76results about How to "Accurate distinction" patented technology

The invention belongs to the technical field of power transmission line modeling, and provides a power transmission line channel point cloud modeling method and system, and the technical scheme is that based on standardized point cloud data, global statistical features are extracted, channel scene categories are automatically judged, and adaptive feature weight vectors are generated according to the judged scene categories; performing weighted calculation on the similarity by using the feature weight vector, and performing superbody clustering segmentation on the standardized point cloud data to obtain an optimized superbody set formed by a plurality of superbodies with similar internal features; based on multi-dimensional feature statistics and context rules, screening the optimized superbody set, and rejecting small non-ground interference superbodies to obtain a ground candidate superbody set; and performing identification, merging and curved surface fitting on the ground candidate superbody set to generate a continuous power transmission line channel ground model. High-precision identification of ground points is realized, and a high-quality ground reference can be provided for safety analysis of a power transmission channel.

The invention relates to the field of genes, in particular to molecular markers for judging flowering time of brassica napus L. The molecular markers are Indel-282, Indel-326 and Indel-353 respectively and are located on an A02 chromosome of the brassica napus L.. The molecular marker has high specificity and stability, and is significantly related to the flowering time. The markers show high specificity and stability in brassica napus plants under different genetic backgrounds. By detecting the molecular markers through PCR amplification and electrophoresis, the flowering time of different plants can be quickly and accurately identified in the early stage of breeding. Therefore, the flowering time of the plant can be determined in the early stage, so that the breeding efficiency is improved. The molecular markers can be used for molecular marker-assisted selective breeding, and plants with early flowering or late flowering characteristics can be quickly screened by detecting the markers, so that the breeding process of excellent varieties is accelerated, and the functions and regulation mechanisms of related genes are further disclosed.

The invention relates to the technical field of medical instruments, in particular to an enteric cavity segment filling controllable device for colon cancer preoperative T staging, which comprises a central mandrel, a flexible head end cap is arranged at the foremost end of the central mandrel, and a near-end balloon inflating pipeline, a far-end balloon inflating pipeline and a main pipeline which are mutually independent are arranged in the central mandrel; the near-end balloon is arranged on the central mandrel in a sleeving manner and is communicated with the near-end balloon inflating pipeline; the far-end balloon is arranged on the central mandrel in a sleeving manner, is positioned on the far side of the near-end balloon and is communicated with the far-end balloon inflating pipeline; at least two main pipeline lateral holes are formed in the tube wall of the central mandrel between the near-end balloon and the far-end balloon, are communicated with the main pipeline and are used for injecting a contrast agent or normal saline into the closed enteric cavity section between the two balloons. The near-end balloon and the far-end balloon are synchronously filled at the two ends of a target intestinal segment to form a closed segment, and a contrast agent is uniformly injected through lateral opening, so that the intestinal wall is fully expanded and does not collapse, and the display definition of the five-layer structure of the intestinal wall is remarkably improved.

A method and system for identifying wireless communication devices based on radio frequency fingerprinting is disclosed. The receiver extracts radio frequency fingerprint features from the wireless signals transmitted by the wireless communication device to identify its identity. The method includes: Step 1. Pre-training a feature extraction network and a classification head using known device label data; Step 2. Jointly estimating the number of known and unknown device categories in the mixed signal through feature space stripping and effective rank estimation; Step 3. An adaptation phase, freezing the anchor points of the known class classification head and constraining the anchor points of the unknown class to actively guide isolation; Step 4. Obtaining pseudo-labels by performing semantic geometric registration on unlabeled samples based on balanced entropy regularized optimal transmission; Step 5. Adaptively training the feature extraction network based on progressive manifold anchoring and reliable pseudo-replay; Step 6. Outputting the identity of the device to be identified using the trained feature extraction network and a unified classifier. This invention solves the stability-plasticity dilemma of known class forgetting and unknown class confusion, achieving accurate device identification and number estimation in dynamic environments.

This application provides a method and apparatus for open-set skeleton action recognition based on outlier prototype learning. The method includes: constructing a neural network model; preprocessing human skeleton data by obtaining initial features through a multi-branch feature extraction network; processing the initial features using a classifier and a hypersphere feature mapper to obtain logical prediction values ​​for action categories and branch features; training the neural network model based on a training set; after the first iteration, optimizing the feature space of in-distribution samples using multi-class loss; synthesizing virtual outliers in the optimized feature space of in-distribution samples; optimizing the energy boundary by combining in-distribution samples and virtual outliers; after the second iteration, selecting the optimal model weights based on the combined performance of open-set recognition and closed-set classification on the validation set; and determining the action type and triggering the corresponding robot operation based on the energy score output by the trained neural network model and the action classification result.

