229 results about "Support vector regression model" patented technology

The invention discloses a power equipment fault prediction method based on support vector regression, and aims to improve the prediction accuracy. The method comprises the following steps: firstly, collecting multi-dimensional operation parameter data of power equipment, and constructing an original monitoring data set; a data set is segmented through a sliding time window, wavelet transform is used for noise reduction, and then a normalization method is applied to generate a standardized data set. Time domain and frequency domain features are further extracted, a feature vector space is constructed, an optimal feature subset is screened through a mutual information method, and a training data set is generated; performing preliminary training on a support vector regression model by using the training data set to generate candidate support vectors, and screening a support vector set with the maximum contribution through an optimization algorithm; a grid search method is adopted to adjust kernel function parameters and penalty factors, and a support vector set is further optimized. And evaluating the performance of the model through cross validation, applying the optimized support vector regression model to power equipment fault prediction, and generating a fault occurrence probability prediction value.

The invention relates to the technical field of environmental impact assessment, in particular to an intelligent environmental impact assessment method and system, and the method comprises the following steps: collecting environmental data including pollutant discharge, water resource consumption, air quality change and biodiversity, building an impact factor data set, and segmenting the data set to obtain an impact factor data set; and calling a kernel function type to carry out numerical prediction on pollutant emission. According to the method, by integrating data processing and machine learning technologies, the accuracy and efficiency of environmental impact assessment are improved, numerical prediction is performed on environmental data by using a kernel function and a support vector regression model, the adaptability and precision of a prediction model are enhanced, model parameters are optimized through automatic population generation and crossover mutation operation, and the accuracy and efficiency of environmental impact assessment are improved. The model can better learn and adapt in a complex data environment, sensitivity and response speed to environmental changes are further improved through dynamic clustering analysis and time sequence adjustment, and scientific support is provided for formulating timely and effective environmental protection measures.

The invention discloses a network security situation awareness and dynamic analysis method based on a graph neural network, and the method comprises the following steps: S1, collecting network related data, and constructing a dynamic security knowledge graph; s2, preprocessing the collected data to generate a node feature matrix; s3, based on the node feature matrix, generating a node embedding representation through an improved graph neural network and a graph attention mechanism, and optimizing the node embedding representation in combination with the graph attention mechanism; s4, establishing a time sequence prediction model based on the improved support vector regression model, and identifying potential abnormal behaviors and security threats; s5, evaluating the identified security threat, and predicting the risk level of the security threat; s6, generating a response strategy according to a security threat assessment result; and S7, executing the response strategy. According to the method, the improved graph neural network and the support vector regression model are utilized to carry out network security situation awareness and dynamic analysis, and the method has efficient and accurate security threat identification and adaptive response capabilities.

The invention relates to a signal emission control method and device based on optical communication and a signal emission laser. The method comprises the following steps: acquiring output light intensity data of the vertical cavity surface emitting laser in a preset temperature range and a bias current range, and determining a bistable working characteristic curve; acquiring a vibration resonance gain signal; gaussian distribution transformation and target phase sampling are performed on the vibration resonance gain signal, and a signal nonlinear compensation model is established; applying a driving current pulse synchronous with communication data, symmetrically sampling the output signal subjected to transient response control at a rising edge and a falling edge, and constructing a signal quality evaluation vector; and inputting a support vector regression model, establishing a mapping relation between the light intensity characteristic parameter and the driving control parameter, and generating a real-time control signal of the laser. According to the invention, the self-adaptive accurate control of the vertical cavity surface emitting laser is realized, and the stable operation of the signal emitting laser under the temperature and bias current fluctuation condition is ensured.

The invention provides an industrial internet security situation analysis method and system based on support vector regression, and relates to the technical field of industrial internet security, and the method comprises the steps: collecting alarm data, and recognizing a homologous alarm event through a time window sliding mode; establishing a time sequence propagation matrix, calculating a propagation probability, constructing an attack scene graph and extracting features; calculating a security situation index by using a support vector regression model; and selecting an optimal protection node combination based on the node centrality and the protection decision tree. According to the method, the threat propagation path can be accurately identified, the security situation can be quantified, and the precise protection of the industrial internet environment can be realized.

