47results about How to "Improve migration ability" patented technology

The invention discloses a voiceprint recognition system attack resisting method and device, equipment and a medium, by constructing a shadow voice sample set and a substitution model, dependence on target speaker data and query capability is reduced, and privacy risk and query overhead are reduced; meanwhile, two types of substitution models are trained based on an information theory method, so that the mobility and stability of the models are improved, and the generalization ability is enhanced; besides, the attacker model adopts a specific architecture and combines an alternate training strategy, so that attack loss and prediction probability difference are optimized, imperceptibility and attack effectiveness are balanced, and tone quality reduction or attack failure is avoided; and finally, through systematic training and optimization, a more uniform evaluation caliber is expected to be provided, and more reliable guidance is provided for engineering landing and risk evaluation.

The invention relates to the technical field of clinical medical treatment, and discloses a disease deterioration risk identification method and system based on depth time sequence clustering enhancement, and the method comprises the steps: 1, obtaining multi-modal clinical sequence data of a patient, including physiological indexes of a time sequence, a laboratory detection result, historical diseases and medication data; 2, performing feature extraction and classification on the patient sequences by adopting a knowledge enhanced sequence clustering method, and grouping the patient sequences according to future outcome distribution of the patient sequences; 3, enabling the model to quickly adapt to a prediction task of a new patient subgroup through a meta-training process; and step 4, based on the trained meta-model, carrying out rapid adaptation on the new patient subtype, and predicting the possibility that the new patient subtype has a deterioration event in a certain time window in the future. The method and the system can effectively learn the disease change mode of the patient under the condition of limited clinical data, improve the prediction accuracy of the new patient subgroup, and are especially suitable for clinical prediction scenes under the condition of small samples.

This invention relates to an air electrode material and its preparation method. The preparation method includes: using metal salts of Ce, Pr, La, Ba, and Sr, metal salts of Co, and a fluorine source as raw materials, and preparing a fluorine-containing perovskite precursor according to a stoichiometric ratio; pretreating the precursor and then sintering it to obtain a composite powder; the composite powder includes a fluorine-containing perovskite main phase and a fluorine-containing CeO2 nanophase; pulverizing and sieving the composite powder to obtain air electrode powder; the fluorine-containing perovskite main phase is an A-site high-entropy perovskite oxide; the fluorine-containing CeO2 nanophase is precipitated through the self-assembly of some Ce elements during sintering; the fluorine-containing CeO2 nanophase is dispersed inside the main phase grains or at grain boundaries, and forms a strongly coupled heterogeneous interface with the main phase. This invention solves the technical problems of poor surface exchange activity, bulk diffusion transport, and long-term stability of existing air electrodes.

The invention discloses a vehicle confrontation camouflage generation method and system for DE-YOLO cross-modal detection, and the method comprises the steps: employing a three-dimensional neural network renderer to initialize a full-coverage texture as an attack patch, covering the surface of a target object, and preliminarily simulating a basic attack scene; simulating a multi-factor interference scene in combination with a diffusion generation model, and generating target vehicle pictures in different scenes; forming an input batch by using the generated target vehicle pictures in different scenes, and inputting the input batch into a model DE-YOLO cross-modal detector to obtain three different-scale infrared feature maps and visible light feature maps in different scenes; and fusing the feature patterns of different scales into a cross-modal loss item, and performing back propagation optimization to generate infrared and visible light cross-modal full-coverage confrontation textures which are attached to the surface of the vehicle. According to the method, the neural network renderer is adopted to generate the full-coverage texture to serve as the attack patch to cover the surface of the target object, the target detection system is attacked, and the attack failure problem under multi-factor interference is solved.

The invention discloses an active polypeptide for promoting healing of oral ulcer and application thereof, and belongs to the technical field of biological medicines. The amino acid sequence of the active polypeptide NPTX3 capable of promoting healing of the oral ulcer is DLKPIAEDIPSLEK. An in-vitro experiment result shows that the polypeptide NPTX3 can remarkably promote the scratch closure of the human immortalized keratinocytes at the concentration as low as 500 pM. Transwell migration experiments further prove that the polypeptide NPTX3 can obviously enhance the migration ability of HaCaT cells at the concentrations of 1nM and 10nM, which prompts that the polypeptide NPTX3 may accelerate wound repair by promoting epithelial cell migration. In an SD (Sprague Dawley) rat oral ulcer model, the ulcer healing rate can be obviously improved by locally applying the polypeptide NPTX3 (1 nM, 10 nM). The polypeptide NPTX3 provided by the invention shows excellent activity of promoting cell migration and tissue repair at an extremely low dose, has the advantages of low effective dose, strong biological activity, high safety and the like, and provides an efficient and mild new choice for drug treatment of oral ulcer.

