90 results about "Sequence matching" patented technology

The invention relates to the technical field of occupational simulation training evaluation, and discloses a system for college student occupational scene simulation training evaluation. The system comprises a task analysis unit, a dynamic behavior modeling unit, a multi-mode interaction management unit, a state evolution tracking unit, a strategy adaptation unit and a comprehensive evaluation generation unit. The system receives the initial scene description and generates a structured task framework, and behavior parameters are configured for student roles; constructing a multi-modal interaction atlas by fusing aligned voice, action and sight line data in real time; identifying a key decision node, and dynamically tracking a team state evolution path; matching and adjusting strategies in a behavior rule base according to the evolution sequence, and outputting an adaptive intervention scheme; and finally generating a multi-dimensional structured evaluation report. According to the system, deep insight and dynamic and objective intelligent evaluation of complex interaction behaviors in the simulation training process are realized.

The invention relates to the technical field of traffic event identification, and discloses an abnormal traffic event identification method and system based on a traffic large model, and the method comprises the steps: obtaining a traffic data flow, extracting an abnormal feature vector, and obtaining an abnormal signal candidate set; grouping the candidate sets and calculating a deviation degree, and if the deviation degree exceeds a threshold value, taking the deviation degree as a risk signal to form an input subset; environment variables are extracted from the subsets, a mapping relation is established, and anomaly recognition embedding representation is obtained; classifying the embedded representation, judging a congestion precursor and generating an early warning signal to obtain an early warning signal sequence; matching the sequence to obtain an abnormal event chain; if the integrity is higher than a threshold value, analyzing the type to obtain an abnormal event type; extracting a correlation feature vector from the type, pushing the correlation feature vector to a traffic management platform to obtain an instruction, and obtaining an emergency response trigger instruction sequence; and executing the instruction sequence to extract a feedback data stream, inputting the traffic large model to judge the accuracy rate, and if the judgment accuracy rate is met, determining an optimized anomaly recognition framework. The method can solve the problem of insufficient early warning capability.

The invention provides a control method and a control device of an underwater robot. The control method of the underwater robot comprises the following steps: establishing an underwater robot system model; performing obstacle avoidance path planning by using a sampling algorithm to obtain a discrete geometric path point set connecting the starting point and the ending point; a full-state reference sequence matched with the control method underwater robot system model is generated based on the control method discrete geometric path point set; constructing a cost function for trajectory optimization based on the control method underwater robot system model and the control method full-state reference sequence; generating an initial control sequence and an initial state sequence according to the deviation between the current state of the underwater robot in the control method and the full-state reference sequence in the control method; and performing iterative optimization by taking the control method cost function as an optimization target and taking the initial control sequence and the initial state sequence of the control method as hot start initial values to obtain a control instruction sequence of the underwater robot of the control method.

The application discloses a lossless compression storage method and system based on heterogeneous data in a Xinhua environment, first extracts the file type and structure feature mapping relationship by scanning the target file, classifies and groups, and then carries out preliminary lossless compression, further adopts a sliding window matching and sequence matching algorithm to optimize the compressed data for the redundant mode, then generates a metadata structure containing a check code and compression parameters, encapsulates into a data packet supporting cross-platform transmission, finally transmits to the target storage device through a secure write protocol, and verifies the data consistency and recovery path feasibility by using the check code. The application significantly improves the data compression efficiency, ensures the compatibility of cross-platform transmission and the integrity of storage data, and provides efficient and reliable technical support for complex electronic file processing.

This application provides a game testing method, system, and storage medium based on virtual characters. Starting from the first-person perspective of the virtual character under test, it first acquires the game scene currently visible to the player through the visible range, and then obtains scene function data, such as interactive items, NPCs, and quest points. Subsequently, this function data is combined with the virtual character's own information, and a series of simulated operation paths are generated using a behavior planning algorithm. Each path represents a series of actions the player might perform from that perspective, with the expected result pre-set for each step. Finally, each path is executed step-by-step in the game, and the actual results are compared with the preset results to obtain the pass / fail information for each path. In this way, by automatically generating a complete operation sequence matching the player's perspective, it helps to trigger hidden vulnerabilities, thereby improving test coverage and defect discovery efficiency.