The invention relates to the technical field of soil detection and multi-source information fusion, in particular to a soil moisture content cooperative detection method and system.The method comprises the steps that a target area is determined, and a target thermal infrared image of surface soil of the target area is obtained; extracting target characteristic parameters related to the moisture content from the target thermal infrared image, inputting the target characteristic parameters into the trained BP neural network, and predicting to obtain a surface soil moisture content distribution diagram; based on the surface soil moisture content distribution diagram, determining a target range region with abnormal moisture content through threshold comparison; after the air coupling stepping radar is driven to be aligned with the thermal infrared imaging system in a space-time mode, scanning is conducted in a target range area, and a radar reflection coefficient extracted from an obtained radar image and phase difference information serve as radar characteristic parameters; and performing modeling analysis on the radar characteristic parameters based on a support vector regression (SVR) model, and obtaining soil profile moisture content distribution of the moisture content abnormal region by constructing a nonlinear mapping relation and optimizing model hyper-parameter inversion.

The invention relates to the technical field of marine ecological environment monitoring and data analysis, and particularly discloses a sea area phytoplankton biodiversity index prediction method and system based on multi-model integration and feature engineering. After preprocessing, constructing four groups of nonlinear interaction characteristics of a temperature-salt relationship, oxygen-salt balance, chlorophyll chemical oxygen demand coupling and a nitrogen-phosphorus ratio based on environmental factors, combining station characteristics with basic environment and interaction characteristics, carrying out variance threshold screening, inputting a characteristic set into a multi-model integration framework containing models such as linear regression and gradient lifting, and carrying out multi-model integration; training and tuning according to a time sequence segmentation strategy, selecting model output according to a decision coefficient, using a result if the decision coefficient of the support vector regression model is within a preset range, and otherwise, taking a gradient lifting and extreme gradient lifting tree model to predict a mean value. And the prediction accuracy and the model generalization, stability and reliability are improved.

The invention discloses a multi-parameter cooperative temperature control system of an energy-saving industrial oven, and relates to the technical field of temperature control of industrial ovens, load data in the oven are collected in real time through a high-precision sensor, and a Kalman filter is used for cleaning, calibrating and filtering environmental noise and drift to obtain smooth and stable data; analyzing a real-time load based on a support vector regression model, capturing a nonlinear relation in combination with a radial basis function kernel, and predicting temperature recovery time and expected energy consumption; a reinforcement learning algorithm is adopted, a high-dimensional state space is analyzed through a deep Q network, the heating power and the airflow speed are dynamically optimized, and optimal control parameters are output; and the control module executes the adjusted setting, monitors the temperature and the energy consumption in real time, feeds back data to a prediction model and a learning strategy, and periodically updates parameters to adapt to load and environment changes. The adaptive capacity and long-term stability are remarkably improved, and the method is suitable for industrial scenes with load change, equipment aging or environment fluctuation.

The invention relates to the technical field of artificial intelligence, in particular to an injection molding process parameter optimization method and system based on hybrid algorithm and model fusion, and the method comprises the steps: optimizing a parameter combination of a support vector regression model through a simulated annealing algorithm, building a weighted fusion model based on the optimized support vector regression model and a random forest, and optimizing the model; constructing a hybrid model of an adaptive selection weighted fusion model and an optimized support vector regression model; constructing a three-objective optimization model including buckling deformation, volume shrinkage and production energy consumption, and searching a Pareto optimal solution set in a process parameter space by adopting a multi-objective genetic algorithm by taking the hybrid model as a target value evaluation tool; carrying out local correction on the key process parameters by adopting a gradient descent method until the deviation falls back to be within a preset threshold value, and obtaining optimized process parameters; the defect rate of products can be reduced, and meanwhile production energy consumption is reduced.

A method and system for power grid optimization control based on a linear time-varying model is provided. The system model is first estimated offline using historical data combined with a piecewise linear regression method that supports a vector regression model. For online applications, a time-varying linear model prediction control framework based on a piecewise linear model coordinates the optimization control of both fast and slow regulating devices in the power grid. This approach does not require precise system model parameters, instead learning the power grid model from historical data. The method optimizes voltage distribution, reduces operating costs, and addresses issues like bad data and collinearity in the historical data, improving the voltage quality and enabling optimal operation even with incomplete models.

The invention relates to the technical field of concrete quality control, in particular to a concrete strength prediction system and method based on SVR. According to the invention, multi-dimensional data of rebound test, compressive strength test and carbonization depth value measurement are collected through the data acquisition module, rich input characteristics are provided for the model, and the prediction capability of the model is enhanced; and abnormal data is eliminated by adopting a Grubbs criterion, so that the quality and consistency of input data are ensured, and the influence of noise on model training is reduced. According to the method, a concrete rebound curved surface regression equation and a support vector regression model are established respectively and serve as prediction criteria, data fusion is carried out through a Kalman filter, and the prediction precision and reliability are further improved. Through automatic and standardized data processing procedures including data acquisition, preprocessing, regression prediction, SVR prediction, data fusion and the like, rapid prediction of the concrete strength is realized, and the prediction efficiency is improved.

