936 results about "Model learning" patented technology

The invention discloses a wire drawing machine operation state evaluation method and system based on deep learning, and relates to the technical field of intelligent manufacturing, and the method comprises the steps: deploying a multi-source sensor at a preset part of a wire drawing machine, and forming a multi-dimensional operation data flow; the deep learning evaluation model is trained based on the standardized time series data set, model parameters are optimized through a cross entropy loss function, the model learns the characteristic difference between a normal working condition and an abnormal working condition, and the deep learning evaluation model is trained based on the standardized time series data set. And inputting a multi-dimensional operation data flow collected in real time into the trained deep learning evaluation model to generate a dynamic evaluation report. According to the wire drawing machine operation state evaluation method and system based on deep learning provided by the invention, a visual operation state score can be output, detailed fault type probability distribution is given, and intelligent support is provided for equipment maintenance decision.

The invention relates to the technical field of rehabilitation training detection, in particular to a rehabilitation training detection method and system based on artificial intelligence, a standard action library is constructed through standard action videos shot at multiple angles, and track and angle features of key joints are extracted for user training comparison; the skeleton key points of the user are extracted in real time through a MoveNet network, and efficient posture recognition in a home scene is achieved; analyzing position difference, angle change and acceleration characteristics by combining a space-time sequence matching algorithm, generating a dynamic matching degree index, and positioning a deviation joint to generate a correction prompt; introducing an attention mechanism model, learning the contribution degree of each joint to cycle recognition, dynamically selecting a dominant joint for action counting, recognizing starting and ending points of an action cycle through an acceleration curve, and finishing effective action statistics in combination with a dynamic threshold value, so that the counting accuracy and the self-adaptive capability are improved; therefore, the training cost is reduced, the evaluation credibility is enhanced, and accurate statistics and analysis of rehabilitation training data are realized.

A federated object detection learning method based on representation enhancement and weighted aggregation under cloud-edge-terminal environment comprises the steps of: 1) building a centralized federated learning framework under cloud-edge-terminal environment; 2) locally conducting representation enhancement training to strengthen model learning for few-shot category after receiving a model from the server at the client; 3) carrying out the weighted aggregation for client models in accordance with sample distribution to obtain the global model after receiving models from all clients at the server. With regard to the problem of existing federated object detection learning on low global model accuracy and weak generalization ability, the present invention can improve the accuracy and generalization ability of global object detection model.

The invention discloses a medical image recognition system based on label noise robust learning, and provides a hard sample label refinement strategy based on a confidence perception weighted prototype and an effective noise sample joint correction method to process divided subsets in correction before training so as to obtain training data with higher quality; in progressive hard sample reinforcement learning, data are input according to the learning difficulty of samples for training, the ability of model learning discrimination feature representation is improved by integrating cross entropy loss, consistency loss and credible contrast loss, and the influence of medical tag noise is reduced. According to the method, by integrating label refinement and a progressive hard sample reinforcement learning technology, the accuracy and robustness of medical image recognition under the condition of noise label data are effectively improved, overfitting of the model to noise samples is relieved, and the method can be used for correctly building a disease auxiliary diagnosis model under the condition that the noise label samples exist in training data.

The invention belongs to the technical field of pathological diagnosis, and discloses an artificial intelligence-based lung cancer pathological section analysis and system, which introduces identification basis visual information generated by an artificial intelligence model, and combines the correction operation of a plurality of pathologists on a model identification result and the acquisition of correction reasons, so that the pathological section analysis of the lung cancer is realized. Further performing consistency analysis on the correction results to generate uniform data, and finally performing incremental training on an artificial intelligence model by using the data, thereby constructing a closed-loop feedback optimization mechanism combining artificial intelligence and expert knowledge. The model performance problem caused by insufficient training data coverage, expert labeling efficiency and result inconsistency of an artificial intelligence model in pathological diagnosis is solved, and model learning and performance improvement are achieved.

The invention discloses a rapid heat transfer simulation method and device based on a neural network, and relates to the technical field of physical simulation. The method comprises the steps that a hybrid neural network model is trained, the model learns operator mapping from an input function to a temperature or heat flow field, and meanwhile physical constraints such as a heat conduction partial differential equation are coded into a loss function; for a new simulation task, single forward inference is carried out by using the operator mapping, and an initial prediction result is rapidly generated; then, according to physical constraints of coding, calculating a physical residual error of initial prediction, and when the residual error exceeds a preset threshold value, executing a small amount of optimization iteration by taking the prediction as an initial value to carry out rapid local correction; the problems that a traditional numerical method is long in calculation time and an existing neural network method is insufficient in physical fidelity are solved, and high efficiency and high precision of heat transfer simulation are achieved.