The invention relates to the technical field of fault detection, in particular to a building drainage pipe network anomaly detection system which comprises a water pressure chain construction module, a sudden drop identification module, a flow resistance classification module, a structure mapping module and an attribution verification module. According to the method, the continuous change chain is constructed between the water pressure monitoring points of the drainage pipe network, and the spatial correlation analysis is realized according to the water pressure direction and amplitude change of the adjacent nodes, so that the abnormal region can be quickly identified by taking the chain type water pressure gradient fracture as a clue; in abnormal sudden drop identification, timing sequence matching of upstream springback and downstream delay is used as a verification condition, so that abnormal judgment has a dynamic response basis, flow resistance type division is formed through water pressure curve trend segmentation matching, and hierarchy and pipe section attribution positioning of abnormal positions is realized; a closed-loop detection mechanism from signal chain change identification to spatial path confirmation is realized on the whole, so that the drainage system abnormity diagnosis has continuity, traceability and improved positioning precision.

The invention discloses a tumor specific signal analysis device and equipment for uncertain pulmonary nodules and a storage medium, and relates to the technical field of electrical digital data processing.The device constructs a LungTCR database containing lung cancer tissue enrichment type and blood enrichment type CDR3 sequences through TCR beta chain sequencing and IMGT database comparison, and the LungTCR database is used for analyzing the lung cancer tissue enrichment type and blood enrichment type CDR3 sequences. A basis is provided for tumor specific signal recognition; a Needleman-Wunsch algorithm and an editing distance threshold value are adopted for screening, so that high-precision sequence matching is realized; through quantification of lung cancer tissue scores, blood scores and mutation specificity scores, a TCR feature group with biological significance is formed; and finally, the TCRnodseek plus model trained by clinical and image features is combined, a reliable malignant probability prediction value can be output, the model AUC can reach 0.90 or above, and the model is obviously superior to traditional imaging diagnosis and tumor markers.

The invention relates to the technical field of digital identity authentication, in particular to an online insurance policy verification method and system and a storage medium. The method comprises the following steps of: authenticating a user identity by adopting a living body detection and certificate OCR (Optical Character Recognition) technology, identifying tampering by using an image tampering detection model, and generating a structured request data packet through an improved OCR content extraction technology; retrieving benchmark insurance policy data in a distributed insurance policy database, and calling a multi-version data fingerprint sequence from a block chain; comparing the reference insurance policy data with the structured request data packet, and outputting a first verification result; matching the structured request data packet with the multi-version data fingerprint sequence, and outputting a second verification result; and generating a final verification conclusion based on the first and second verification results. According to the method, multiple technologies are fused to construct a whole-process verification system, compared with a single verification mode, the anti-fraud capability is improved through multi-dimensional cross verification, and related requirements of financial supervision on Internet insurance services are met.

The invention relates to the field of lifting intelligent control, and particularly discloses a lifting control method and system based on user habits, which transfer a lifting control trigger source from a single time dimension to a digital behavior of a user, and identify a workflow context with semantic depth by performing atomic classification and sequence matching on an application activity flow. Furthermore, a dynamically updated habit portrait is constructed, and the portrait introduces an enhancement mechanism for a semantic isolation defect: during decision making, not only direct historical preferences are extracted, but also prior knowledge of adjacent scenes is aggregated based on a context graph, and a Bayesian formula is utilized to perform weighted fusion on direct experience and reasoning prior. Therefore, when the real-time working state of the user can be captured, intelligent generalization reasoning is carried out on a new scene which is not seen by using experience of similar scenes, and non-inductive lifting control with personalized accuracy and cold start robustness is realized.

The invention discloses a dynamic quantitative evaluation method and device for comprehensive risks of network public opinions, and the method comprises the steps: constructing an index system of network public opinion risk symptoms, the index system comprising a public opinion symptom index set and a public opinion symptom index association graph; based on an actually occurring public opinion risk event sequence, matching a public opinion symptom index set, generating an incidence matrix according to the public opinion symptom index association graph, and quantifying an association risk value of each public opinion risk event through risk intensity; based on a risk time sequence attenuation model, performing time attenuation calculation on the associated risk value of each public opinion risk event to obtain a real-time risk value; on the basis of all real-time risk values, the comprehensive risk index of the network public opinions is calculated through an improved logistic model, factors in multiple aspects are considered, mathematical modeling is combined to simulate a real scene, and a wait index association diagram, a risk time sequence attenuation model and the improved logistic model are introduced in the calculation process, so that the evaluation result tends to the real situation more.