The application discloses a characteristic peptide of Pheretima guangxiensis and application thereof, and belongs to the technical field of biological medicines. The characteristic peptide is LLNYVDYK, EFTEEQLR, LQAELAR, ALETQVK, FVVLK, ARDDPHCCSAQDR or GILAADESTATIGK. The characteristic peptide is highly specific, and a detection method thereof is simple in operation and high in sensitivity, so that whether the Pheretima guangxiensis component is contained in the animal medicine can be accurately identified, and a scientific method is provided for guaranteeing the quality of the Pheretima guangxiensis and products thereof.

The invention relates to the technical field of wireless communication security, and discloses a security encryption processing method and system for a wireless access area control instruction. The method comprises the following steps: acquiring network state data and an instruction data packet; performing risk quantification processing according to the network state data to obtain a risk vector; performing threat grouping analysis according to the risk vector to obtain a threat distribution combination; determining an encryption parameter combination according to the threat distribution combination; performing fragmentation and encryption processing on the instruction data packet according to the encryption parameter combination to obtain encrypted instruction content; and performing path verification according to the encrypted instruction content to obtain a secure transmission path. According to the invention, the security encryption processing of the control instruction can be realized under the dynamic network condition, and the reliability and security of instruction transmission are improved.

According to the metabolism-oriented clustering-based tumor-associated macrophage metabolism typing method provided by the invention, the single-cell RNA sequencing data set for analysis is collected and arranged, and then the tumor-associated macrophage population is identified and extracted from the single-cell RNA sequencing data set, so that the technical deviation is eliminated; a core metabolism gene set is constructed based on a tumor-associated macrophage population, then weight distribution and optimization of core metabolism genes are carried out, contribution of the metabolism genes in clustering analysis is enhanced, metabolism-oriented clustering is carried out based on the optimized weight, and a metabolism typing result is generated. Therefore, enhancement of metabolic gene expression characteristics and more accurate distinguishing of TAMs subgroups are achieved, typing results are analyzed and verified through GSEA enrichment analysis or GO enrichment analysis, typing categories are output, challenges confronted when high-dimensional single-cell data are processed through a traditional method are effectively overcome, the method is not only suitable for TAMs, but also can be expanded to other tumor-related cell types, and the method has a good application prospect. And the blank in the prior art is filled.

The invention discloses a rapid detection method for echinococcus granulosus based on RPA-CRISPR / Cas, and belongs to the technical field of molecular diagnosis. The method comprises the following steps: extracting DNA of a sample to be detected; the specific primers are used for RPA isothermal amplification, a reaction system contains a buffer solution, upstream and downstream primers, RPA enzyme, magnesium acetate and template DNA, and reaction is carried out at 39 DEG C for 20 minutes; and mixing the amplification product with LbCas12a protein, sgRNA and a fluorescence report probe, carrying out CRISPR / Cas reaction at 37 DEG C for 30 minutes, and judging the result through a real-time fluorescence signal. Constant-temperature rapid amplification is achieved through RPA, high-sensitivity and high-specificity detection of the echinococcus granulosus can be achieved under the site condition by combining high-specificity recognition and trans-cleavage activity of CRISPR / Cas12a, the sensitivity reaches the single-site copy number, operation is easy and convenient, complex temperature control equipment is not needed, and the method is suitable for primary medical treatment and site quarantine.

This application relates to a method for identifying key load regions based on spatiotemporal load curves. The method includes: performing event-based decomposition of spatiotemporal correlation data in the region to be identified to obtain a load behavior event sequence; performing spatiotemporal transmission analysis on each spatial unit in the region to be identified based on the load behavior event sequence and the spatiotemporal correlation data to obtain a spatiotemporal dynamic relationship diagram; performing dynamic core potential analysis on each spatial unit based on the spatiotemporal dynamic relationship diagram to obtain core dynamic seed units; performing dynamic boundary growth on the associated spatial units in each spatial unit based on the spatiotemporal dynamic relationship diagram and the core dynamic seed units to obtain candidate core dynamic regions; and performing disturbance removal simulation on the candidate core dynamic regions to obtain the core dynamic regions. This method can improve the accuracy of the core dynamic region identification results.