The invention discloses a weak surrounding rock tunnel deformation intelligent monitoring method and system. The monitoring method comprises the following steps: acquiring original time sequence data of tunnel deformation monitoring; decomposing the original time sequence data to obtain decomposed time sequence data; obtaining a feature vector set based on the decomposed time sequence data; the feature vector set is input into a tunnel deformation monitoring model, a tunnel deformation monitoring result is obtained, the tunnel deformation monitoring model is constructed through a support vector regression model and is obtained through training of a training set, and the training set is historical monitoring data. Through fusion application of various algorithms, the tunnel deformation monitoring precision and reliability are remarkably improved, and powerful support is provided for safe operation of the tunnel.

The invention discloses a battery RUL prediction method, device and equipment based on variable mode decomposition, and a medium, and relates to the technical field of batteries, and the method comprises the steps: obtaining battery capacity degradation data and a battery capacity failure threshold value, carrying out the smoothing of the battery capacity degradation data through variable mode decomposition, obtaining a plurality of mode components, and carrying out the smoothing of the multiple mode components; and obtaining battery capacity key feature data according to the plurality of modal components, inputting the battery capacity key feature data into the optimized support vector regression model to obtain a battery capacity predicted value, and when the battery capacity predicted value is smaller than a battery capacity failure threshold, performing calculation to obtain the remaining service life of the battery. According to the method, the battery capacity degradation data is smoothly processed through variable mode decomposition, key features are extracted and input into the optimized SVR model to predict the battery capacity, when the predicted battery capacity is lower than a failure threshold value, the remaining service life of the battery is calculated, and the RUL prediction precision and robustness are improved.

The invention relates to the technical field of laser frequency stabilization control, and discloses a multi-mode laser frequency stabilization error feedback correction system driven by machine learning. The system comprises a multi-modal data acquisition module for acquiring various data to generate a multi-modal time sequence data set; the error feature extraction module is used for extracting an error feature tensor by using a kernel principal component analysis algorithm; the dynamic compensation modeling module is used for constructing a support vector regression model to generate a dynamic compensation strategy matrix; and the feedback control optimization module is used for designing a self-adaptive model prediction control framework to generate a closed-loop correction instruction sequence. In addition, a self-adaptive correction execution module, an error traceability analysis module and an abnormal mode recognition model are further arranged. Through multi-modal data acquisition and analysis and intelligent modeling and control, high-precision laser frequency stabilization is realized, the system can effectively adapt to a complex environment, the stability and reliability of the system are improved, and the system has a wide application prospect in the fields of laser processing, optical communication and the like.

The invention discloses a Raman spectrum fluorescence baseline correction method and device, and the method comprises the steps: collecting a time domain signal of a to-be-detected sample, and obtaining an IRF function of a to-be-detected sample collection system; pre-processing the time domain signal, and carrying out deconvolution on the pre-processed time domain signal and an IRF function to obtain a deconvolution result; fitting the second half data of the deconvolution result by using a support vector regression model to obtain a fitted curve; wherein a radial basis function is adopted as a kernel function, and hyper-parameters of the support vector regression model are optimized through a grid search and cross validation method; performing difference operation on the deconvolution result and the fitting curve to obtain a Raman attenuation curve; and windowing the Raman attenuation curve and averaging to obtain a Raman spectrum. According to the method, a fluorescence attenuation model and parameters thereof do not need to be preset, the IRF deconvolution result is processed and the time domain baseline is predicted by utilizing the characteristics of the time domain signal data and the support vector regression model, the signal-to-noise ratio of the extracted Raman signal can be greatly improved, a good effect is achieved, and the method is suitable for practical application.

