151results about How to "Overcoming feature" patented technology

The invention belongs to the biotechnology field, and in particular relates to escherichia coli TolC antigen and antibody. The antibody has the specificity aiming to a sequence shown by SEQ ID NO:1. The invention also provides a gene sequence for encoding the escherichia coli TolC antigen and application of the antibody to preparing a drug for inhibiting the bacterial resistance. The anti-TolC fragment SNGYRDANGI antibody enhances the sensitivity of bacteria to antibiotic and can specifically and efficiently inhibit the functions of drug resistance outer membrane protein of escherichia coli, thereby solving the problem of the bacterial resistance. Compared with the traditional antibody, the antibody has higher safety and operability on the application of the bacterial resistance drug.

The invention relates to the technical field of substation state monitoring, in particular to a method and a device for monitoring gas in transformer oil on line. A standard oil sample is subjected to factory delivery calibration during delivery test, the delivery recording of a specified gas sensor is recorded, the peak height test and factory delivery recording of the standard gas are compared and corrected, and the characteristic drift of the gas sensor is corrected and reduced. The method has profound significance for improving the long-term stability of the gas in the transformer oil through the online monitoring system, guaranteeing the transformer safety and stabilizing the operation.

The invention discloses a question-answer matching method fusing deep representation and an interaction model. The method comprises the following steps: firstly, pre-training word vectors and character vectors of a text pair consisting of spoken questions input by a user and standard questions in a knowledge base; and then fusing the fusion depth representation and the interaction model in a loosecombination or tight combination mode, and performing question and answer matching on spoken questions input by the user by using the fused model. The universality of the service can be enhanced, therecognizable spoken language category can be expanded, the answer matching accuracy can be improved, and accurate semantic matching of wide semantic recognition can be realized.

The invention discloses a method for preparing intelligent target medicine carrying composite micelles. The surfaces of the composite micelles are modified by target ligands and pH / temperature synergy sensitive polymers at the same time; and reversible transformation between the shielding state of the target ligands during blood circulation (the temperature is 37 DEG C, and the pH value is 7.4) and the deshielding state of the target ligands in tumor tissues (the thermal therapy temperature ranges from 40 DEG C to 44 DEG C, and the pH value ranges from 6.5 to 6.8) is achieved. The pH and temperature dual sensitivities used in the method have the synergy; and the method is more suitable for the complex physiological environment of the human body. The method has the beneficial effects that targeting of the target ligands and the sensitive phase transformation characters of the pH / temperature synergy sensitive polymers are sufficiently used; and a target reversible shielding nanometer drug delivery system with the tumor targeted specificity and the blood circulation stability is built.

The invention relates to a method for detecting and classifying pixel defects in inkjet printing of a printed display panel and a device thereof. The method comprises the following steps: obtaining the image of a panel to be tested in blocks; when an original image is received, preprocessing the original image; when the pre-processed original image is received, stitching all the partitioned imagesof the same panel ; when the original image is received, segmenting the image is segmented; when the segmented original image is received, using a classifier in machine learning to classify the image, if there is no defect, the image entering the subsequent process, and if there is any, giving the prompt; when the original image of the defective pixel is prompted after the classification is received, positioning and subclassifying the defective pixel of the image, and carrying out statistical analysis of the category and the number of the defective pixels of the image to indicate corresponding defects. The invention aims at solving the problems that the manual detection and classification exist for a long time, the efficiency is low, the detection result is easy to be influenced by humansubjective factors, and the consistency is poor.

The embodiment of the invention provides a network encrypted traffic identification method and device based on deep learning, relates to the technical field of network security, and can improve the accuracy and efficiency of network encrypted traffic identification. The method comprises the steps of obtaining message information and communication behavior information of a preset number of data packets belonging to a communication establishment stage in to-be-identified network traffic; constructing a message two-dimensional data matrix according to the message information, and constructing a behavior two-dimensional data matrix according to the communication behavior information; and inputting the message two-dimensional data matrix and the behavior two-dimensional data matrix into a network traffic identification model, and determining the protocol type of the to-be-identified network traffic, wherein the network flow identification model is a sample two-dimensional data matrix passing through the sample network flow and a protocol type label corresponding to the sample network flow; and training the deep learning network to obtain a model, wherein the sample two-dimensional datamatrix comprises a sample message two-dimensional data matrix and a sample behavior two-dimensional data matrix corresponding to the sample network flow.