The invention belongs to the technical field of identification of variety resources and germplasm innovation, and discloses an SSR (simple sequence repeat) core primer combination for identifying the content of'high saikoside 'in a radix bupleuri variety, a kit and application. According to the invention, a pair of specific SSR core primers (CHSSR1F / CHSSR1R) are screened by using a microsatellite molecular marker development technology. The primer combination can specifically amplify a characteristic band of 136 bp in a Chinese thorowax root variety with high saikosaponin, so that the Chinese thorowax root variety with low saikosaponin can be quickly and accurately identified and distinguished from other Chinese thorowax root varieties with low saikosaponin. The SSR core primer combination, the kit and the identification method provided by the invention overcome the defects that the traditional chromatographic identification method is time-consuming, high in cost, dependent on equipment and the like, have the advantages of reliable result, simplicity and convenience in operation, high detection efficiency and the like, and are beneficial to popularization and protection of a'high saikosaponin 'radix bupleuri variety and medicinal material quality control.

The application discloses a data quality evaluation system and method based on artificial intelligence, and relates to the technical field of data quality evaluation.The total quality loss score is determined according to the first state parameter value and the second state parameter value of the media data; the time cumulative damage component, the abnormal jitter frequency and the jitter peak cumulative amount are calculated according to the instantaneous transmission rate, the rate representative value and the transmission time delay jitter in the media data transmission process; the minimum transmission guarantee strength is calculated according to the maximum jitter peak value, the transmission guarantee failure coefficient is determined, the concurrent scheduling pressure loss coefficient is calculated according to the concurrent scheduling layer and the jitter peak cumulative amount, and the transmission scheduling comprehensive damage value is obtained in combination with the transmission guarantee failure coefficient; the BP neural network model is constructed, the time cumulative damage component, the abnormal jitter frequency, the jitter peak cumulative amount and the transmission scheduling comprehensive damage value are input into the model, the data transmission quality loss score is obtained, the data receiving and transmitting quality loss score is further obtained, and the quality optimization suggestion is generated.

The application discloses a kind of soil water salt dynamic monitoring and quality evaluation method based on multispectral remote sensing, it is related to geological monitoring field, the method includes by satellite remote sensing image, unmanned aerial vehicle multispectral image, ground sample and meteorological driving data, scattering correction and roughness correction are carried out to observation spectrum, obtain soil intrinsic reflectivity, vegetation cover component and salt crust component;Water salt inversion model is established in combination with thermal infrared and red edge band information, and multi-time phase soil volume moisture content and surface salt index are inverted;By introducing meteorological conditions and vegetation dynamic characteristics, water salt coupling partial differential equation and graph time and space constraint model are constructed, and the continuous time and space distribution of soil moisture content and salt is obtained;Salinity, water, vegetation response and soil health index are extracted, and soil quality comprehensive evaluation model is established, and high-risk patch and management priority order are output.The method realizes the dynamic monitoring and quality classification evaluation of soil water salt, and is suitable for saline-alkali soil management and ecological restoration.

The invention relates to the technical field of jadeite detection, in particular to a method for identifying organic matter types in jadeite based on a laser Raman spectrum D / G peak. According to the invention, the trace unsaturated hydrocarbon component (1t) of solid wax oil in jadeite is found for the first time; according to the present invention, the laser Raman spectrum D / G peak can induce the D peak (1350 + / -20 cm <-1 >) and the G peak (1600 + / -20 cm <-1 >) under the thermal effect of the laser, such that the method for determining the type of the organic matter in the jadeite based on the laser Raman spectrum D / G peak is developed for the first time, if the D peak and the G peak appear at the same time or only the G peak appears in the laser Raman spectrum, the organic matter in the jadeite is determined to be the solid paraffin, and if the D peak and the G peak appear in the laser Raman spectrum, the organic matter in the jadeite is determined to be the organic matter in the jadeite is determined to be the organic matter in the jadeite. Therefore, the corresponding relation between the unsaturated olefin component in the wax oil and the D / G peak of the Raman spectrum is established, and the blank of qualitative analysis of the organic matter of the jadeite is filled.

