field of buildings, and particularly relates to a cable defect positioning and intelligent identification system and method based on a broadband impedance spectrum. The system comprises a vector network analyzer which can inject a sweep frequency excitation signal into a target cable to be detected, a high-precision receiver which is used for collecting signal reflection / transmission response so as to obtain full-band complex impedance spectrum data, and computer equipment which receives test data. The positioning method based on Nuttall-Kai ser mixed window and cepstrum coupling does not need to depend on an intact cable parameter database, is high in anti-interference capability, can accurately identify defect positions, and is high in positioning precision; the one-dimensional impedance spectrum is converted into the two-dimensional image through the GAF, the automatic identification of the defect type is realized by combining the residual neural network (ResNet), the manual intervention is reduced, and the automation degree is high; the deep feature learning capability of ResNet is utilized, and a CBAM attention module is combined, so that the capture of a fine defect mode is enhanced, and the recognition accuracy is improved.

The invention provides a passive radiation noise multipath channel estimation method based on a complex residual neural network. According to the method, complex number time-frequency characteristics of ship radiation noise are fully utilized, in a non-cooperative scene without prior information, time delay and amplitude parameters of a multipath channel are accurately estimated through a deep neural network, and the problems that a traditional method is unstable in performance and low in estimation precision under the condition of a low signal-to-noise ratio are solved. According to a simulation experiment, the estimation precision, robustness and path resolution capability of the algorithm are systematically analyzed from multiple angles of different signal-to-noise ratios, path strength and the like, the result shows that the method has stable and reliable performance under the ship radiation noise background, and the effectiveness and popularization potential of the method are verified.

The invention provides a carbon monoxide plume identification and quantification method based on a domestic hyperspectral satellite, and the method comprises the steps: carrying out the radiometric calibration and atmospheric correction of the data of a visible short-wave infrared hyperspectral camera AHS I carried by the domestic hyperspectral observation satellite (GF-5B, Gaofen-5 B satellite); generating a carbon monoxide unit absorption spectrum by using a medium-resolution atmospheric transmission model MODTRAN; inverting a carbon monoxide enhancement value by adopting a matched filtering method; generating a binary plume mask through threshold segmentation; and constructing a deep learning model based on a residual neural network ResNet network, and realizing plume pixel-level identification and carbon monoxide column concentration enhancement value prediction. The method can identify the carbon monoxide plume with high precision and high efficiency, and is suitable for atmospheric pollution monitoring and pollution source quantification.

The invention discloses a rock-fill dam deformation digital twin construction method based on generative AI, and the core is that a conditional denoising diffusion probability model is adopted, a conditional sampling mechanism is constructed through classifier-free guidance and continuous conditional vectors, and finite element simulation data, monitoring data and operation data are efficiently fused. And further combining a residual neural network ResNet-18 and a K-means clustering method to carry out finite element data unsupervised classification, introducing combinatorial optimization, and identifying a finite element data category which is most matched with monitoring data, so that the condition-guided deformation field is generated. The framework is applied and verified on the highest two-estuary rockfill dam (303 meters) in the current built world. The result shows that the rockfill dam deformation digital twinborn body construction method based on the generative AI can efficiently reconstruct rockfill dam deformation, has high precision and real-time performance, remarkably improves the global deformation thorough sensing ability of the rockfill dam, and provides key technical support for safe operation of the rockfill dam.

The invention provides an alignment method for territorial space planning multi-modal data, and belongs to the technical field of big data processing.The alignment method comprises the steps that a quadtree structure is adopted to segment a remote sensing image, a three-dimensional index is established, a graph segmentation algorithm is utilized to divide vector data space subgraphs, a space-time weight matrix fusing space weight, time weight and credibility weight is constructed, and the three-dimensional index is established; inputting the preprocessed multi-modal data into a multi-modal semantic fusion model comprising a residual neural network, a bidirectional encoder and a graph convolutional network to extract a unified semantic representation vector, identifying conflict data by calculating semantic similarity, constructing a rule priority directed graph, and calculating a rule importance score by adopting a random walk algorithm; and executing the conflict resolution rule according to the priority and outputting the aligned data set. The technical problem that efficient alignment of the multi-source heterogeneous territorial space planning data on the semantic level is difficult to realize is solved.