The application discloses a kind of low-carbon cementing material by waste incineration fly ash excited ore powder and preparation method thereof, by weight percentage, including the following components: modified waste incineration fly ash 8~12%, ion complex component 1~3%, composite shrinkage component 2~6%, ore powder 55~65%, water 23~28%.Preparation method includes the following steps: S1, preparation modified waste incineration fly ash, ion complex component, composite shrinkage component;S2, modified waste incineration fly ash, ion complex component, composite shrinkage component, ore powder, mixing water, fully mixed and stirred evenly, can obtain the low-carbon cementing material by waste incineration fly ash excited ore powder described in the application.In the aspect of component design, the application is environmentally friendly, low cost, in the performance aspect, the flow rate can be adjusted, the mechanical properties are excellent, the volume shrinkage is small.

The application discloses a SnTe-based thermoelectric material and a preparation method thereof, and belongs to the technical field of thermoelectricity. The application adopts high-purity metal elements as raw materials, and the raw materials are mixed according to a stoichiometric ratio, and then a cast ingot is obtained through a melting reaction after vacuum packaging. Subsequently, the cast ingot is subjected to high-temperature annealing treatment. After the cast ingot subjected to the high-temperature annealing treatment is ground into powder, a dense block-shaped SnTe-based thermoelectric material is formed through discharge plasma sintering. The chemical formula of the SnTe-based thermoelectric material is Sn 2.9‑X W X Te 2.9 -(In2Te3) 0.03 -(Cu2Te) 0.18 The application significantly improves the thermoelectric figure of merit of the SnTe-based thermoelectric material and improves the thermoelectric performance of the SnTe-based thermoelectric material by doping tungsten elements in the SnTe matrix and compounding the two metal compounds of In2Te3 and Cu2Te.

The invention discloses a wind turbine generator gearbox state monitoring and fault early warning method and system. The method comprises the following steps: defining a key part of a gearbox of a wind turbine generator as a flexible body, and constructing a multi-body dynamic model based on the flexible body; the multi-body dynamics model receives input multi-source time sequence data, then performs feature extraction and multi-index fault early warning algorithm processing, and outputs signal abnormal information for the current multi-source time sequence data; the signal abnormal information is gearbox fault data; constructing a mixed data set guided by a physical mechanism and performing data enhancement; and inputting a spatial-temporal feature coding fault diagnosis model based on the enhanced mixed data set, and then outputting a health index. According to the intelligent early warning system for the wind turbine generator gearbox, through the technical innovation of making up for data missing through modeling simulation, improving diagnosis precision through an intelligent algorithm and guaranteeing real-time early warning through edge calculation, a set of intelligent early warning system for the wind turbine generator gearbox which is high in precision, low in time delay and strong in generalization is successfully constructed.

The application belongs to the field of treating blood vessel-bone coupling imbalance diseases in traditional Chinese medicine, and particularly relates to a composition for treating blood vessel-bone coupling imbalance diseases, a pharmaceutical preparation and application. Geniposide alone or in combination with crocin I can effectively reduce endothelial cell damage under pathological stimulation conditions, reduce the level of cell apoptosis and the degree of oxidative stress, improve the vascular endothelial function state, and can also significantly inhibit osteoclast differentiation, reduce the number of mature osteoclasts, and reduce bone resorption activity, thereby being conducive to restoring the dynamic balance between bone formation and bone resorption.