The invention belongs to the technical field of knowledge graph construction, and discloses a supply chain knowledge graph construction method based on time sequence dynamic perception and a large language model. Constructing a time sequence dynamic sensing model, and respectively generating a time-sensitive embedding matrix for the entity and relationship of each sub-graph in a historical time window; inputting the time-sensitive embedded matrix into an aggregator to further mine structure and semantic information of entities and relationships; establishing a dependency relationship of an autoregression model learning sub-graph in a time sequence, and generating a time sequence evolution representation; and inputting the time sequence evolution representation into a pre-trained large language model, generating candidate entities or candidate relationships to complement the fact tetrad, and updating the sub-graph sequence of the current timestamp. According to the method disclosed by the invention, the supply chain domain knowledge is adapted while the general semantic understanding capability is reserved, and the balance of dynamic evolution modeling, long-period dependency capture and efficient utilization of the domain knowledge is realized, so that the reliability and interpretability of a construction result are ensured.

The invention discloses a reward model training method, a large language model optimization method and correlation, and the reward model training method comprises the steps: obtaining a preference training sample pair and a to-be-trained reward model, the preference training sample pair comprising a preferred response sample and a non-preferred response sample; calculating an award score difference between the preferred response sample and the non-preferred response sample based on a to-be-trained award model; constructing a cost matrix based on the reward score difference and the semantic association degree between the preferred response sample and the non-preferred response sample; calculating a loss margin based on the cost matrix; and based on the loss margins, carrying out calculation to obtain paired preference loss values of the band margins, and updating parameters of the to-be-trained reward model by taking minimization of the loss values based on the band margins as an optimization target to obtain a trained reward model. The learning ability and overall generalization performance of the model for difficult samples are improved, excessive dependence on simple samples is avoided, and then the generation quality of the large language model in complex tasks is improved.

The invention relates to the technical field of deep counterfeit image detection, and discloses a deep counterfeit detection method and system, a storage medium and computer equipment. The method comprises the following steps: firstly, constructing a reference data set containing a forged image and an original real image; secondly, through an integrated model, generating antagonistic samples for the reference data set, and integrating the successfully attacked antagonistic samples into an antagonistic sample set; and finally, merging the reference data set and the adversarial sample set, and constructing a robustness enhanced data set containing four types of samples. In the model training stage, multi-classification cross entropy loss and comparative learning loss are combined, and expression of the model in a feature space is optimized through comparative learning constraint, so that the model learns discriminative features with more compact intra-class features and more dispersed inter-class features. The model trained by the method not only can effectively defend against attack and improve robustness, but also surpasses original detection performance on clean samples, and has remarkable technical advantages and application value.

Methods for training a series of neural networks to output driving behavior control parameters is disclosed. The training dataset for the neural networks includes sensor-based vehicle driving recordings that may be categorized by geographical area, by qualitative driving behaviors, or by some combination, such that various training data subsets are used to train the series of neural networks. By learning either city-specific driving behavior control parameters, qualitative driving behavior specific driving behavior control parameters, or both, the resulting parameters may then be provided to a motion planning model for use in modeling predictive control for an autonomous vehicle. Rather than relying on XYZ trajectories of agent vehicles when planning future trajectories of the ego vehicle, the motion planning model is adaptive, due to the use of the learned driving behavior control parameters.

The invention discloses an analysis method and system for AI recognition of physical exercise trajectory data. The method comprises the following steps: step 1, constructing a space-time-environment-biological feature fusion network, integrating collected multi-source data, capturing association of trajectory coordinates, skeleton joint angles and electromyographic signals, constructing a space-time dynamic graph, and comprehensively sensing the influence of a dynamic environment on a motion trajectory; 2, introducing a generative adversarial network to simulate extreme environment interference, and improving model robustness; forcing the model to learn a defensive decision through physical layer disturbance and behavior layer disturbance; distinguishing a real track from a generated track by using a discriminator, and trying to cheat the discriminator by a generator to enable the model to adapt to a noise environment; 3, cross-athlete skill migration is conducted, a parameter freezing strategy and dynamic weight adjustment are adopted, the basic exercise mode recognition capacity is reserved, and meanwhile individual differences are adapted; and step 4, in combination with deep reinforcement learning and an interpretable rule engine, carrying out instant tactical adjustment, and generating an action through a DRL strategy network.