The invention discloses a watermarking method and device suitable for any resolution image input, and the method comprises the steps: generating a fixed low-resolution watermark residual error through a watermark encoder according to an input carrier image and a watermark message, restoring the watermark residual error to the size of the carrier image, and superposing the watermark residual error with the carrier image to obtain a watermark-containing image; designing a watermark discriminator based on the residual neural network, wherein the watermark discriminator is used for distinguishing the difference between the watermark-containing image and the carrier image; a watermark noise layer is used for simulating various distortion operations suffered by an image under a digital channel, and disturbance is applied to the watermark-containing image processed by a watermark discriminator; and transmitting the watermark-containing image to which the disturbance is applied in a real channel, inputting the watermark-containing image into a robust watermark decoder, and decoding and extracting a watermark message. According to the method and the device, image input of any resolution can be processed, good invisibility and robustness can be kept, and copyright protection and leakage traceability of the image of any resolution can be realized.

The invention discloses an automatic control method and system for fumaric acid production. The method comprises the following steps: collecting process variables in a fermentation process; constructing a soft measurement model fusing physical modeling and a residual neural network to obtain an acid production rate prediction value and prediction confidence; calculating a feed-forward control quantity, and adjusting the weight according to the prediction confidence to obtain a feed-forward control output; constructing a state vector, and establishing a Markov decision process; training a neural network control strategy by adopting a near-end strategy optimization algorithm, and outputting a proportionality coefficient, an integral coefficient and a differential coefficient; calculating feedback control quantity, fusing the feedback control quantity with feed-forward control output, and applying change rate and amplitude limitation to obtain final control output; and switching to a feedback priority control mode under a specific condition, and regularly updating a soft measurement model and a control strategy parameter. According to the invention, high-precision stable control of pH in the fumaric acid fermentation process is realized, the production efficiency and the process robustness are improved, and the method is suitable for automatic control of the large-scale fermentation process.

A multi-stage evaluation electric power system short circuit and open circuit fault diagnosis method comprises the steps that voltage and current sensors and fault recording devices are installed at key nodes such as a main transformer and a distribution line, and three-phase voltage and current waveform data and zero-sequence components are collected in real time and transmitted to a fault diagnosis server; the server firstly performs preliminary judgment and alarm triggering on short circuit, open circuit and one-way grounding / electric leakage faults, then performs discrete wavelet decomposition and multi-scale time-frequency domain feature extraction on acquired waveforms by using a deep fault classification model based on wavelet analysis and neural network fusion, and realizes fine classification of the faults through a multi-layer residual neural network. And finally, establishing a fault diagnosis result self-adaptive evaluation and response mechanism, dynamically adjusting an alarm level and a switching scheme according to a fault type, realizing protection equipment tripping and fault isolation for a short-circuit fault, generating a standby line switching suggestion for an open-circuit fault, and effectively improving the accuracy and coping capacity of power system fault diagnosis.

The invention discloses a power distribution equipment state detection method and device based on multi-modal data fusion, terminal equipment and a storage medium, and belongs to the field of power distribution equipment. Performing feature analysis on the acquired multi-modal data by using a time sequence convolutional network, a residual neural network and a Transform network to obtain a local abnormal fluctuation feature, an infrared temperature spatial distribution feature and a video global feature representation; performing feature splicing on the local abnormal fluctuation feature, the infrared temperature spatial distribution feature and the video global feature representation to obtain a fusion feature, and inputting the fusion feature to a long-short term memory network to obtain a historical hidden state sequence; and determining the equipment state of the to-be-detected power distribution equipment according to the historical hidden state sequence, the fusion features and edge equipment deployed in the local network domain of the equipment. By implementing the application, the problem of low accuracy of state analysis of the power distribution equipment caused by only depending on single modal data in the prior art can be solved.

