75 results about "Bipartite graph matching" patented technology

The invention relates to the technical field of Internet of Things management, in particular to a lightweight Internet of Things management system based on a star flash protocol stack, which comprises a plurality of modules such as a star flash protocol communication module, a dynamic topology management module and a resource virtualization module. The satellite flash protocol communication module realizes low-power-consumption efficient connection of equipment; the dynamic topology management module optimizes the network topology by using an improved multi-agent Q learning algorithm; the resource virtualization module realizes accurate resource allocation through a weighted bipartite graph matching model; the safety protection module adopts an attention mechanism to detect abnormities and guarantee data safety; the edge co-processing module realizes intelligent task unloading by means of deep reinforcement learning; all the modules cooperatively work under the overall planning of the cross-module coordination controller, and the information barrier is broken. According to the invention, the problems of high communication energy consumption, poor resource allocation, weak security protection, low task processing efficiency, insufficient module collaboration and the like of a traditional Internet of Things system are effectively solved, the system communication efficiency, the resource utilization rate and the security are remarkably improved, and the task processing capability is enhanced.

The invention provides an all-weather autonomous inspection method and system based on cluster task dynamic load balancing, and relates to the technical field of inspection monitoring, and the method comprises the steps: collecting the remaining electric quantity, the positioning position, the task queue length and the sensor health state of an unmanned aerial vehicle cluster in real time through an airborne terminal; fusing the meteorological data and the multi-source environment sensing data, and constructing a dynamic obstacle map and a meteorological influence model; dividing an inspection area into a plurality of sub-areas based on an electronic fence, dynamically adjusting the inspection priority of each sub-area, setting task weights of a water taking head and a raw water pipeline facility, and distributing inspection tasks according to the priorities; based on the remaining power of the unmanned aerial vehicle, the task priority, the current load and the meteorological data, a bipartite graph matching model of the unmanned aerial vehicle and the task is constructed, the matching weight is dynamically calculated, the optimal matching distribution of the unmanned aerial vehicle and the inspection task is completed, and the efficient inspection operation of the unmanned aerial vehicle cluster in the complex environment is realized.

The invention discloses an intelligent port operation vehicle scheduling system and a scheduling robot, and relates to the technical field of intelligent port operation. The problems that in traditional port vehicle scheduling, the manual scheduling response is slow, the error rate is high, resource distribution is uneven, the labor cost is too high, and the efficient and intelligent requirements of modern ports are difficult to meet are solved. The vehicle no-load distance is reduced and the heavy load rate is improved by using a bipartite graph matching strategy, the prediction scheduling module plans in advance, task overstock and vehicle idleness are reduced, the operation efficiency is effectively improved, the constraint processing module optimizes task allocation according to task and vehicle conditions, and the multi-target optimization module gives consideration to the heavy load rate and order dispatching fairness. The scheduling robot integrates system functions, and can adapt to different working environments and realize intelligent scheduling through data acquisition and continuous optimization of an online learning technology, so that the scheduling accuracy and efficiency are improved, the labor cost is reduced, and the overall competitiveness of a port is improved.

The invention relates to a standardized project management and intelligent professional scheduling system, and belongs to the technical field of project management and human resource intelligent scheduling. The project standardization module is disassembled into standardization task steps through a business process association rule mining algorithm, and project complexity and resource tensity are adapted by using a task attribute dynamic weight algorithm; the archive matching module constructs professional archives, extracts feature vectors through a multi-criterion decision-weighted bipartite graph matching algorithm, and generates optimal matching pairs in combination with adaptive weight adjustment and an integer linear programming model; the dynamic scheduling module plans a task execution scheme according to a resource constraint scheduling mechanism, and realizes efficient scheduling in combination with a skill supply and demand prediction algorithm and calendar integration; and the quality optimization module adopts a deliverable anomaly detection algorithm to monitor compliance, feeds back an iterative matching and scheduling strategy through a time sequence prediction optimization algorithm, and perfects a skill map based on a map increment updating algorithm. The system realizes a project full-process closed loop.

The invention discloses a task dynamic decomposition management system and method based on an AI framework, and relates to the technical field of task management, and the method comprises the steps: dividing task parameters into a plurality of dimension features, carrying out the fusion of the multi-dimension features through a space-time attention network, and constructing a three-dimensional feature tensor; recursively decomposing the three-dimensional feature tensor through an adaptive secondary screening algorithm, and outputting a decoupled modal set; predicting a resource load and quantifying a modal demand level based on a long short-term memory network, constructing a bipartite graph matching model, and adaptively allocating resource instances through dynamic matching; time performance and resource consumption data of task execution are collected in real time, a modal deviation value is calculated, if the modal deviation value exceeds a deviation threshold value, dynamic re-decomposition is triggered, resources are reallocated based on an updated modal set, and dynamic capture, efficient decomposition and accurate resource allocation of task features are achieved; therefore, the task execution efficiency and the resource utilization rate are improved.