The invention relates to the technical field of laser spectrum gas detection, and particularly discloses a TDLAS (tunable diode laser absorption spectroscopy) double-gas concentration synchronous calibration and decoupling method in the presence of spectrum aliasing interference, and the method comprises the following steps: acquiring a multi-band harmonic signal subjected to adaptive optimization while acquiring environmental parameters; an optimal frequency band signal is determined based on the signal quality index; and calculating a spectral aliasing coefficient according to the characteristics of the optimal frequency band signal, dynamically adjusting operation parameters of a search algorithm based on a numerical value interval in which the spectral aliasing coefficient is located, and determining an initial range of dual-gas concentration in a search space. According to the method, the improved genetic algorithm based on the spectral aliasing coefficient dynamic adjustment search strategy is combined with the incremental support vector regression model, stage processing from coarse calibration to fine decoupling is achieved, and the calculation efficiency and the search precision can be automatically balanced when the spectral aliasing degree changes; and online correction is carried out on the model in combination with aging characteristics.

The invention discloses an agricultural load prediction method and system based on multivariate time sequence decoupling multi-modal learning, and belongs to the technical field of agricultural load prediction. Comprising the steps of collecting historical agricultural load and meteorological data; decomposing historical agricultural load and meteorological data by using multivariate variational mode decomposition to obtain cycle, trend and residual mode components; and respectively constructing a time convolutional neural network, a bidirectional gating cycle unit and a support vector regression model for the decomposed period, trend and residual modal component data set, and fully mining feature information of each mode after decomposition, thereby realizing accurate prediction of agricultural load. According to the method, the potential nonlinear space-time coupling relationship between the agricultural load and the meteorological factor is captured, the prediction effect in a seasonal periodic fluctuation scene of the agricultural load and a long-term trend and agricultural load abnormal scene is improved, the agricultural load prediction precision is improved, and a support is provided for reliable and stable operation of a power grid.

The invention discloses a cable early fault early warning method and system based on multi-feature fusion and an iterative prediction model, and relates to the technical field of wire and cable health management, and the method comprises the steps: carrying out the standardization processing of the time series data of a plurality of early fault features, screening out important features through a gray correlation analysis method, and carrying out the calculation of the early fault features; the weights of the multiple features are calculated through a subjective and objective comprehensive weight method, and the multiple features are fused into a comprehensive feature; constructing an augmented support vector regression model; training an augmented support vector regression model by using the known comprehensive features, and carrying out iterative prediction on unknown comprehensive features; and determining the residual service life of the cable according to the comprehensive characteristic threshold value corresponding to the cable failure and the iterative prediction result, and setting a residual service life early warning threshold value to complete early fault early warning. According to the method, the problems of insufficient data utilization and low feature change prediction accuracy in the early fault early warning process of the cable are solved, and the method is of great significance for guaranteeing safe operation of a cable system.

The invention provides a reactor cooler performance evaluation and predictive maintenance method and device, and relates to the technical field of performance evaluation, and the method comprises the steps: collecting the historical data of a reactor cooler in real time; obtaining an input variable feature set; inputting a prediction model, and cooperatively predicting cooling efficiency and future energy consumption data in combination with real-time monitoring data to obtain a performance degradation curve of the reactor cooler; calculating to obtain a temperature difference, respectively inputting the temperature difference into a support vector regression model and the prediction model, and generating a corresponding predicted reactor cooler temperature difference value; the final temperature difference prediction value of the reactor cooler is obtained, and the service life curve of the reactor cooler is calculated in combination with the cooling efficiency; and determining a performance evaluation grade of the reactor cooler, and determining a maintenance time point and a maintenance strategy according to the performance evaluation grade. Through efficient data processing and feature extraction, the performance evaluation grade of the reactor cooler is determined, and the accuracy of performance determination of the reactor cooler is improved.

The invention relates to the technical field of data processing, in particular to a fan tower fastening bolt on-line monitoring system based on a sensor, the system comprises a processor and a memory, and the processor executes a computer program of the memory to realize the following steps: acquiring a multi-dimensional monitoring data sequence of any bolt on a fan tower at a current sampling moment; dividing the multi-dimensional monitoring data sequence into subsequences and a current subsequence, and obtaining a temperature and wind pressure influence coefficient between every two subsequences; according to a time interval between a sampling moment and an initial sampling moment of each sub-sequence, obtaining a regression weight of each temperature wind pressure influence coefficient, constructing a weighted loss function, training a support vector regression model to obtain an optimal prediction model, and obtaining a prediction value; and according to the difference between the temperature wind pressure influence coefficient between the current sub-sequence and each sub-sequence and the predicted value of the current sub-sequence, abnormal early warning is performed on any bolt, so that the accuracy of a bolt fastening state monitoring result is improved.