The invention discloses an engine state estimation and system modeling correction method based on double-layer variational inference, which relates to the field of engine state estimation and comprises a variational inference stage aiming at component performance states and kinetic model parameters; a system output prediction stage based on an observation equation; and a solving stage of performing objective function optimization through an evidence lower bound. The structure clearly presents information flow and key calculation links of the proposed algorithm in state estimation and model learning. According to the method, combined reasoning of state variables and model parameters is achieved by building a probability modeling structure, the modeling problem when system dynamics is partially or completely unknown is solved by combining a modeling method of a stochastic differential equation, and while the state variables and the model parameters are optimized, the modeling efficiency is improved. Precise inference of component states and reliable identification of fault features are achieved, the fault detection accuracy of sudden gas circuit abnormity reaches the standard, and meanwhile the performance is better in the aspect of tracking long-term performance degradation.

The invention belongs to the technical field of artificial intelligence, and particularly relates to an exercise psychology multi-round dialogue data set construction method. In order to solve the problems that in the prior art, single-round question and answer data is simple in structure, multi-round dialogue construction lacks logicality and field adaptability, and a data organization form is not beneficial to model learning, the main scheme comprises the steps that original single-round question and answer data in the sports psychology field is collected and arranged, and an original single-round question and answer data set is formed; based on the problem type classification system, a splitting mode and a reconstruction rule are designed for each type of problems, and a problem splitting rule base is generated; after subject recognition and classification are carried out on the original question and answer data, a sub-question sequence is generated according to a splitting rule, corresponding sub-answers are reconstructed, and multi-round dialogue data is formed; and storing the multi-round dialogue data according to a structured format to generate a structured dialogue data set.

The invention relates to the technical field of intelligent wharfs, in particular to a wharf container truck dynamic optimization scheduling method and system combining machine learning and path planning. Comprising a behavior data acquisition and feature coupling unit; a learnable incentive and behavior guide unit; a scheduling demand prediction unit; and a path planning and scheduling unit. According to the method, on the basis of the coupling characteristics, the excitation coefficient is optimized through reinforcement learning, the excitation instruction is dynamically pushed, and targeted guidance of the non-operation staying behavior of the container truck is achieved; according to the method, based on standardized time series data, an association rule of a historical staying period and a working condition is learned through an LSTM model, a prediction result is optimized in combination with real-time data, a prospective constraint basis is provided for scheduling, and meanwhile, a time, space, resource and priority multi-dimensional path constraint system is constructed through a structural causal model; container truck-berth matching and dynamic path planning are completed by matching with an improved A * algorithm fused with dynamic weights, and scheduling conflicts are effectively avoided.

Devices and methods for adaptively classifying network traffic associated with a new application are provided. A network device, for example, an edge device, stores a Machine Learning (ML) model pre-trained based on historical network traffic associated with a set of applications. The network device receives network traffic associated with a new application, for example, a zero-day application, that is different from the set of applications. The ML model learns one or more patterns associated with the received network traffic. The ML model detects whether the learned pattern(s) is similar to previously learned patterns of at least one application. The ML model classifies the received network traffic as legitimate traffic or anomalous traffic based on the detection. The ML model is scalable, providing timely classifications for different types of network traffic, while handling protocol and application diversity, variability in traffic patterns, and emergence of zero-day application traffic.

The invention relates to the technical field of artificial intelligence, discloses a multi-modal self-supervision abnormal mode detection method and system, and aims to solve the problems that in the prior art, a large amount of annotated data is relied on, modal fusion is insufficient, the anomaly discrimination ability is weak, and the dynamic environment is difficult to adapt. The method comprises the following steps: synchronously acquiring videos, audios, sensor time sequences and log text data, and carrying out time alignment; extracting spatial-temporal characteristics of each modal through a modal specific encoder; constructing a contrast learning task under a label-free condition, generating positive and negative sample pairs by utilizing data enhancement, and driving model learning discriminative representation; and cross-modal feature alignment and dynamic weighted fusion are realized by adopting an attention mechanism, and joint representation is generated. According to the scheme, efficient anomaly detection without annotation data is realized, the multi-modal fusion representation capability is remarkably improved, the false alarm rate is reduced, the environmental adaptability is enhanced, and the real-time monitoring requirement is met.