The application discloses a kind of carbon source recovery and water cycle control methods of kitchen waste leachate and municipal sewage, and it relates to wastewater treatment technical field, the method includes: kitchen waste leachate is sent into hydrolytic acidification unit after being removed slag, oil separation and homogenization;In the hydrolytic acidification process, volatile fatty acid, soluble chemical oxygen demand, pH and oxidation-reduction potential are obtained, and the ratio of volatile fatty acid and soluble chemical oxygen demand and its change rate are used to determine the taking time point of acidification product;The removed acidification product is detected oil, suspended solids, ammonia nitrogen, pH and flow, and the carbon source availability is calculated and divided into direct dosing grade, bypass domestication dosing grade and reflux reacidification grade;According to the nitrate load of municipal sewage biochemical system, total nitrogen effluent trend and carbon source availability, determine the dosing amount and dosing position;Thus realize the graded recovery of organic carbon source in kitchen waste leachate, low impact utilization and water recycling.

The application discloses a kind of based on artificial intelligence's consumer goods harmful substance intelligent screening method, it is related to artificial intelligence technical field, it is difficult to effectively identify and analyze the technical problem that a large number of natural language, commercial name or commonly known as etc. non-standardized description exists in ingredient list, it is also difficult to accurately convert these fuzzy text information into the standardized terminology that can be accurately matched by regulation database;Including the following steps: the fuzzy ingredient list containing natural language or commercial name description is preprocessed by multi-task text regularization model;Then, utilize embedding semantic similarity model and the multi-evidence fusion presumption unit of improved attention mechanism, fuzzy component is intelligently analyzed and matched, and standard substance identification and confidence are output;Finally, through multi-agent risk decision framework combines real-time regulation database, dynamic compliance screening and risk assessment are carried out;The application improves the intelligent level of consumer goods safety compliance management.

The invention provides a mixed library paired read genome type detection method and a related product. The specific embodiment of the method comprises the following steps: executing the following feature extraction operation on at least one read of a to-be-detected paired read: for each candidate cytosine site in the read, extracting the candidate cytosine site in the to-be-detected paired read, determining whether the candidate cytosine site is a non-CG site or not according to the dinucleotide context where the candidate cytosine site is located, counting the number of all candidate cytosine sites which are judged to be non-CG sites in the read, and determining the number of characteristic basic groups of the read; and generating a feature parameter set of the read based on the number of feature bases of the read. According to the method, data source detection is performed on the to-be-detected paired read segments on the basis of the feature parameter set of at least one read segment in the to-be-detected paired read segments, so that the analysis efficiency of the mixed sequencing data is improved while the experiment cost is reduced and the sample consumption is reduced.

This invention discloses a transgenic recombinant immune cell that specifically targets tumors. The recombinant immune cell comprises a chimeric polypeptide and a chimeric antigen receptor; the chimeric polypeptide includes a first extracellular region, a first transmembrane region, and a first intracellular region with HLA-G protein binding activity, wherein the N-terminus of the first transmembrane region is connected to the C-terminus of the first extracellular region, and the N-terminus of the first intracellular region is connected to the C-terminus of the first transmembrane region; the first transmembrane region includes the transmembrane region of the Notch receptor protein from Xenopus laevis; the antigen-chimeric receptor does not have HLA-G protein binding activity. This recombinant immune cell significantly improves the precise recognition of tumor cells by immune cells, reduces off-target effects of cell therapy, and accurately and effectively distinguishes tumor cells from normal cells, providing an effective target and targeting method for the treatment of broad-spectrum tumors (especially solid tumors).

The utility model relates to the technical field of display control, and discloses a double-vision double-control module and a vehicle, two wearable devices are introduced, each wearable device can generate unique frequency signals and transmit the frequency signals to a display screen through the body of a user, and therefore accurate recognition and distinguishing of touch operations of different users are achieved. According to the innovative scheme, the problem that user operation is difficult to accurately distinguish in a multi-user interaction scene of an existing double-vision display screen is effectively solved, and the user experience is remarkably improved. Specifically, according to the scheme, the recognition accuracy and the response speed of the touch operation are improved, and the stability and the anti-interference capability of the module are enhanced. Besides, through respective control of display contents of different visual angles, the interaction experience of multiple users is further optimized, the operation habits and requirements of different users are better met, and wide market application prospects and remarkable economic benefits are achieved.