The invention provides a main and standby switching method and system for a water dispatching automation system, and relates to the field of water dispatching automation and data processing. The method aims at solving the problems that in an existing method, business event flow sequential logic is ignored, and instructions are prone to being disordered and lost due to switching. The method is applied to a framework comprising a main system, a standby system and a time source, and the main system and the standby system are synchronous in time. When executing the water dispatching control instruction, the main system generates a water dispatching event containing instruction content and a timestamp, and stores the water dispatching event in a shared event sequence cache region; the standby system periodically pulls an event, simulates execution after verification by a business rule engine in combination with an associated context, constructs an event node tree diagram with a weight, and prunes based on an actual result; and when the switching is triggered, comparing the actual result of the main system with the execution sequence of the standby system, matching an adaptive path, and avoiding error instruction marks. The method guarantees the time sequence consistency and logic correctness of the event flow, achieves the intelligent seamless switching of a main system and a standby system, and guarantees the continuous and stable operation of a water dispatching service.

This invention discloses a dynamic data security assessment system and protection method, belonging to the field of information security technology, and is used to solve the problem of dynamic collaborative protection of terminal device access control and behavioral risks. It generates fingerprints based on terminal device characteristics and verifies authorization. Authorized terminal devices input access feature slices into a time-series model. A trust score is generated based on the cosine similarity between the behavioral sequence and the predicted sequence. A Markov model is then constructed to dynamically select the authentication method. A risk index is calculated by combining network and geographic information and assigned to the real or virtual environment. Unauthorized terminal devices are assigned to a virtual environment after registration. Differential monitoring is implemented: a time-series database is established for virtual environments, and a suspicious score is generated through sequence matching; for real environments, a state deviation is calculated using a Hidden Markov Model and multi-dimensional features are fused to generate a risk score. Based on result-based access control, suspicious terminal devices have their permissions restricted, compliant terminal devices undergo authentication migration, and access is controlled according to the principle of least privilege and a time window mechanism.

The invention discloses a link log sampling method and device, equipment and a medium. The method comprises the steps of obtaining a target log set, wherein a target log comprises at least one burying point and original business data; processing each target log according to each burying point in the target log to obtain a target sequence set, the target sequence set comprising at least one target sequence and a sequence type matched with the target sequence; processing each target sequence in the target sequence set based on a pre-trained sequence pattern mining algorithm to obtain at least one service sequence type; and obtaining a sampling log set, and performing screening operation on each sampling log in the sampling log set according to each service sequence type to obtain a result log set. Through the technical scheme of the invention, sampling of the link logs can be realized, the efficiency of log acquisition work is improved, and the comprehensiveness, accuracy and reliability of an acquisition result are improved at the same time.

The present application relates to the technical field of railway traffic, in particular to a detection vehicle accurate positioning system and method based on multi-channel wireless transmission, which comprises a positioning sending unit, a positioning receiving unit and a detection analysis unit; the detection analysis unit matches and calculates the real-time collected train running distance and track circuit code sequence information with the theoretical path length and expected track circuit code sequence in the line account database to automatically execute route switching decision; the present application realizes the automation of route switching based on the route decision algorithm of train real-time running distance deviation, track circuit code sequence matching and route priority, avoids the lag and misjudgment of artificial presetting, solves the problems of route switching dependence on artificial operation, low efficiency and easy positioning deviation; the present application improves the matching precision of positioning data and line account by combining the multi-dimensional route matching of track circuit code sequence and mileage parameters.

The application discloses a kind of visual inertial SLAM loop detection method and system based on geomagnetic information optimization, first to the data collected is solved coarse matching on low resolution level and obtains first geomagnetic sequence matching result, then first geomagnetic sequence matching result is optimized in local neighborhood and obtains second geomagnetic sequence matching result, then based on second geomagnetic sequence matching result introduces key frame image and calculates the modulus of geomagnetic intensity corresponding to the time stamp of key frame image;The application realizes the function of first coarse screening and then fine matching in loop detection with the way of deep fusion of high-efficiency geomagnetic sequence matching algorithm and visual key frame, and through Fast-EDLines algorithm, gradient calculation acceleration, edge detection, line extraction and verification can be carried out, ensuring the accuracy and high confidence of line feature extraction, and through the use of FastDTW algorithm and geomagnetic sequence matching and dynamic adjustment of key frame buffer, efficient and robust loop detection can be carried out, which is suitable for wide promotion and use.