The invention relates to the field of artificial intelligence security, in particular to a copyright protection scheme and system oriented to a split learning model and based on an adversarial sample. Aiming at the characteristics of separation and gradient interaction of a client and a server in a split model structure, a fingerprint sample is constructed by adopting an adversarial sample generation technology, so that the model can be induced to generate specific misclassification output. In a training stage, a fingerprint sample is mixed into data loading of a client in an extremely low proportion, and model learning is gradually guided to generate an expected response to the fingerprint sample in model training. The fingerprint embedding mode can effectively verify the copyright of the model on the premise of not influencing the normal classification performance of the model. In addition, the method has high robustness, and common attack means such as model pruning and label reasoning can be resisted. In conclusion, the invention provides a copyright protection scheme suitable for the split learning model, the blank of the copyright protection technology in the split learning field is filled, and a technical means is provided for verification of the split learning model copyright.

The invention discloses an angle steel connecting piece shear strength prediction method and system based on a physical information neural network, electronic equipment and a storage medium thereof. The method comprises the following steps: establishing a database containing a plurality of groups of test data based on a numerical simulation result of an angle steel connecting piece finite element model verified by a push-out test; embedding an angle steel connecting piece shear bearing capacity physical constraint condition based on an elastic foundation beam theory into a loss function of the data-driven neural network model DNN; training the neural network embedded with the physical constraint condition by using a database, and adjusting and selecting a physical item weight factor and a model learning rate so as to construct a shear bearing capacity prediction model of a physical information neural network (PINN); predicting the shear strength of the angle steel connecting piece by using the trained physical information neural network (PINN) shear capacity prediction model; according to the prediction method and system, the electronic equipment and the storage medium thereof provided by the invention, the accuracy and reliability of the shear resistance prediction of the angle steel connecting piece can be improved.

The invention belongs to the field of artificial intelligence and natural language processing, and discloses a cross-topic rumor detection method based on a time perception attention mechanism, and the method comprises the steps: constructing an input sample containing a text, a timestamp and a topic label based on multi-source social media data, and carrying out the semantic coding of the text through a pre-training language model; a topic attention enhancement mechanism is introduced, semantic differences of samples in different topics are captured, and the perception ability of a topic structure is improved; a time perception attention mechanism is introduced, and the robustness of time sequence characteristics is enhanced by modeling a sample time interval and dynamically adjusting an information weight; dividing positive and negative sample pairs by using topic tags, and guiding the model to learn topic-independent discriminative representation by comparing a loss function; and performing rumor judgment on the features fused with the multi-dimensional information. The method has good cross-topic migration ability and time-sensitive modeling ability, and can significantly improve the detection performance in the early propagation stage of unknown topics.

The invention provides a cross-domain binocular stereo matching method based on multi-scale information dynamic fusion and feature deviation correction, and belongs to the field of computer vision. Specifically, the model adaptively fuses a convolutional neural network (CNN) and a Transform structure according to dynamic change of a scene so as to capture local detail information and global context dependence at the same time. Secondly, through a multi-stage progressive cost body fusion and aggregation mechanism, full fusion of multi-scale matching cost bodies is promoted, redundant information is effectively inhibited, and matching precision is improved; and finally, guiding the model to learn more stable domain invariant representation by gradually reducing the deviation between different domain features, thereby remarkably enhancing the cross-domain robustness and adaptability. Compared with the most advanced method for performing cross-domain experiments in different scenes, the method disclosed by the invention shows better performance compared with an existing advanced model: 1) the cross-domain stereo matching precision is remarkably improved in a complex scene, and stronger domain migration capability and generalization stability are shown; and 2) while the cross-domain robustness is maintained, relatively high accuracy and calculation efficiency are maintained in a non-cross-domain environment, and double consideration of cross-domain and non-cross-domain performance is realized.