The invention discloses a method for intelligently identifying mineral information based on spectral information. The method comprises the following steps of: (1) extracting the spectral information comprising spectral waveform and spectral characteristic parameters; (2) reading in hyperspectral image data; (3) resampling each reference spectrum according to the wavelength of the image spectrum; (4) removing continuum of the image spectrum and the resampled reference spectrum; (5) performing least square fitting on the image spectrum and each reference spectrum processed by the step (4) at a primary absorption characteristic band and a secondary absorption characteristic band to obtain an initial matching value; (6) performing three kinds of constraint processing, namely matching of spectral angle at a characteristic band, judging of existence of specific absorption characteristic and setting of a reflectivity threshold value of the characteristic band, on the initial matching value to obtain the final matching value; and (7) obtaining the final identification result of the hyperspectral image by using the constraint of spatial distribution continuity of a mineral, wherein the reference spectrum corresponding to the maximum matching value in the final matching values obtained by the step (6) is a preliminary identification result of a pixel.

The invention discloses an image defogging method based on weak supervision generation antagonism network. The present invention mainly solves the problem of instability of defogging effect of the real foggy day image in the prior art. The implementation scheme is as follows: 1) obtaining a real training set and a composite training set; 2) respectively constructing a generation network and a confrontation network; 3) updating the parameters of the countermeasure network according to the loss function of the countermeasure network; 4) establishing a loss function of the generating network andupdating parameters of the generating network according to the loss function; 5) judging whether the update times of the generation network and the antagonism network reach 100, if so, inputting the real foggy day image into the generation network for defogging, otherwise, returning to step 3). The image after the method of the invention has rich details, improves the brightness, contrast and clarity of the image, reduces the phenomenon of supersaturation and distortion, and can be used in the field of computer vision.

The invention provides a rotor system typical fault diagnosis method based on motor current, belonging to the field of rotating machinery equipment. According to the method, firstly a current signal of a rotor system driving motor is collected, and an improved wavelet threshold denoising method is used for current signal denoising pretreatment on the current signal; then the current signal is subjected to EEMD decomposition, an meaningless IMF component is removed, the amplitude domain and time-frequency domain feature of each IMF component are extracted to form a feature vector, then a principal component analysis method is used to carry out dimension reduction on the feature vector, and a new low dimensional feature vector is obtained; and finally a support vector machine of particle swarm optimization is used to establish a nonlinear mapping relationship between the feature vector and different fault types, and the fault diagnosis is realized. According to the method, through the fault diagnosis of the motor current signal, the signal collection is convenient, the external interference resistance ability is high, and the method is easy to implement. The invention can be applied to the diagnosis of faults of mass unbalance, deflection angle misalignment, parallel misalignment and the like of a rotor system.

The invention relates to an intelligent configuration system and an intelligent configuration method for a wireless sensor network. The system comprises a sensor network module, a gateway module, an output module, an execution module and an application layer interface module. The system and the method solve problems of poor adaptability, poor reliability and relatively great control cost existing in a device and a method of a traditional sensor network. The system and the method for the wireless sensor network have advantages of intelligent performance, strong robustness, good versatility and small control cost.

The invention discloses a preparation method for a super-hydrophobic metal surface. The method comprises the following steps: constructing a micrometer structure on the surface of a frequently used engineering material (such as copper, iron, stainless steel, nickel, aluminum, an aluminum alloy and a high-molecular section covering the above-mentioned metal through chemical plating) by using an electrocoppering process; then constructing a nanometer structure by using a chemical silvering process; and carrying out modification with low-surface energy substances like octadecanethiol so as to obtain excellent and stable super-hydrophobicity. The method has a wide application scope and can be used on pure metal, alloys and metalized high-molecular sections; an obtained surface has stable and reliable surface performance, strong film-substrate bonding force and good corrosion resistance, especially good acid resistance, can be extensively applied in fields like self-cleaning, water prevention and dehumidification, corrosion prevention, fluid drag reduction, frost prevention and condensation and heat exchange, and has an industrialization value.