The invention relates to the technical field of oil type gas monitoring, in particular to an oil type gas laser monitoring and early warning device and method, and the device comprises a processor control panel, a laser driver, a laser device group, a light path change-over switch, a diffusion type light absorption cell and a tuning fork detector; emergent laser with different gas spectrum absorption wavelengths is generated; switching each path of laser beam to a diffusion type absorption cell; respectively demodulating based on the acquired acquisition signals to obtain second harmonic signal voltages of optical signals of methane, ethane and hydrogen sulfide gas; inputting the second harmonic signal voltage of the optical signal absorbed by the to-be-detected gas into the residual neural network, identifying the methane and ethane gas overlapping spectrum, and outputting the methane and ethane gas concentration in the to-be-detected gas; and generating a multi-parameter early warning signal according to the methane and ethane gas concentrations and the hydrogen sulfide concentration. The light absorption detection can be accurately carried out on methane, ethane and hydrogen sulfide in the oil type gas, and the methane and ethane gas overlapping spectrum can be effectively identified.

This invention belongs to the field of fault detection and relates to a bearing fault detection method based on a residual neural network using a cross-attention mechanism. The fault detection method comprises the following steps: First, preprocessing the input data; second, extracting features from the data; third, fusing the data features; fourth, detecting data defects; and fifth, inputting the processed data into the network to obtain the corresponding fault detection category and result. This invention has a wider applicability to bearing fault data and higher accuracy, effectively improving the accuracy and speed of bearing fault detection in complex environments. Through the cross-attention mechanism, data features can be extracted more effectively.

The embodiment of the invention provides a high-precision single well stratigraphic division method and device, and relates to the field of oil and gas field exploration and development, and the method comprises the steps: determining a multi-dimensional original vector group according to parameter data of all sampling points of a single well and a preset window value; inputting the multi-dimensional original vector group into a residual neural network model for window division and feature processing to obtain a feature vector group of each window; calculating the feature vector groups of every two adjacent windows through a clustering algorithm, and determining the clustering distance of every two adjacent windows; clustering the sampling points in each window according to the clustering distance to obtain a clustering result; and performing stratigraphic section division on the single well based on the clustering result to obtain a corresponding stratigraphic section division result. Through the method provided by the invention, the accuracy of stratigraphic division is improved.

The application discloses a non-intrusive power load decomposition method based on micro machine learning, comprising the following steps: S1, collecting power load data and performing feature extraction, fast Fourier transform and normalization processing, and dividing the processed data into a training data set and a test data set; S2, performing iterative training on the power load data of the training data set based on a residual neural network structure to obtain a power load decomposition model; S3, quantizing and compressing the power load model; S4, inputting the power load data of the test data set into the compressed power load decomposition model and evaluating whether the decomposition model meets the use requirement; and S5, decomposing the real-time power load data of the user by the power load decomposition model to monitor the operating power of each power equipment. The application uses high-frequency sampling data and performs FFT decomposition, can effectively construct load characteristics and improve the model accuracy.

This invention discloses a method and system for magnet pose estimation based on magnetic dipole constraints. The method includes estimating the initial orientation and position of the magnet based on the triaxial magnetic field vectors measured by each magnetic sensor in a magnetic sensor array to determine the initial estimated orientation unit vector and initial estimated position of the magnet; iteratively solving the initial pose state parameters of the magnet using a nonlinear optimization algorithm based on a magnetic dipole model to obtain the final estimated orientation unit vector; obtaining the position residual and orientation residual using a pre-trained residual neural network model; and compensating the initial estimated position and the final estimated orientation unit vector based on the position residual and orientation residual to obtain the final magnet pose estimation result composed of the magnet orientation. This invention aims to overcome the influence of initial value deviation, sensor error, and environmental interference factors, improve robustness and generalization ability in complex environments, and meet the high-precision pose detection requirements of magnetic positioning.