The invention relates to a hafnium zirconium oxide all-component ferroelectric characteristic and fatigue optimization method based on machine learning potential. The construction method comprises the following steps: step 1, acquiring atomic configuration, energy, stress and electric dipole moment data of HfxZr1-xO2 under different components x (0 < = x < = 1) as an original data set; 2, training a multi-task machine learning potential model by using a deep learning algorithm, wherein the model comprises an energy-force prediction branch and an electric dipole moment prediction branch; 3, performing molecular dynamics evolution under different temperatures and external electric fields to obtain a change curve of the polarization intensity along with the cycle period; and step 4, screening an optimal material component according to the nonlinear corresponding relationship between the fatigue rate and the Hf content obtained by simulation. Compared with the prior art, the method has the advantages that the training data set covering all-component space is constructed, the machine learning potential model capable of synchronously describing energy, force and electric dipole moment is obtained through deep neural network training, and the regulation rule of the Hf doping proportion on the crystal phase stability is disclosed through large-scale molecular dynamics simulation; and the relationship between the domain wall movement rate and the fatigue rate is quantified, and theoretical guidance of atomic scale is provided for component design of the FeRAM device.

The application belongs to the technical field of continuous casting and particularly relates to a high-nitrogen Fe-Ni-based alloy protective slag. According to mass percentage, the high-nitrogen Fe-Ni-based alloy protective slag comprises SrO: 14-16%, CaF2: 11.5-13.5%, SiO2: 24-26%, Na2O+K2O: 12-14%, B2O3: 7-9%, Al2O3: <1%, MgO: 8-10%, La2O3: 10-12%, and the rest is carbonaceous material and inevitable impurities. Through accurate slag system design and component regulation, the protective slag has excellent performances of rapid slag formation, high lubrication and phased heat control, effectively solves the crack problem in the high-nitrogen Fe-Ni-based alloy continuous casting process, and meets the actual needs of the large-scale production of the continuous casting.

The invention discloses a biomarker for diagnosing pulmonary fibrosis and application of the biomarker, and belongs to the technical field of biomarkers in clinical medicine. The biological marker for preparing and diagnosing pulmonary fibrosis is a nerve growth factor NGF, the nerve growth factor NGF can remarkably promote differentiation of lung fibroblast MRC5 cells, and along with expression increase of a receptor TrkA of the lung fibroblast MRC5 cells, the migration capacity and the multiplication capacity of the lung fibroblast MRC5 cells are remarkably enhanced. Meanwhile, the NGF also aggravates the formation of the BLM-induced pulmonary fibrosis of the mouse. Therefore, the nerve growth factor NGF as the biomarker has accurate and specific detection and auxiliary diagnosis on pulmonary fibrosis, especially idiopathic pulmonary fibrosis (IPF).

This invention proposes a transferable adversarial attack method based on generative adversarial networks (GANs). This method introduces an auxiliary generator into the generator to improve generation quality and sample diversity, and uses a pyramid attention mechanism to focus on features at different scales. Before the generator network, auxiliary information is first generated by the auxiliary generator network. This generated auxiliary information is then concatenated with the original image, and the concatenated input image is fed into the generator network. To further improve the performance of the generator network, a pyramid attention module is introduced, allowing the network to focus on important regions and features of the image at different scales. Through a series of neural network layers and transformations, a high-resolution, realistic image containing perturbations is generated. Such an image has a higher success rate in black-box adversarial attacks and increases the diversity and quality of adversarial samples. Compared to existing black-box attack methods, the adversarial samples generated by this scheme are more realistic and effective, have a higher success rate, and possess a certain degree of transferability. This invention provides an effective solution for adversarial attacks in black-box scenarios.

The application relates to a high-strontium mineral water preparation method for improving strontium absorption and application of the same. The method comprises raw water input, pretreatment and purification, two-stage nanofiltration grading purification, strontium concentrate liquid and low-mineral dialysis, constant-temperature targeted chelation reaction, chelated strontium mother liquor, sterile blending, calcium-strontium ratio precise control, improvement of strontium absorption and utilization, double strategies of calcium-strontium ratio lock control and ion removal, solution of competition of calcium and magnesium ions on strontium absorption from the root, mild reaction in the whole process, no high temperature and high pressure, no introduction of harmful by-products, all raw materials meeting national food safety standards, no taste difference, no peculiar smell and no astringency.

The application discloses a frequency domain guidance-based multi-feature attention adversarial sample generation method and system. In the method, the target recognition loss of the initial adversarial sample of each round is calculated in the iterative generation of the final adversarial sample; the multi-scale features of the initial adversarial sample of the round are extracted, and the multi-scale feature gradient is calculated; the spatial domain feature weight is calculated and determined according to the target recognition loss and the multi-scale feature gradient; the spatial domain feature weight is subjected to frequency domain transformation and band-stop filtering processing to obtain the filtered weight; the filtered weight is reconstructed back to the spatial domain to obtain the frequency domain guidance weight; the frequency domain guidance weight and the multi-scale feature gradient are point-multiplication fused to obtain the fusion gradient; the initial adversarial sample of the next round is obtained by updating the initial adversarial sample of the round through the fusion gradient, and the initial adversarial sample of the next round is obtained. The generated adversarial sample has good cross-model migration.