A unit name consistency discrimination method, comprising: constructing a standard name library, dividing each standard unit name and setting a category; obtaining a user input name, dividing the input name and setting a category; constructing a consistency judgment library; constructing a combined name set for each standard unit name in the consistency judgment library; calculating the cosine similarity of the input name and each standard unit name, and selecting the maximum cosine similarity and the corresponding standard unit name therefrom; using a sequence matching method, calculating the sequence similarity of the input name and each standard unit name, and selecting the maximum sequence similarity and the corresponding standard unit name therefrom; judging whether the standard unit name with the maximum cosine similarity and the standard unit name with the maximum sequence similarity are consistent, and if so, the standard unit name is consistent with the user input name. The present application belongs to the field of information, and can accurately identify, match and obtain a standard unit name consistent with the user input name.

The invention provides a pulse timing method and system based on mixed supervised learning and electronic equipment, the method comprises a self-supervised learning part and a weak supervised learning part, and the self-supervised learning part comprises the following steps: carrying out random sub-sequence sampling on sampling waveforms of an analog-to-digital converter at the same or lower sampling rate to obtain sampling sub-sequence matching pairs, recording a label of a self-supervised learning loss function by using the relative position of the sampling sub-sequence matching pair, inputting the sampling sub-sequence matching pair into a neural network, and calculating self-supervised loss; the weak supervision learning comprises the steps of obtaining a reference time sequence and a timing label, inputting the reference time sequence into a neural network, and calculating weak supervision loss; and finally, forming a mixed supervision loss function by the sum of the self-supervision loss and the weak supervision loss, and jointly applying the mixed supervision loss function to neural network reverse conduction training. According to the method, the neural network timing model can be trained under the condition that only a small number of labeled samples are provided, and the purpose of high-precision pulse timing is achieved.

This invention provides a network attack detection method and system based on distributed intelligent probes. It receives real-time traffic data synchronously collected by distributed intelligent probes at various network nodes, models the interaction relationships between nodes in the real-time traffic data, identifies traffic communication patterns between different network nodes, and generates a traffic correlation graph containing node connection relationships and communication frequencies. Based on the traffic correlation graph, it mines abnormal communication paths, extracting node communication sequences with abnormal patterns by analyzing the deviation degree of node connection relationships and the fluctuation characteristics of communication frequencies. It then performs pattern matching processing between the node communication sequences and attack behavior templates in a preset attack feature library, calculates the sequence matching similarity, and generates an attack matching score set. Finally, it determines the network attack type and the location information of the attack source node, generating a network attack detection result. This invention improves the practicality and effectiveness of network attack detection.

The application discloses a security audit and high-risk event mining method and system based on TDS protocol log and flow control, and relates to the technical field of information security.The application comprises the following steps: performing protocol analysis, service shunting, executing data preprocessing, statistical analysis, constructing a feature baseline, and extracting behavior features to construct a behavior feature operation chain to establish an operation chain library; online matching the real-time received TDS protocol ticket service data with the operation chain library to determine the legality of the flow; based on the online matching result and the feature baseline, auditing the compliance of the ticket service, mining high-risk events, and outputting the audit result.The application effectively reduces false positives and false negatives by combining behavior sequence matching and statistical feature analysis of different dimensions of detection technology, and realizes the security protection of ticket services including but not limited to gate control, ticket sales, and automatic sales through the whole-process closed loop of "flow control shunting-offline training-online matching-security audit".

The invention relates to the technical field of intelligent system architecture design, and discloses an intelligent system architecture design method based on hierarchical collaboration and multi-model fusion. The method comprises the steps of generating a task situation fingerprint capable of accurately representing a task resource demand mode by analyzing a cross-level interaction event flow generated during system operation; based on the fingerprint, a dynamic resource allocation strategy is established to predict and cope with future resource competition, a dynamic weaving process of architecture configuration is started, and a customized architecture topology matched with the task sequence is generated. By injecting a resource elastic adjustment mechanism into the topology, the data flow load is monitored in real time, the resource quota of the model instance in each level is dynamically adjusted, and continuous incremental optimization is performed according to the actual resource utilization rate. According to the method, fine management of computing resources and real-time on-demand reconstruction of a system architecture are realized, and the resource utilization efficiency and the self-adaptive capability of the overall performance of the system in a complex task scene are improved.

This application relates to the field of medical data processing technology and discloses a method and system for screening infertile patients with polycystic ovary syndrome (PCOS). The method includes: acquiring data on the number of antral follicles, anti-Müllerian hormone (AMH) concentration, and ovarian volume from multiple patient monitoring cycles; performing feature extraction and normalization to obtain a set of normalized feature vectors; determining the synchronous relationship between the number of antral follicles and ovarian volume based on the vector set, and identifying signs of asynchronous changes; combining hormone concentrations for joint grouping to obtain the distribution of abnormal indicator clusters and identifying ovarian function stage markers; performing sequence matching based on the stage markers of continuous cycles and historical medical records to generate an ovarian function progression trajectory curve; extracting inflection point features from the curve and performing causal association rule mining to construct a logical chain between abnormal indicators and clinical outcomes; and determining conditions layer by layer according to the logical chain to output the final screening results. This application improves the accuracy of screening infertile patients with PCOS.