The invention discloses a method for identifying a shielded face on basis of blocks and identification of non-negative matrix factorization. The method is adopted for solving the problem of low identifying rate for the large-area continuously shielded face of the prior art. According to the technical scheme, the method comprises the following steps: 1) constructing a data matrix and a No.k block non-negative matrix of a training dataset; 2) constructing a weight matrix and a new inter-category divergence matrix according to the cosine similarity between the categories of the data matrixes, thereby forming a new target function; 3) performing optimization solution on the target function, thereby acquiring a basis matrix and a coefficient matrix; 4) constructing a No.k block test data matrix of a test set, and projecting the No.k block test data matrix on the basis matrix, thereby acquiring a projecting coefficient matrix; 5) calculating distances from each column vector in the projecting coefficient matrix to all the column vectors in the coefficient matrix, and utilizing a weight fusion criterion to acquire the category of each image in the test set. An experiment proves that the identifying rate for the large-area continuously shielded face is increased according to the method provided by the invention and the method can be applied to the field of personal identification and information safety.

The invention discloses an image processing method in an underwater vision SLAM (Simultaneous Localization And Mapping) system. The image processing method comprises the following steps: establishing an underwater imaging model; processing influence of underwater environment factors on a camera imaging; carrying out image feature extraction and matching. According to the method, extraction on feature points of an underwater environment image in the later period is reinforced; according to a data correlation method which is newly disclosed on the basis of the improved SLAM system, the feature points are extracted in a matched manner, so that the feature points can be more rapidly and accurately extracted; real-time performance of an SIFT algorithm is improved. The method which uses a relative position factor of a binocular camera and a way point position factor as auxiliary conditions for carrying out correlation can effectively solve the problems of mismatching and poor matching efficiency in the data correlation.

The invention relates to a multi-block slide calculation method for a giant landslide, and belongs to the field of geological disaster engineering. The method is used for landslide disaster stability evaluation and engineering prevention and control. The method comprises the steps of: (1) finding out a landslide region range, deformation features and hydrogeology and engineering geology conditions; (2) based on a finite element program, establishing a finite element numerical model of a side slope; (3) inputting formation mechanical parameters of the side slope; (4) inputting reduction parameters and physical and mechanical characters of landslide massif for performing finite element calculation; (5) determining a side slope failure region; (6) continuously reducing failure strength parameters until slide surfaces are communicated; (7) determining a tension failure zone of each stage of slide surface; (8) setting the tension failure zone as a null unit in the model; and (9) re-calculating a new numerical model and searching out multiple stages of slide surfaces. According to the multi-block slide calculation method for the giant landslide, provided by the invention, the multiple stages of slide surfaces of the giant landslide can be accurately searched out and an important basis is provided for giant multi-stage landslide stability evaluation and engineering prevention and control.

The invention discloses a fuzzy-intelligence-based rail traffic car suspension system fault analysis method. The method comprises the steps that 1, a rail traffic car suspension system model is constructed, and dynamic characteristic analysis is performed on the model; 2, according to the dynamic characteristic analysis result of the rail traffic car suspension system model, an acceleration sensor is arranged; 3, multiple data time domain and frequency domain characteristics collected by the acceleration sensor are extracted, and distance characteristics are extracted through power spectrum analysis; 4, dimension reduction processing is performed on an original characteristic sample in the step 3, and a fault characteristic sample is obtained; 5, on the basis of the fault characteristic sample, fuzzy intelligence is utilized for performing fault classification on the car suspension system. According to the scheme, the defect that time frequency domain characteristic indexes describe signal changes from a certain aspect of a time domain or a frequency domain is overcome, meanwhile, meanwhile, the defects that time frequency domain characteristic indexes are easily added and average calculating operation submerge difference characteristics are obtained are overcome, and the characteristic sample quality is improved.

The invention provides a fluorogenic culture medium of Enterobacter sakazakii, a test kit and a test method thereof, and relates to a method for testing microbe and a composition used by the method. The fluorogenic culture medium comprises the compositions as follows: 8 to 20 grams of peptone, 4 to 7 grams of beef extract powder, 4 to 7 grams of sodium chloride, 1.0 to 2.0 grams of cholate, 12 to 20 grams of agar, 0.05 to 0.5 grams of X-a-glucopyranoside, 0.01 to 0.04 grams of Na2CO3, 0.3 to 2.0 grams of ferric ammonium citrate, and 0.3 to 2.0 grams of sodium thiosulfate. The test kit consists of the fluorogenic culture medium of Enterobacter sakazakii, an enriched liquid A ,namely buffer peptone water culture medium, and an enriched liquid B, namely modified lauryl sulfate pancreas peptone broth. The test kit is configured simply, and the test method has high sensitivity in detection, so the method is suitable for treating large flux samples.