The invention relates to the technical field of training assistance, and provides a training assistance system for five-poultry paddling and tiger paddling actions, which comprises a data acquisition module, an action recognition module, a whole body coordination evaluation module and a user interface module. The data acquisition module is used for acquiring motion data when a user executes a tiger play action through a plurality of sensor groups; the action recognition module performs preprocessing and feature extraction on the collected data, decomposes the tiger play action into a basic action sequence, and recognizes the matching condition of the current action and a standard action template through a deep learning network; the whole body coordination evaluation module establishes a biomechanical model based on a reverse dynamics method, calculates multi-dimensional coordination indexes including joint angle change, mechanical load distribution and dynamic stability, and generates an action coordination score through a support vector regression model; and the user interface module is used for displaying identification and evaluation results and providing improvement suggestions when the actions do not conform to specifications or the coordination is insufficient.

The invention relates to the technical field of disease course trend modeling, in particular to a subarachnoid hemorrhage disease course trend modeling system fusing multi-source data. Four kinds of signals of intracranial pressure, blood flow velocity, cerebrospinal fluid pressure and electroencephalogram of a monitored object are synchronously collected, an instantaneous phase is extracted through Hilbert transform, a time window is adaptively adjusted according to the brain blood vessel conduction delay characteristic of an individual, and phase locking indexes among three pairs of signals are calculated. A multivariate coupled oscillator model is established, phase track topology invariant features are extracted, and comprehensive trend indexes are generated through tensor fusion. An individualized four-dimensional phase entropy baseline mode is established, and a double-layer early warning mechanism is adopted: when second derivative continuous symbol overturning occurs in all three phase locking indexes, early warning is directly performed, and when any two phase locking indexes are overturned, a trend index needs to be synthesized for confirmation. And predicting a state level, a trend level and an expected evolution trajectory based on a support vector regression model. According to the invention, precise disease course prediction and early warning are realized, and a basis is provided for clinical decision making.

The invention relates to a wide-temperature-interval battery health state estimation method based on an electrochemical impedance spectrum, and belongs to the technical field of battery health state monitoring. Aiming at the problem of battery state of health (SOH) estimation misalignment caused by electrochemical impedance (EIS) temperature interference, the invention provides the following scheme: collecting EIS data within 10-30 DEG C, including extracting temperature interference resistance features, principal component dimensionality reduction features and low variance phases, constructing combined features # imgabs0 # and # imgabs1 #, and realizing SOH estimation through a support vector regression (SVR) model. The method breaks through the dependence of a traditional method on a constant temperature condition or a temperature sensor, realizes SOH high-precision estimation in a wide temperature interval, and is suitable for lithium / sodium ion battery on-line health monitoring.

The invention discloses a method for detecting types and concentrations of trace organic matters by combining terahertz spectroscopy with machine learning, which specifically comprises the following steps: preprocessing terahertz time-domain spectroscopy data of a sample, and optimizing model parameters by using an improved particle swarm optimization algorithm; training the optimized model based on 5-fold cross validation, and evaluating the performance of the model; and inputting test set data into the trained model for classification prediction, and outputting a result. According to the method, terahertz signal sensitivity is enhanced through metasurface structure optimization, a support vector classification model and a support vector regression model are optimized in combination with an improved particle swarm algorithm, a qualitative and quantitative double-model collaborative analysis system is established, and sensitive and efficient detection of elementary substance types and concentrations of organic substances in a million fraction concentration (ppm) magnitude is achieved.

The invention relates to the technical field of shield tunneling machine cutter detection, and discloses a shield tunneling machine cutter wear automatic detection system and method, and the system obtains a shield tunneling machine cutter image collected by a camera assembly, and carries out the image preprocessing, noise suppression, cutter region segmentation and region-of-interest extraction to obtain a cutter target ROI image, image segmentation based on a gray threshold is adopted in cutter region segmentation, the gray threshold is determined through the cutter geometric features and the cutter texture features of each sliding window region, and the method can improve the accuracy of image segmentation and the accuracy of the cutter geometric features and the cutter texture features. And finally, performing tool wear feature extraction on the tool target ROI image, and inputting the tool wear feature into an intelligent tool wear detector based on a support vector regression model to obtain a detection result. Therefore, automatic evaluation of the wear degree is realized, human intervention is reduced, and the detection efficiency and accuracy are improved.