The invention discloses an enterprise appeal intelligent perception and closed-loop processing system based on multi-modal AI, and relates to the technical field of intelligent government affairs, the system realizes accurate semantic understanding and deep intention recognition of multi-modal fusion, and the intelligent perception and closed-loop processing of enterprise appeals are realized through a cross-modal attention mechanism and a unified semantic representation model. According to the method, deep fusion and complementary analysis of multi-source heterogeneous data such as voices, texts, images and the like are realized, the limitation of a traditional single-mode or simple splicing mode is broken through, the accuracy and robustness of intention recognition are remarkably improved, and misjudgment is fundamentally reduced; the method comprises the following steps: constructing a dynamic self-adaptive routing mechanism based on Actor-Critic reinforcement learning, constructing a work order assignment problem into a sequence decision problem, driving a model to learn a dynamic fusion and weight distribution strategy of multiple decision factors through a reward mechanism, and adopting an online strategy iteration optimization mechanism to realize an optimal assignment decision, so as to improve the work order assignment efficiency. And the shunting accuracy and efficiency are obviously improved.

The invention discloses an agricultural machinery intelligent path planning system and method based on multi-sensor fusion, and relates to the technical field of agricultural machinery automation control. Comprising an AI large model learning module, a visual identification module, a data analysis module and an intelligent control module. The AI large model learning module learns an operation instruction of the agricultural implement controller; the visual identification module collects and extracts agricultural implement controller panel feature information; the data analysis module analyzes, diagnoses and analyzes the feature information; and the intelligent control module transmits the analyzed data to an intelligent driving system of the tractor and executes corresponding control actions. The method solves the problems that the agricultural implements cannot be subjected to associated control under different communication protocols, and the existing scheme is high in cost, complex, low in efficiency and the like, has the advantages of high compatibility, low cost, high efficiency, high adaptability, high safety and the like, and can be widely applied to the field of modern agriculture.

The invention provides an agricultural image background removal method, and relates to the technical field of agricultural image processing, and the method comprises the steps: carrying out the adaptive illumination condition judgment of a preprocessed agricultural image; calculating and selecting corresponding vegetation indexes according to the light condition judgment result, and calculating to generate a plurality of vegetation index feature maps; inputting the plurality of vegetation index feature maps into a machine learning model, learning the relative importance of each vegetation index through the machine learning model, and generating a comprehensive vegetation feature map; performing vegetation texture region judgment on the preprocessed agricultural image to obtain a target vegetation texture region judgment result; judging a foreground mask based on the comprehensive vegetation feature map and a target vegetation texture region judgment result; performing morphological processing on the foreground mask to obtain an optimized foreground mask; and according to the optimized foreground mask, removing a background from the agricultural image in the original RGB format to obtain a target vegetation image.

The invention discloses an audio deep forgery detection method and device, a terminal and a storage medium, and relates to the technical field of multimedia information security and artificial intelligence, and the method comprises the steps: constructing an audio deep forgery detection network model; according to the training set, performing decoupling stage training on an audio deep counterfeiting detection network model, and determining an initial training network model; performing meta-learning on the initial training network model according to the training set, and determining a target training network model; and obtaining a to-be-detected audio, inputting the to-be-detected audio into the target training network model, and determining an audio category corresponding to the to-be-detected audio. According to the method, meta-learning is adopted in the training stage of the audio deep forgery detection network model to force the audio deep forgery detection network model to learn general knowledge with higher generalization, so that the problem of poor detection performance caused by model overfitting in the face of an unknown vocoder and a natural scene in the prior art can be effectively solved.

The invention provides a multi-modal emotion recognition method and system based on knowledge distillation. The method comprises a pre-training stage, a knowledge distillation stage, a fine adjustment stage and a prediction stage. The pre-training stage comprises the following steps: independently training a preset neural network by using electroencephalogram, electrocardiogram and facial expression data to obtain three independent teacher models; the knowledge distillation stage comprises the step of migrating the representation learned by the teacher model to the student model through a knowledge distillation technology; the fine tuning stage comprises the step of carrying out joint optimization on the student model by utilizing a small amount of labeled multi-modal data; the prediction stage comprises the steps of dynamically fusing the multi-modal features by adopting an attention mechanism, and inputting the fused features into a classifier to obtain an emotion recognition result. The method is oriented to three modes of electroencephalogram, electrocardio and facial expression, and the characterization ability of the model to complex emotion and state change is improved; and a knowledge distillation mechanism is introduced, so that the complexity of the model is reduced, and the deployment feasibility and the operation efficiency of the model in practical application are improved.