The invention provides a multi-human-body tracking method based on a human body part model and relates to the video monitoring technology field. The method comprises the following steps of acquiring image data, extracting the human body part model of each person and calculating a color characteristic, acquiring the part characteristic set of each person, and gathering and acquiring a multi-human-body part characteristic set list; then, calculating the similarity of the part characteristic set of each person acquired from a current frame and each person in the multi-human-body part characteristic set list acquired from a previous frame and acquiring a matching matrix; and according to the matching matrix, calculating the similar confidence of each person acquired from the current frame to each person acquired from the previous frame, according to the similarity and the similar confidence, matching a target acquired from the current frame with the target of the previous frame, and acquiring a tracking result. By using the method, the detection target of the current frame can be accurately matched to the position of the previous frame and a new target is discovered, and the problem oftracking failure caused by a characteristic difference generated because of human body deformation during a multi-human-body tracking process is effectively solved.

A ferroelectric FET with bismuth titanate (BIT) film as storage medium layer is prepared through washing Si substrate, coating photoresist, exposing, developing to obtain pattern, etching to form recess, coating photoresist, phootoetching to form injection windows of source and drain areas, ion implantation to form source and drain areas, depositing ferroelectric BIP film, etching off the film over source and drain areas to obtain the windows of source and drain, plating AG layer, and reverse etching to obtain source, drain and grid. Its advantages include good interface performance, long information storage time and high rate of finished products.

The invention discloses a determination method of high-pressure water injection and radial gas permeability of gas-containing raw coal and belongs to the technical field of coal mine disaster prevention. According to the method, a water injection hole is drilled in the top center of a selected and cut cylindrical sample, a water injection floral tube is mounted in the water injection hole, two ends of the water injection floral tube are sealed, the sample is placed in an annular gas permeating cylinder, cement is poured on a base of the annular gas permeating cylinder, a gas permeable tube is arranged between the inner wall of the annular gas permeating cylinder and the cylindrical sample, cement is poured into the gap formed between the cylinder wall and the cylindrical sample, a platen is arranged at the top of the cylindrical sample, after the cement in the annular gas permeating cylinder is set, a support rod is taken out, and a raw coal sample is obtained; then the raw coal sample is placed in an experiment tank, and the permeability of the high-pressure water invasion gas-containing raw coal sample is finally calculated through cooperation of a pressure loading system, a high-pressure water injection system and the like which are connected with the experiment tank. The method provides scientific instruction for prevention of coal mine disasters with the coal seam water injection technology.

The invention discloses a dynamic partition electricity price calculation method based on partition clustering analysis. The method comprises constructing a direct-current optimal power flow optimization target, determining a power system operation constraint condition, solving to obtain optimized hydroelectric output, and then solving node electricity price, conventional unit and new energy unitoutput, power flow data, an active dual factor and node injection power; generating an effective clustering attribute; determining an optimal clustering number; generating a dissimilarity matrix through standardization processing and a Euclidean distance method, and dividing a clustering algorithm to realize clustering; dividing each electricity price area according to the clustering result, and finally determining the unified pricing of each area; feeding back a calculation result of the regional electricity price, and carrying out settlement on market participants in each region according tothe regional electricity price. According to the invention, the electricity price of each subarea can be determined, and finally, each node in the same area is settled in a unified pricing manner, sothat the practical operability is improved, and the demand of obtaining relatively stable price by most users is met.

The invention discloses a short-term prediction method for marine environment monitoring factors. According to the method, an autoregressive model is adopted for a short-term prediction of marine environment input factors, and then a deep neural network method is adopted for a short-term prediction of output factors. When the characteristics of the autoregressive model and the characteristics of a deep neural network model are efficiently used, short-term input predicted values are obtained by means of the autoregressive model, then predicted values of the output factors are obtained by means of the deep neural network model, and thus the limitations that prediction of a traditional neural network model to time series data is insufficient in extracted substantive characteristics and limited in gradient descent are overcome. The method achieves the prediction of the marine environment monitoring factors, and has an important significance in a short-term forecast of the marine ecological environment, a short-term pre-warning of marine disasters and the like.