The invention provides a lithium-rich manganese-based positive electrode material and a preparation method thereof, a positive plate and a battery. The lithium-rich manganese-based positive electrode material comprises a lithium-rich manganese-based core material and a coating layer coating the lithium-rich manganese-based core material, the alkali metal layer of the lithium-rich manganese-based core material at least comprises Mg element, and the coating layer comprises lithium fluoride. The lithium-rich manganese-based positive electrode material disclosed by the invention has high capacity, high structural stability and excellent interface durability, and can meet the performance requirements of a high-energy-density battery under complex working conditions.

The invention provides an anomaly detection method and device for graph data. The detection method comprises the steps that graph data of a target graph are obtained, the graph data comprise n nodes and the connection relation between the nodes, and each node has first feature data composed of d attributes; and partial nodes in the n nodes have labels representing the abnormal degrees thereof. And then determining second feature data, related to the graph structure, of each node, including processing the first feature data of each node by using a group of band-pass filters determined according to the connection relationship to obtain high-frequency abnormal features. The table data is constructed, and the single row data records the comprehensive feature data of the single node and comprises the first feature data and the second feature data. Then, the table data and the labels of the partial nodes are input into the table basic model, and the abnormal degree of the other nodes without the labels is predicted.

Embodiments of the present application provide an image recognition method and device based on self-supervised learning, equipment and medium, relating to the technical field of image processing. The method comprises: inputting a to-be-detected monitoring image into a pre-trained self-supervised visual model, obtaining attention weight information output by the self-supervised visual model in processing the to-be-detected monitoring image; determining a semantic salient region of the to-be-detected monitoring image based on the attention weight information; extracting an image feature vector corresponding to the semantic salient region using the self-supervised visual model; and performing an object recognition task based on the image feature vector. Embodiments of the present application can effectively improve the recognition ability of unknown objects and improve the migration performance and generalization ability of the model by using the self-attention mechanism of the self-supervised visual model to adaptively determine the semantic salient region of the image and extract the image features of the semantic salient region for object recognition.

The invention relates to the technical field of tissue engineering, and particularly discloses a pilose antler tissue cell culture method which comprises the following steps: S1, cell in-vitro culture; s2, directional induced differentiation; step S3, ossification induction; the induction culture medium 1, the induction culture medium 2 and the ossification induction culture medium are added, amphotericin B capable of covering the fungal pollution risk is added into the induction culture medium 1, fetal calf serum contained in the induction culture medium 1 can neutralize trypsin, recombinant human HGF contained in the induction culture medium 2 can improve the migration ability of cells, and meanwhile, the ossification induction culture medium can be used for preparing the ossification culture medium. The recombinant human bFGF can synergistically promote proliferation, so that emigrated cells are quickly accumulated, and the situation that the cell density is too low due to insufficient proliferation after emigration is avoided; ascorbic acid-2-phosphate contained in the ossification induction culture medium can avoid oxidation failure of ascorbic acid, and the cell emigration efficiency is improved through the synergistic effect of the substances.

The invention discloses a database migration method and device and electronic equipment, and relates to the technical field of database migration, and the method comprises the steps that under the condition that a target system needs database migration, a first module in the target system is recognized; under the condition that the first module exists in the target system, updating the first module based on a target transaction isolation function supported by the target database; under the condition that the first module does not exist in the target system, identifying a second module in the target system; under the condition that the second module exists in the target system, updating the second module based on a preset mark; after updating of the first module / the second module is completed, database migration operation is executed on the target system. According to the database migration method and device, the technical problems that in the database migration process, due to the fact that the function supported by the source database and the function supported by the target database are different, the system function is abnormal after migration, and the performance is reduced are solved.

This invention discloses a sulfur-containing cathode material and its preparation method, belonging to the field of lithium-sulfur battery technology. The method uses a quaternary ammonium polymer containing nitrate as a cathode binder, dispersed with a conductive agent and a sulfur-containing cathode material in deionized water to obtain a cathode slurry. The slurry is then coated onto aluminum foil and vacuum dried to obtain a cathode sheet. The quaternary ammonium cation backbone in the binder can adsorb polysulfides, improving lithium-ion migration capacity, while simultaneously preventing the rapid and uncontrollable consumption of nitrate, thereby extending the lifespan of the lithium-sulfur battery.