The invention belongs to the technical field of remote sensing image segmentation, and particularly relates to a double-branch water body small target image segmentation method based on feature efficient interaction, which comprises the following steps: preparing a data set, constructing a network model, training the network model, selecting a proper loss function and evaluation index, and determining a segmentation model. According to the multi-scale detail feature interactive aggregation encoder, efficient fusion of multi-scale features and effective supplement of implicit relative position encoding information are achieved; the long-distance feature efficient capture encoder increases the edge segmentation effect on the small target image of the water body by processing features along a specific space direction; the cavity space convolution pyramid module based on the large convolution kernel is used for improving the ability of the model to learn a large-scale effective receptive field; the non-significant feature extraction module is used for filling up the deficiency of non-significant features. The whole network adopts a learning strategy of parallel connection of global and local features and series connection from a large scale to a small scale, and effectively extracts different scale information of the water body image.

The invention relates to the technical field of human activity recognition, and particularly discloses a human activity recognition method and system based on a lightweight hybrid neural network and self-attention migration and a medium thereof, and the method comprises the steps: 1, employing an HHAR data set, and carrying out the preprocessing; step 2, constructing a teacher model based on a CNN-LSTM-Transform hybrid architecture, and constructing a teacher model based on the CNN-LSTM-Transform hybrid architecture; step 3, constructing a student model of a hybrid architecture based on CNN-LSTM-Transform; and step 4, utilizing a self-attention migration mechanism to guide the student model to learn the attention distribution mode of the teacher model, and realizing efficient migration of knowledge. According to the method, the calculation complexity and the storage requirement can be remarkably reduced while the expression ability of the model is maintained.

The invention relates to the technical field of power equipment operation and maintenance, in particular to a fault diagnosis method and device for an electricity utilization information acquisition terminal. The method comprises the following steps: establishing an anomaly model according to an anomaly cause of an electricity utilization information acquisition terminal; obtaining a to-be-trained sample set according to the real-time power consumption data and environmental data collected by the power consumption information collection terminal and the abnormal event label of the abnormal model; deploying an initial teacher model on the cloud platform, and training the initial teacher model to obtain a teacher model; training the initial student model to learn fault diagnosis knowledge by adopting a knowledge distillation mode and a to-be-trained sample set according to the teacher model, and deploying the trained student model to an edge computing node corresponding to the electric information acquisition terminal; and fault diagnosis is carried out on data acquired by the electricity utilization information acquisition terminal by adopting a student model. According to the method, the problems of high model complexity and limited computing resources caused by performing fault diagnosis on the power utilization information acquisition terminal by adopting edge computing in the prior art can be solved.

The invention discloses a multi-variable measurement sensor state lightweight evaluation method based on knowledge distillation, and belongs to the technical field of electric digital data processing and multi-sensor data fusion. The method comprises the following steps: firstly, carrying out time synchronization and physical consistency constraint modeling on original data of multiple sensors, and extracting feature representation; a high-precision teacher model is trained at the cloud end, and the high-dimensional mapping relation of the sensor state is learned; intermediate features, soft output and uncertainty information of a teacher model are extracted to serve as distillation knowledge, lightweight student model training is guided, and effective migration of discrimination knowledge is achieved in combination with soft label constraint, feature alignment and an uncertainty guiding mechanism; and finally, compressing, quantifying and optimizing the student model, and deploying the student model to a vehicle-mounted end to realize real-time evaluation and dynamic updating of the states of multiple sensors. According to the method, the model complexity is greatly reduced while the evaluation precision is ensured through a knowledge distillation framework, and efficient and reliable state perception and fault-tolerant control support is provided for an intelligent driving system.

The invention relates to the technical field of image classification, and discloses a tongue picture classification method, a tongue picture classification model acquisition method and electronic equipment, and the tongue picture classification model acquisition method comprises the steps: inputting a preprocessed unlabeled tongue picture image into a neural network model for self-supervised learning; masking the unlabeled tongue picture image according to the reconstruction mask loss to obtain a pre-trained neural network model; inputting the labeled tongue picture image into a pre-trained neural network model for supervised learning; guiding supervised learning of a pre-trained neural network model based on a centroid classifier to obtain a trained neural network model; and testing the trained neural network model, and taking the neural network model meeting a test condition as a tongue picture classification model. According to the tongue picture classification model acquisition method, the unlabeled tongue picture image and the labeled tongue picture image are fully utilized to complete knowledge-assisted model learning, and the accuracy of small sample learning in a tongue picture classification task is improved.