The application discloses a method for identifying matcha produced by different enterprises by using fNIRS, and belongs to the field of food detection. The method detects matcha produced by different enterprises by using functional near-infrared spectroscopy (fNIRS), carries out pretreatment and discriminant analysis on full-channel characteristic wave bands, screens characteristic channels, and further uses a discriminant model established by using characteristic wave bands of the characteristic channels to accurately identify matcha produced by different enterprises. Compared with traditional sensory evaluation methods and ingredient detection technologies, the method is objective and fair, does not need to train an evaluation personnel, time consumption is shortened, and the method can be used for rapidly identifying matcha produced by different enterprises.

The invention discloses a cable joint detection method, system, equipment and medium based on multispectral and laser technologies, and belongs to the technical field of cable joint detection, and the method comprises the steps: synchronously collecting multiband optical data and spatial depth data of a target cable area; constructing an energy function based on correlation constraints of the spatial gradient field and the geometric gradient field, solving a spatial transformation matrix by optimizing the energy function, and performing spatial registration on the corrected multi-band optical data and the processed spatial depth data by using the spatial transformation matrix; multi-dimensional features are extracted from the registered data, and when the same spatial position meets a preset multi-dimensional feature judgment condition, the cable joint is judged to be a cable joint; and converting the identified coordinates of the cable joint from the sensor coordinate system to a global coordinate system, and outputting positioning information of the cable joint. According to the method, strict time synchronization of multiband optical data and space depth data is realized through hardware trigger pulses, and the time-space consistency of heterogeneous data in an acquisition stage is ensured.

The present application belongs to the technical field of intelligent manufacturing control, and particularly relates to a pressure-sensitive self-repairing asphalt cold patch material mixing intelligent control system and method, which comprises the following steps: collecting torque time series data of a mixing motor; constructing a dynamic feature vector of the torque time series data at each time point based on the torque time series data, which is used to represent the asymmetry and transient peak degree of the torque time series data fluctuation; performing unsupervised analysis on a feature space formed by the dynamic feature vector by using a local anomaly factor algorithm to calculate a local anomaly factor corresponding to each time point; and executing a closed-loop control strategy according to the local anomaly factor and the feature representing asymmetry in the dynamic feature vector to realize intelligent control of the mixing process. The present application can accurately distinguish different mixing conditions, can timely stop production to protect product activity, and can reasonably judge mixing uniformity through the asymmetry feature to improve production efficiency.

This invention discloses a target trajectory background modeling method and system based on bidirectional difference and grayscale correction, belonging to the field of image data processing technology. The method includes: S1, periodically acquiring basic image trajectory data and performing preprocessing on the acquired data; S2, constructing a frame index sequence, generating a frame pairing table and a grayscale coefficient table, performing grayscale correction and bidirectional difference calculation on the paired frame sequence to obtain the candidate motion mask for the current frame; S3, obtaining the initial background region of the current frame, evaluating the background confidence level of each pixel, and generating an effective background region; S4, comprehensively evaluating the stability level and fluctuation constraint capability of the background, dividing the stable and weak background regions, and outputting the frame-level motion mask to complete the background closed-loop output. This solves the problem that traditional background modeling methods directly use arithmetic mean modeling in a few-frame sequences, leading to the background grayscale being lowered by high-contrast moving pixels, thus generating motion artifacts and causing weak target detection failure.

The present application belongs to the technical field of data processing, and particularly relates to a kind of software data of case-related electronic equipment traceless detection method, device, computer equipment and storage medium, method includes the following steps: based on the concurrent execution of multiple detection tasks of case-related electronic equipment software data, detection task at least includes file keyword scanning task, browser history scanning task, social account scanning task and remote control software scanning task;Risk classification is carried out according to the detection results of all detection tasks, and risk research information is generated.Through concurrent execution of multiple detection tasks, the overall detection time is significantly shortened, and the efficiency of on-site evidence collection is improved, and finally risk classification is carried out according to the detection results of all detection tasks and risk research information is generated, which provides intuitive and operable decision support for law enforcement personnel.

The utility model provides a multispectral imaging device for detecting stored grain pests, which comprises a sealed box body as well as a sample table, a light source and a bracket which are arranged in the box body, and is characterized in that the bracket is a T-shaped bracket, the tail end of the transverse part of the T-shaped bracket is connected with an imaging sensor, the bottom of the imaging sensor is connected with a lens facing the sample table, and the lens is connected with the light source. A plurality of optical filters switched through optical filter wheels are arranged below the lens, and a sample protection cover is arranged between the lens and the sample table. The grain pest detection device realizes rapid, lossless and high-precision detection of pests in grains, reduces grain storage loss, is simple in overall structure and wide in detection range, can improve the detection efficiency of the pests in the grains, overcomes the defects of low efficiency, poor accuracy, unfriendliness to the environment and the like of a traditional detection method, and is suitable for popularization and application. And an efficient and environment-friendly solution is provided for storage, processing and quality control of grains.