The invention discloses an audio-visual navigation method combining explicit acoustic geometric reasoning and auxiliary supervision, and belongs to the field of intelligent and robot autonomous navigation. The invention aims to solve the problems of weak geometric perception capability and poor generalization of the existing audio-visual navigation method in a perception limited and unknown environment. The method comprises the following steps: acquiring a binaural spectrogram and a visual image of an intelligent agent and extracting features; inputting the auditory features into an explicit geometric reasoning module (AGR), and outputting a high-dimensional geometric representation vector for decision making and a sound source distance and azimuth angle prediction value for auxiliary supervision in parallel through a shared coding and multi-head decoupling structure; and fusing the geometric representation vector and the visual feature, and inputting into a recurrent neural network to generate a navigation action. In a training stage, real geometric information of a simulation environment is utilized to construct auxiliary regression loss, and a dynamic weighting mechanism based on uncertainty is introduced to adaptively balance weights of a navigation task and a geometric auxiliary task. According to the method, the agent is endowed with the explicit perception ability of sound listening and position distinguishing, the model is forced to learn the acoustic physical law, and the navigation robustness and generalization ability in a visual blind area and an unknown environment are remarkably improved.

This application provides a method for preparing a positive electrode sheet, a positive electrode sheet, and a lithium-ion battery. The method for preparing the positive electrode sheet includes lithium intercalation treatment of a niobium-based material to obtain a lithium-containing niobium-based composite. The lithium-containing niobium-based composite, a positive electrode active material, a conductive agent, and a binder are mixed to obtain a slurry. The slurry is then formed on the surface of a current collector to prepare the positive electrode sheet. This application first performs lithium intercalation treatment on the niobium-based material. When lithium ions are intercalated into the crystal lattice of the niobium-based material, not only is electron transfer involved, causing a change in the valence state of the niobium ions in the niobium-based material, but the crystal structure of the niobium-based material is also altered, creating effective electron transport paths between crystals and enhancing electron migration capabilities. By mixing the lithium-containing niobium-based composite obtained after lithium intercalation with the positive electrode active material, conductive agent, and binder, the excellent ionic and electronic conductivity of the lithium-containing niobium-based composite can improve fast charging efficiency.

The application discloses a music style migration method, which comprises the following steps: obtaining to-be-combined music and to-be-migrated music; inputting the to-be-combined music and the to-be-migrated music into a preset style decoupling model respectively to obtain a first pitch content code, a second pitch content code and a rhythm code; inputting the first pitch content code and the rhythm code into a first decoder to obtain migrated music; inputting the second pitch content code and the rhythm code into a second decoder to obtain reconstructed music; detecting the migrated music and the reconstructed music by using a discriminator to obtain a style migration detection result; and determining the migrated music as target music when the style migration detection result represents successful migration. The preset style decoupling model is used for completely decoupling music, thereby avoiding mutual mixing between pitch content and rhythm style. The discriminator is used for detecting the migrated music and the reconstructed music, thereby ensuring successful rhythm style migration and improving the effect of music style migration.

This application relates to a method and apparatus for generating adversarial examples based on an artificial intelligence image generation model. The method includes: acquiring a pre-constructed adversarial training dataset; the adversarial training dataset includes original image samples and their corresponding text labels, and initial adversarial examples generated based on the original image samples; constructing a constraint function based on a pre-constructed semantic matching loss function, the initial adversarial examples, and the semantic mask of the original image samples; the semantic matching loss function is constructed based on the original image samples and their text labels; adjusting the weights of target neurons in each layer of the artificial intelligence image generation model based on the constraint function to obtain an adjusted artificial intelligence image generation model; and using the adjusted artificial intelligence image generation model to generate target adversarial examples corresponding to the original image samples. This solves the problem of low adversarial example transferability and improves adversarial example transferability, thereby enabling adaptation to black-box model scenarios.