The invention discloses a power distribution secondary equipment operation state intelligent discrimination method based on multi-dimensional signal deep analysis, and relates to the technical field of power system state monitoring, and the method comprises the following steps: in a short-time parallel operation state of a power distribution network, uniformly expanding a zero-sequence protection signal and an overcurrent protection signal on the same time scale, constructing a combined time chain for capturing transient jump; and positioning a plurality of time points where the action directions deviate on the joint time chain, and connecting the continuous time points where the action directions deviate in series according to a time sequence to form a conflict time zone for marking a protection signal direction divergence region. According to the method, by constructing the joint time chain and the conflict time zone, the abnormal action trend is accurately recognized, blocking before tripping is achieved in combination with reverse superposition sequence matching, and misoperation is avoided; after the opening signal is frozen, the opening signal is introduced into a breathing type coordination channel, and the impulse current is dispersed and the contact loss is reduced through amplitude-limiting probing and delayed flexible closing signal alternative output and micro-scale dislocation control.

The invention provides a liquid chromatography-mass spectrometry combined oligonucleotide sequence confirmation method, and relates to the technical field of instrument analysis. The confirmation method comprises the following steps: detecting a sample solution containing oligonucleotide by adopting an IPRP-UPLC-MS / MS method to obtain fragment ion data; and processing the fragment ion data by adopting CONFIRM Sequence to obtain a sequence matching rate. According to the method, IPRP-LC-MS / MS is adopted to detect a sample solution, CONFIRM Sequence is adopted to process data, and the sequence matching rate is obtained. The method disclosed by the invention adopts an IPRP high performance liquid chromatography and QTOF MS combined method for analysis, has the advantages of short analysis time, strong directness, high analysis precision, high resolution, high accuracy, simplicity in operation and the like, and achieves the purpose of identifying the oligonucleotide sequence.

The application discloses a zero sample video classification method based on frame-level cache and video-level cache updating. First, visual features and text features of a test video are extracted. Then, cosine similarity calculation is performed on the test video and text prototypes of each category to obtain cross-modal zero sample matching results, which contain predicted pseudo labels and prediction result entropy. According to the prediction result entropy and time sequence probability difference criterion, the credibility of each frame is measured, and a two-stage Top-K strategy is adopted. First, high-confidence frames satisfying a preset value are selected according to the prediction entropy, and then frames most discriminative in time sequence are further selected according to the time sequence probability difference for caching. A video-level cache updating module is used to maintain the cache. Visual features of the test video are matched with visual prototypes in the cache of each category to obtain same-modal matching results. The cross-modal zero sample matching results and the same-modal matching results are fused to obtain final inference results.

The invention relates to the technical field of rehabilitation follow-up visit management, and discloses a multi-dimensional data-based follow-up visit scheme dynamic adjustment method and system, and the method comprises the steps: obtaining historical rehabilitation index data and historical follow-up visit frequency records of a subject, and carrying out the dynamic analysis of a time sequence, and obtaining an index fluctuation mode and a frequency accumulation effect value; carrying out causal inference processing to obtain a causal association path and a weight; extracting a real-time rehabilitation index state according to the path, and performing risk prediction when fluctuation exceeds a limit to obtain an evaluation result; performing frequency arrangement sequence causal simulation in combination with the features, and determining a target follow-up sequence of which the target deviation is smaller than a preset threshold value; matching historical tracks, and performing node analysis and opportunity missing risk assessment to obtain an instant adjustment instruction; and finally, retrieving the resource record and carrying out continuity verification, and outputting a personalized follow-up scheme. According to the method, accurate allocation of follow-up visit resources and closed-loop adaptive optimization of individual rehabilitation tracks can be realized.

This invention discloses a method, apparatus, device, medium, and product for large-scale model inference. The method includes acquiring the execution sequence of target kernel functions matching the currently input large-scale model inference task; completely distributing logic based on the target startup descriptor package matching the execution sequence to obtain all required kernel function description information; generating a logical task graph matching the large-scale model inference task based on the kernel function description information corresponding to each target kernel function; instantiating the graph to obtain an executable task graph; and during the large-scale model inference process, distributing the target startup descriptor packages obtained from the executable task graph to the device to generate and output each lexical unit matching the large-scale model inference task. This technical solution effectively reduces the resource overhead of inference task execution and improves the scheduling efficiency of inference tasks.