The invention discloses a method for extracting parameters of a bipolar transistor and an equivalent circuit of the bipolar transistor. The method comprises the following steps: performing direct current test and S parameter test on the bipolar transistor; and performing curve fitting by use of ADS (Application Development System) software and origin software so as to acquire parameters of the bipolar transistor. The equivalent circuit established based on the method for extracting the parameters of the bipolar transistor comprises three inductors, five capacitors, six resistors and an intrinsic module BJT (Bipolar Junction Transistor), wherein the intrinsic module BJT comprises two resistors, three charge controls and three current controls. The method can be used for establishing bipolar transistor equivalent circuits suitable for circuits in different conditions, and also can be used for establishing the bipolar transistor equivalent circuit for high-frequency circuits. The bipolartransistor equivalent circuit established based on the method can reflect the device characteristics of a group of working points of the bipolar transistor, and can be used for circuit simulation of different situations; and the equivalent circuit can reflect the high frequency characteristic of the bipolar transistor, and can be used for high frequency circuit simulation.

The invention discloses an intelligent radar radiation source signal classification method based on a long-short time memory model. The method mainly solves the problem that the prior art is low in identification rate and slow in identification speed. The realization scheme is as follows: 1) generating a radar radiation source signal data set, and performing data preprocessing on the same; 2) acquiring a training sample set, a testing sample set and a verification sample set from the preprocessed data set; 3) constructing a seven-layer long-short time memory unit network and setting parametersof a network model; 4) adjusting the hyper-parameter of the network model and training the long-short time memory unit network by utilizing the training sample set and the testing sample set; and 5)inputting the verification sample set into the trained long-short time memory unit network model, thereby acquiring a radar radiation source signal classification result. Through the classification method disclosed by the invention, the automatic feature extraction and accurate signal classification can be performed on the one-dimensional signal; and the classification result is excellent, the time complexity is low, the stability is good, and the method can be used for the radar radiation source signal identification under the complex electromagnetic environment.

The invention discloses a multiplex PCR detection kit and a detection method for detecting thirteen types of transgenic soybeans. The multiplex PCR detection kit is capable of rapidly detecting thirteen types of transgenic soybeans imported under approval of the Ministry of Agriculture of China and comprises MON87701, GTS40-3-2, MON89788, CV127, A2704, A5547, DP356043, DP305423, MON87769, MON87708, 305423*40-3-2, MON87701*MON89788 and MON87705. The multiplex PCR detection kit is based on a multiplex PCR method mediated by a public primer and comprises detection of the public primer, a multipleprimer and a multiplex PCR system; a newly-designed annealing temperature method is used for improving specificity and sensitivity of the multiplex PCR system. The multiplex PCR method has inclusiveness; detection targets can be continuously increased; the detection throughput of multiplex PCR can be improved; the multiplex PCR detection kit has the advantages of high sensitivity, economy, rapidspeed and simplicity in operation, and can be applied to food detection.

The invention discloses a construction method of a morphological filter based on adaptive multi-scale AVG-Hat transformation. The construction method comprises following steps of: 1, determining the number of initial multi-scale structure elements of a signal and an initial structure element value according to a set parameter index based on a bearing fault signal; 2, constructing a set formed by the initial multi-scale structure elements through morphological expansion operation; 3, calculating results of the corresponding morphological AVG-Hat transformation of the bearing fault signal under the initial multi-scale structure elements, and constructing a set of the result; 4, selecting a ratio of the arrangement entropy of the filtered bearing fault signal to sparseness degree of an envelope spectrum as an evaluation index, and adaptively determining an optimal weight coefficient corresponding to the filtered initial multi-scale structure elements; and 5, according to the weight coefficient, constructing an optimal multi-scale morphological AGV-Hat filter. The construction method is advantageous in that the construction method is compatible to signal de-noising and fault feature extraction and accuracy of filtering results can be kept.

The invention relates to a high-frequency ultra-wideband dual polarization full wave radiation unit. The high-frequency ultra-wideband dual polarization full wave radiation unit includes a first dielectric substrate, four folding oscillators on the upper surface of the first dielectric substrate, four groups of band-shaped feeders which feed the four folding oscillators on the upper surface of the first dielectric substrate, four folding oscillators on the lower surface of the first dielectric substrate, four groups of band-shaped feeders which feed the four folding oscillators on the lower surface of the first dielectric substrate, and a radiating surface of a full wave oscillator, wherein the radiating surface is formed by copper coated layers being consistent in the upper layer and the lower layer on the first dielectric substrate; the upper and lower groups of folding oscillators respectively realize external feeding through a first feed bridge and a second feed bridge; the first feed bridge and the second feed bridge are respectively arranged on a second dielectric substrate and a third dielectric substrate; the second dielectric substrate and the third dielectric substrate are fixed in four balance baluns in an orthogonal manner; the internal and external conductors of coaxial cables respectively feed the band-shaped feeders through the balance baluns and the feed bridges. The high-frequency ultra-wideband dual polarization full wave radiation unit has the advantages of optimizing the folding oscillator circuit characteristic, realizing the ultra-wideband standing-wave ratio band width, and overcoming the problem that the full wave oscillator is hard to match high input impedance and is hard to realize the ultra-wideband frequency range requirement.