The invention discloses a biomass gasification product prediction method based on mechanism and data fusion driving, and the method comprises the steps: constructing a biomass gasification product prediction mechanism model, and inputting biomass gasification operation parameters into the model to obtain a prediction result; calculating a residual vector between a prediction result and actual experimental data; establishing a support vector regression model between the biomass gasification operation parameters and the residual vectors; and taking the output of the support vector regression model as an error compensation item of a biomass gasification product prediction mechanism model, constructing a biomass gasification product prediction model based on mechanism and data fusion driving, and predicting a biomass gasification product by using the model. Through fusion of a mechanism model and data driving, the problem of prediction deviation caused by model simplification hypothesis in a traditional method is solved, high-precision prediction of biomass gasification products is achieved, a theoretical basis is provided for real-time optimization of process parameters, and the method is suitable for industrial control of biomass gasification systems such as fluidized beds and fixed beds.

The invention discloses an SVR combustion instability prediction method and system, and relates to the field of energy power and intelligent control, and the method comprises the steps: obtaining a fault combustion flame chemiluminescence signal, and generating a signal data set comprising a training data set and a test data set; establishing a support vector regression model based on the training data set; evaluating the support vector regression model by using the test data set, and adjusting according to an evaluation result to obtain a prediction model; using the prediction model to predict to-be-predicted sample data; the historical flame chemiluminescence signal data of the gas turbine under different working conditions are adopted as data input and can be used for predicting the working state of the gas turbine after a certain time delay, and the model can perform algorithm processing on the flame chemiluminescence signal measured by the gas turbine and predict the change trend of the future moment. The occurrence time of the unstable combustion can be predicted, and enough time is reserved for the control process of the unstable combustion.

The invention relates to a cable insulation degradation degree evaluation method based on frequency domain characteristic analysis, and belongs to the field of cable insulation degradation detection.The cable insulation degradation degree evaluation method comprises the steps that cable operation environment data and laboratory accelerated aging data are collected, and a low-frequency impedance characteristic data set with water tree growth stage labels is generated; extracting a dielectric response spectrogram and analyzing a relaxation polarization peak value; when the peak value exceeds a threshold value, extracting harmonic component characteristics and loss factor distribution characteristics by using wavelet transform and principal component analysis; constructing a support vector regression model to predict the length, density and growth stage of the water tree; for middle and later water trees, microstructure evolution characteristic parameters are calculated, and insulation penetration time is predicted; the degradation probability is analyzed through Monte Carlo simulation, and a risk distribution curve is generated; and optimizing the cable replacement time by adopting dynamic planning, and outputting an evaluation report for quantifying the degradation degree and maintenance suggestions. According to the invention, accurate evaluation and prediction of cable insulation degradation are realized, and an important guarantee is provided for safe operation of a power system.

The present invention proposes a short-term wind power forecasting method based on the IJS‑SVR model. The steps are as follows: obtain a wind power generation related data set S, divide it into training data and test data, determine the input and output of the prediction model, and normalize the data; set the parameters of the improved artificial jellyfish (IJS) algorithm and the support vector regression (SVR) model; run the improved artificial jellyfish (IJS) algorithm to obtain the optimal penalty factor C and the optimal kernel function parameter g in the support vector regression (SVR) model; bring the optimized optimal parameter combination into the support vector regression (SVR) model, and train the support vector regression (IJS‑SVR) model optimized by the improved artificial jellyfish algorithm; input the test data into the IJS‑SVR model to obtain the prediction results, and denormalize the prediction results. The results show that in terms of MAE, MAPE, RMSE, R 2 Under the performance index, the prediction performance of the IJS-SVR model proposed in the present invention is better, which effectively makes up for the low prediction accuracy of the existing short-term wind power output power prediction method.

The invention provides a cattle and sheep feed trough residual feed monitoring method based on vision, and relates to the technical field of intelligent livestock breeding, and the method comprises the following six steps: intelligent triggering and multi-modal image acquisition, image preprocessing and fusion, time sequence image segmentation and feature extraction, density adaptive volume calculation, online learning and residual feed estimation, and decision analysis and early warning. According to the method, RGB and near-infrared images are synchronously collected through infrared triggering, after perspective correction and illumination adaptive fusion, an LSTM-U-Net time sequence segmentation model is adopted to solve the dynamic shielding problem, a residual feed area is accurately extracted, the feed type is recognized, density adaptive volume measurement is achieved in combination with monocular depth estimation and texture feature analysis, and the method is suitable for large-scale industrial production. Weight estimation is carried out through a support vector regression model, model parameters are optimized on line based on manual correction data, intelligent early warning of the residual material amount is finally achieved through a cloud platform, and the limitation of a traditional method in the aspects of shielding processing, density adaptation and environment anti-interference is effectively overcome.