The application discloses a glass cup washing score evaluation method and system for testing the performance of a dishwasher, comprising: obtaining a glass cup washing sample of a dishwasher to be tested, collecting images of the washing sample based on a preset washing score evaluation system, obtaining surface images of the washing sample, extracting a background axis region image, distinguishing two modes of coincident area subset matching and full sequence matching between adjacent images, discriminating and removing stain residual points of repeated residual areas, calculating physical areas and point numbers of independent residual areas, and outputting washing performance scores. The method uses stain fluorescence to overcome the interference of fully transparent materials and surface texture characteristics on imaging, improves the collection quality and image clarity of fluorescence signals, and removes repeated images, thereby improving the precision and objectivity of dishwasher washing performance detection.

The application discloses an English spelling pronunciation sequence matching method based on intelligent pronunciation, and relates to the technical field of intelligent pronunciation recognition. The method extracts a target English word and intelligent pronunciation input by a learner for the target English word, carries out letter cluster segmentation to obtain a letter cluster sequence, and carries out preprocessing on the intelligent pronunciation to generate a pronunciation fragment sequence. The letter cluster sequence is written into a letter cluster record, field mapping is performed to generate a plurality of anchor domain units, a letter cluster pointer is established to perform character label consistency comparison, and a pronunciation fragment writing position is recorded synchronously to form a letter cluster anchor domain synchronous chain. Adjacent slot records are read, and a merging determination is performed by taking a fragment form label of a pronunciation fragment in an adjacent slot and a letter cluster type of a next anchor domain unit as a determination condition. After the merging determination is completed, a letter cluster matching record is generated for each anchor domain unit in sequence, and a matching result is determined according to an anchor slot writing state, a letter cluster type and a merging mark.

The invention is suitable for the technical field of computer vision, and provides a GraphMama-based dual-view image matching method, which comprises the following steps of: constructing a model which comprises an InitialProjection unit, a Preminarial Local ContextEncoding unit, a Consensus-Aware Learning Block unit, a GraphMama unit and an Inlier Predictor unit, and constructing a model which comprises an InitialProjection unit, a Preminarial Location unit, a Consensus-Aware Learning Block unit, a GraphMama unit and an Inlier Predictor unit; the method comprises the following steps of: performing high-dimensional feature mapping, local neighborhood enhancement, global consistency optimization and long-range dependence modeling by taking initial matching point pairs of a double-view image extracted by SIFT (Scale Invariant Feature Transform) as input, and performing Consensus-Aware Learning Block and GraphMama iteration for four times; and training the model by using supervised data sets such as YFCC100M, SUN3D and the like, and finally outputting correct matching point pairs and an essential matrix. According to the GraphMama-based double-view-angle image matching method, through the limitation that traditional MLP independently processes matching pairs and RANSAC robustness is insufficient, long-range dependence of long-sequence matching pairs is accurately captured, on YFCC100M and SUN3D data sets, mAP indexes of relative attitude estimation and feature matching are remarkably superior to those in the prior art, complex scenes such as view angle changes and shielding can be dealt with, and the method has the advantages of being high in adaptability, high in robustness and the like. The method can adapt to computer vision tasks such as image stitching and vision positioning.

The invention discloses an information management method and equipment for accessing urban merchants to a cloud platform, and a medium, and relates to the technical field of merchant information management, and the method comprises the steps: calculating the credit scores of all merchants according to a merchant portrait vector, screening a candidate merchant set from all merchants of the platform through a multi-arm bandit algorithm according to a user portrait vector, and carrying out the selection of the candidate merchant set; collecting user situation information, inputting the user situation information, the merchant portrait vector and the user portrait vector into a pre-constructed multi-target matching function for calculation to obtain a comprehensive matching degree, sorting and screening the candidate merchant set according to the comprehensive matching degree, obtaining a merchant priority pushing sequence, and according to credit scores of all merchants, obtaining a merchant priority pushing sequence; a sequence matching platform resource allocation strategy is preferentially pushed for the commercial tenants, and commercial tenant information intelligent management and pushing are carried out on the matched platform resource allocation strategy; according to the urban commercial service management method, honest operation and high-quality development of urban commercial tenant ecology are promoted.