The invention belongs to the field of molecular biology, and particularly relates to a nasopharyngeal carcinoma microRNA detection kit. A purpose of the present invention is to solve the problems of poor stability, poor reproducibility, low specificity and easy false positive generation of the existing nasopharyngeal carcinoma microRNA detection kit. The invention provides the nasopharyngeal carcinoma microRNA detection kit, which comprises a probe composition, wherein the probe composition comprises a capture probe bound with a target microRNA and a signal amplification composition, and the target microRNA is one or a plurality of microRNA selected from hsa-miR-17-5p, hsa-miR-20a-5p, hsa-miR-29c-3p, hsa-miR-223-3p, hsa-miR-34b-3p, hsa-miR-34c-5p, hsa-miR-212-3p, hsa-miR-216a-5p, hsa-miR-217, hsa-miR-151a-3p, and hsa-miR-192-5p. The nasopharyngeal carcinoma microRNA detection kit of the present invention has advantages of good stability, good repeatability, high specificity, less false positive generation, and high sensitivity.

The invention discloses an automatic sorting and conveying device which is used for sorting and conveying materials. The automatic sorting and conveying device comprises a feeding unit, a sorting and conveying unit, a discharging unit, a supporting unit and a single-chip microcomputer, wherein the feeding unit, the sorting and conveying unit and the discharging unit are sequentially hinged and supported by the supporting unit; the sorting and conveying unit sorts and conveys materials arranged on a feeding frame one by one; the discharging unit sorts out and discharges the materials conveyed by the sorting and conveying unit to a next feeding port one by one; the sorting and conveying unit and the discharging unit are both connected with the single-chip microcomputer; the single-chip microcomputer is used for controlling the speed of sorting, conveying and discharging. The structural characteristics that rods are thin, long and bent are effectively overcome through a method of two times of hooking, the production method for sorting and feeding of thin and long round or square rod materials is totally changed, and the automatic sorting and conveying device is good in feasibility, high in intelligence and high in applicability.

The invention discloses a high-resolution electrical tomography conductivity distribution reconstruction method. According to the method, a weighting matrix is adopted to control a data fidelity item;a method for automatically selecting a threshold value in a weighting matrix is provided to ensure the solving accuracy and speed, meanwhile, a scale factor is set to control the weight of the weighting moment, the method is suitable for electrical tomography, an electrical tomography problem is regarded as solving a linear uncertain problem, and an objective function is determined; The method comprises the following steps: obtaining a relative boundary measurement value vector and a sensitivity matrix required by reconstruction according to a measured field domain; Setting initialization parameters; Calculating a gradient and a Hessian matrix of the objective function; Updating the value of the solution by using a Gauss-Newton iteration method; Judging whether iteration is finished or not; And performing imaging according to the finally solved imaging gray value. According to the invention, the quality and image resolution of an electrical tomography reconstructed image are effectively improved, the complexity of the algorithm is simplified, and the applicability of the algorithm is enhanced.

The embodiment of the invention provides a method and a device for classifying texts, a storage medium and a processor, and belongs to the technical field of computers. The method comprises the stepsof performing word segmentation on a to-be-classified text; determining a word vector corresponding to each word obtained by performing word segmentation on the to-be-classified text based on a word vector model, and forming a matrix by the word vectors corresponding to the words belonging to one sentence; processing each matrix based on the sentence vector model to obtain a sentence vector corresponding to each matrix; and processing each sentence vector based on the sentence classification model to obtain a category score vector corresponding to each sentence vector, and determining the typeof the sentence vector corresponding to the category score vector according to each category score vector to realize classification of the to-be-classified text. Therefore, the defects that the constructed word frequency or feature matrix is very sparse and the relationship between words is ignored when the short texts are classified are overcome, and the effect of classifying the texts is improved.