The application discloses a kind of detection devices of parasitic loop, and the capacitor input module includes at least two parallel capacitors, and each capacitor can be respectively connected to the measured loop or disconnected with the measured loop by different capacitor lines;The host control module is used to detect the voltage of two voltage detection points of the measured loop after the air switch of the measured loop is disconnected, before the capacitor is connected;Control capacitor input module in the capacitor is connected to the measured loop by different capacitor lines in turn, and the voltage of two voltage detection points is detected every time a capacitor is connected;According to the voltage variation of two voltage detection points, it is judged whether the measured loop exists parasitic loop or not.The application can accurately distinguish parasitic series electricity and induced electricity, exclude the interference of induced electricity, truly judge whether detection parasitic loop exists or not, without manual wiring, save time and effort, and can be applied to the parasitic loop detection of various secondary circuits.

This invention relates to the field of network protocol technology, specifically to a data processing method for a multi-protocol fusion monitoring terminal based on edge computing. The method includes: acquiring port static level signals and outputting test pulses; calculating the RC time constant; matching communication parameters to generate a port device identifier; based on the exposure time of video frames, backtracking and linearly interpolating in a circular buffer to generate synchronous sensor values; generating a spatiotemporally aligned fusion data stream and uploading it to the main processor; monitoring heartbeat signals sent by the main processor during the transmission of the fusion data stream; if the heartbeat interval exceeds a preset threshold, determining a data transmission link failure, blocking the normal display path of the fusion data stream, and outputting a preset alarm display interface; simultaneously performing data processing and transmission, capturing port surge current waveforms, extracting energy integral characteristic parameters, and comparing them with a benchmark fingerprint; if the device health status is determined to be abnormal, performing port isolation operations.

The invention belongs to the field of microbe counting, and particularly relates to a dead and viable bacteria absolute numerical analysis method based on laser confocal fluorescence space analysis. According to the present invention, the survival state of the bacterial population is evaluated by monitoring the integrity of the cell membrane through the LIVE / DEAD BACLIGHT 1 KIT reagent (LIVE / DEADBacterium Viability Kits); the kit adopts a two-color fluorescence detection technology, and can be used for detecting viable bacteria of various bacteria genus. When the cell membrane is damaged to cause bacterial death or endangered, the cell membrane and the propidium iodide present red fluorescence or orange fluorescence (co-dyeing); and viable bacteria with a complete cell membrane and the SYTO9 dye present green fluorescence. The counting method is convenient and fast, and can accurately distinguish dead and viable bacteria.

This invention belongs to the field of medical image processing technology, specifically relating to a surgical instrument segmentation method based on a visual foundation model and structural prior constraints. The specific process is as follows: The DINOv2 visual foundation model extracts visual features at different levels; a low-rank adaptation module is injected into the attention-related linear mapping layer of the DINOv2 visual foundation model encoder. The low-rank adaptation module freezes the original weight matrix of the linear mapping layer, sets two low-rank matrices to model the weight increment, and performs a linear projection mapping operation on the input features using the frozen original weight matrix; output features are obtained using a lightweight convolutional feature extraction unit constructed using GhostModule; after performing multiple dilated convolutions and average pooling operations with different dilation rates on the features using the Ghost-ASPP module, feature compression and fusion are performed to output multi-scale semantic features; based on the multi-scale semantic features, the surgical instrument segmentation result is predicted.

The application discloses a hash code generation model training method and a hash code generation method, relates to the technical field of artificial intelligence, and comprises the following steps: sampling a batch of data from a text data set; inputting target text data in a text data sample in the batch of data into an initial hash code generation model to obtain a feature projection vector corresponding to the target text data, so as to determine a contrast loss value, an optimal transport distance margin loss value and a uniform distribution regularization loss value, and then determine a total loss value; based on the total loss value, updating model parameters of the initial hash code generation model, returning to the step of sampling a batch of data from the text data set until the model converges, and obtaining a target hash code generation model. The technical problem that a hash function in the related art cannot effectively distinguish similar vectors is solved, and the technical effect of effectively distinguishing similar vectors is achieved.