The invention discloses an automobile ornament laser welding surface defect online identification method and system, and the method comprises the steps: projecting a structured light field to a welding surface: obtaining the coordinate information of a welding region, driving to carry out the spatial displacement compensation, and adjusting the aperture range of a diaphragm; capturing reflected light of the structured light field in real time to form an image sequence, which comprises the following steps: receiving a pulse signal output by a welding equipment encoder to trigger shooting, and adjusting exposure time to a preset mapping interval according to real-time light intensity feedback; performing defect feature extraction and defect classification identification on the image sequence, namely performing edge extraction by adopting an edge detection algorithm, and calculating contour geometric parameters and gray-level co-occurrence matrix parameters to extract features; performing defect classification and identification through a residual neural network integrated with a feature attention mechanism; and a defect recognition result is obtained, and welding process parameters are adjusted based on the defect recognition result. According to the invention, the laser welding surface defect on-line identification of the automobile ornament can be realized, and the identification precision is improved.

The invention relates to the technical field of battery health management, and discloses a lithium battery SOH prediction method and system based on a Koopman gating residual neural network, and the method comprises the steps: obtaining the electrochemical impedance spectroscopy data of a lithium battery; fitting and extracting a first physical feature from the electrochemical impedance spectrum data based on an equivalent circuit model, determining a feature frequency point, and extracting an electrochemical impedance imaginary part value under the feature frequency point as a second frequency domain feature; generating an initial prior estimation value of the battery health state based on the first physical characteristic, and performing dynamic trend correction on the initial prior estimation value to obtain a corrected prior estimation value; and inputting the first physical feature, the second frequency domain feature and the corrected priori estimation value into a pre-trained Koopman gating residual error neural network model, and outputting a final prediction value of the battery health state through a self-adaptive gating fusion mechanism of the model. According to the invention, accurate and high-robustness online prediction of the health state of the battery is realized.

The application discloses a bearing life prediction method based on an attention mechanism and a residual neural network, comprising the following steps: firstly, using time domain features and a k-means clustering algorithm to adaptively identify the degradation state of the bearing; using an empirical mode method to decompose the bearing data signal and remove noise; and training and using a residual neural network with an introduced attention mechanism to complete the life prediction of the bearing. The application has the following advantages: the degradation state of the bearing is adaptively identified, avoiding the interference of data generated in the normal state on the life prediction; the empirical mode method is used to decompose the bearing data signal and remove noise, preventing the mutual interference and cancellation of degradation and fault features at different frequencies, and thus making the features learned by the model more effective; and the residual neural network with the introduced attention mechanism is used, so that the neural network model is always affected by the original signal and the attention is focused on the channel where the fault features are located, preventing the model from overfitting.

Provided are methods for generating a multitask machine learning model based on time series data, that may include receiving input time series data associated with an input time series of data points, calculating a pairwise distance between the input time series and a plurality of time series templates, providing the pairwise distance as a first input to a building block of a residual neural network, where the residual neural network has a plurality of multi-dimensional convolutional layers; generating a first output of the first building block of the residual neural network based on the first input, generating a final output of the residual neural network based on the first output, and generating a first output of a multitask machine learning model using a first output layer and a second output of the multitask machine learning model using a second output layer. Systems and computer program products are also disclosed.

The application discloses a kind of stone automatic grading method and system based on deep learning, the method is using stone picture quality evaluation module, stone grading feature extraction module and stone grading data fitting module realizes industrial belt high-speed transport scene stone automatic grading task, stone picture quality evaluation module is a residual neural network based on multilevel feature, for the high-quality dynamic frame of picture quality is evaluated and screened;Stone grading feature extraction module is a segmentation module based on digital image processing using watershed algorithm and concave point detection algorithm, for the stone grading feature information extraction of screened picture;Stone grading data fitting module is an improved multilayer perceptron, for the data fitting correction error of initial stone grading feature result.The application has higher precision and better generalization ability in different backgrounds, and can accurately detect different types of stone transport scene.