This invention belongs to the field of artificial intelligence algorithm technology and discloses an intelligent transfer adversarial method and system for electromagnetic signal recognition under incomplete information. First, multiple models of alternative target networks with incomplete information are constructed in parallel, and these alternative models are trained using an evolutionary strategy based on a prior knowledge matrix. Then, an ensemble network model is generated based on the trained alternative models, and a fully connected layer and a layer with feature-level weighted fusion are designed. Finally, an optimized adversarial method for electromagnetic spatial signal recognition tasks is designed, generating adversarial samples under the ensemble model, and transferring these adversarial samples to the target network to achieve adversarial capabilities against the target network. This invention can effectively achieve adversarial capabilities against the target network when the target network information is unknown, and it still exhibits good transfer adversarial performance under the same signal recognition task and different unknown target network conditions.

The invention relates to a transferable adversarial sample generation method and device based on wavelet transform and multi-scale filling, and a storage medium, and the method comprises the following steps: obtaining an input image, converting the input image from a spatial domain to a frequency domain through wavelet transform, and carrying out the random block disruption of a high-frequency component block, and obtaining a preprocessed image; mI-FGSM-based sample updating is performed on the preprocessed image, and in the updating process, a global random multi-scale filling method is used, MI-FGSM input is transformed from multiple angles of a space domain and a frequency domain, a plurality of substitution models are combined, combined gradient loss is calculated, and training is performed to obtain an adversarial sample. Compared with the prior art, the method has the advantages of remarkably improving the migration capability of the adversarial sample among the black box models, improving the attack success rate and the like.

The invention discloses an active polypeptide NP-VV-3 capable of promoting healing of oral ulcer and application of the active polypeptide NP-VV-3, and belongs to the technical field of biological medicines. The amino acid sequence of the active polypeptide NP-VV-3 capable of promoting healing of the oral ulcer is VVPAFVLLKKAIRIMFKRNC. The amino acid sequence of the active polypeptide NP-VV-3 capable of promoting healing of the oral ulcer is shown in the specification. In an in-vitro experiment, the polypeptide NP-VV-3 can remarkably promote scratch closure of human immortalized keratinocytes at the concentration as low as 1 nM, and shows dose-dependent healing promotion capability at the concentrations of 1 nM and 10 nM; transwell migration experiments further prove that the polypeptide NP-VV-3 can significantly enhance the migration ability of HaCaT cells at the concentrations of 1 nM and 10 nM, indicating that the polypeptide NP-VV-3 can accelerate wound repair by promoting epithelial cell migration; besides, in an SD rat oral ulcer model, local administration of the polypeptide NP-VV-3 (1 nM, 10 nM and 100 nM) can significantly improve the healing rate of ulcer wounds, and presents a concentration-related trend. The polypeptide NP-VV-3 provided by the invention provides an important candidate molecule for developing a novel, efficient and low-irritation oral ulcer treatment medicine.

This invention discloses a highly transferable adversarial example generation method and its application in a secure test range. The method applies principal component protected feature perturbation transformation to the intermediate layer feature map of a momentum-based iterative fast gradient symbolic attack framework. It identifies the principal and secondary components of the feature space using singular value decomposition (SVD), quantifies the redundancy between feature channels using Pearson correlation coefficient analysis, and achieves precise feature perturbation using an adaptive mask generation mechanism based on Bernoulli sampling and energy conservation constraints. This method can be embedded as a pluggable module into existing iterative gradient attack processes without modifying the main structure of the basic attack algorithm or retraining the source model. By dynamically registering hook functions during forward propagation to intercept and transform intermediate layer features, it significantly improves the transferability of adversarial examples on unknown target models. The method's high transferability and ease of use in the test range environment are verified, providing users with an effective tool for evaluating model robustness.

This invention discloses a multi-level land resource segmentation migration fine-tuning method with input alignment, comprising the following steps: acquisition and processing of remote sensing image data and label data related to the land resource segmentation task; selection and slicing of sampling areas based on prior knowledge of the study area; data augmentation to expand the dataset and pixel value normalization; partitioning the model dataset using stratified sampling techniques; determining the optimal initial learning rate through hyperparameter optimization, entering input feature alignment, and establishing a quick connection between the input alignment module and the model output; employing three Transformer architectures and their pre-trained models for land resource segmentation, and optimizing the model training process using gradient-value-constrained gradient clipping and two-stage training strategies; step six: segmentation prediction and visualization. This invention improves the model's training efficiency and segmentation performance, achieving high-precision multi-level land resource segmentation and recognition.