The invention discloses a single tree trunk structure extraction method and system based on deep learning, and the method comprises the steps: obtaining two-dimensional image data of a single tree, marking the two-dimensional image data, and constructing a single tree trunk segmentation data set; constructing a spatial domain and frequency domain double-branch network based on the segmented data set, respectively extracting spatial domain features and frequency domain features, and fusing the double-branch features to generate a coding feature map; the coding feature map is decoded, in the decoding process, coordinate convolution CoordConv is adopted to enhance position perception, and mask segmentation and semantic label extraction are respectively carried out through a dynamic mask reconstruction branch DRMask Branch and an instance branch Inst Branch; based on a bipartite graph matching strategy, associating results of the mask segmentation and instance branches, and realizing segmentation of a single tree trunk and matching of instance-level labels; according to the method, the boundary precision and the detail reconstruction capability of trunk segmentation are remarkably improved, and the problem of feature loss of a traditional method in a complex under-forest environment is solved.

The invention discloses a new energy power station inspection method based on a multi-modal large model small sample open set, and the method constructs a multivariate text learnable prompt, and guides a detection model to form a finer-grained category decision boundary through the text information related to an aggregation task. Due to the lack of real unknown class samples in the training process, the mining of unknown class pseudo samples is regarded as a bipartite graph matching task for the first time, and the model is optimized to form a compact unknown class decision boundary by adding unknown class virtual nodes, constructing a cost matrix and mining the unknown class pseudo samples. In order to solve the problem that known classes and unknown classes of small samples are prone to confusion, the method proposes unknown class optimization loss based on cost perception, considers the classification and positioning quality of the unknown classes, and improves the open set detection performance of the model. According to the method, transformation from a single-mode vision small model to a multi-mode vision large model is realized, and good generalized small sample open set target detection performance can be obtained only by a small amount of training data.

The invention discloses an end-to-end category level object pose estimation method based on space-time implicit anchor point query, which comprises the following steps of: inputting an RGB (Red, Green and Blue) image sequence into a visual encoder, extracting multi-scale semantic features and fusing depth and normal vector geometric features to generate 2.5-dimensional multi-scale features; a 3D implicit anchor queue is established based on the features, and queries in the queue are processed through confidence screening and camera pose offset transformation. Performing modeling on space-time correlation by using multi-head attention, outputting time sequence enhanced query features, and performing geometric perception feature sampling; and finally, mapping the implicit 3D query into 9-degree-of-freedom attitude parameters through bipartite graph matching to realize end-to-end training.

The invention relates to the field of relation triple extraction, in particular to a relation triple joint extraction method based on information enhancement and bidirectional modeling. According to the method, through entity-to-relation and relation-to-entity double-branch collaborative modeling, semantic information is enhanced through double-branch potential information complementation: a bipartite graph matching method is combined with subjects and objects extracted by an entity extraction module, and relation classification is optimized through an entity boundary mask attention enhancement method; a potential relation extraction module is used for guiding subject and object entity extraction, and entity boundary information is utilized to crosswise mask attention of a large related area; the two branches filter redundant triads through a relation triad cutting module; combining bidirectional results and outputting a complete triple set; according to the method, a bidirectional interaction method is introduced to realize mutual enhancement of entities and relationships, and attention distribution is optimized in combination with entity boundary masks, so that the problem of high dependence on an initial extraction result caused by unidirectional modeling in traditional joint extraction is effectively solved.

The invention provides a character interaction detection method based on open vocabularies in an unmanned aerial vehicle scene, and the method comprises the steps: extracting global features of an image through employing a pre-trained CLIP visual encoder, segmenting the image into a plurality of image blocks, and carrying out the coding of the image blocks; an MLR module is introduced to extract global context information, character interaction is decoded from a multi-level feature map, and when bipartite graph matching is carried out between a prediction result and a real result, a loss function is designed to guide a low-level feature map to correspond to a character pair with a small distance and guide a high-level feature map to correspond to a character pair with a large distance; a large language model is used for generating human body part state description related to character interaction in an image, character interaction category names and the state description of the related human body parts are coded into text embedding, and the embedding is combined with the output of a character interaction decoder and global context information extracted by an MLR module to obtain an instance-level detection score. According to the invention, the model can better adapt to interaction detection requirements under different distances.

The invention discloses an online vectorization high-precision map generation method based on point features, which is characterized in that real-time high-quality generation of an online vectorization high-precision map is realized by utilizing a Point MapNet network, and the Point MapNet network is obtained by improving a BEV feature coding module and a map element decoding module of an original MapTR network; the BEV feature coding module is improved in the following steps: replacing BEV features sampled at fixed positions with position learnable point features, naming the improved BEV coding module as a point feature coding module, enabling the point features to learn a spatial position containing a detection object in a sensing area through bipartite graph matching and distance loss calculation, and updating the position of the point features; the map element decoding module is improved as follows: attention masks based on distance are used, so that map element features are only interacted with point features of a near space. Through the improvement, the reasoning speed is increased and the training cost is reduced while higher map generation quality is achieved.

The invention provides an online driver and passenger matching method, a medium and equipment, belongs to the technical field of online bipartite graph matching, and provides a set of strategy architecture based on hierarchical reinforcement learning to predict future information to guide a current decision, different underlying strategies are trained firstly, and then different underlying strategies are trained; and training a high-level model to select a bottom-level strategy to optimize the long-term total revenue. The earnings and state transition of the matching time slices follow specific rules, the states comprise driver and order features, the action decides whether to change the underlying strategy through a stop strategy, and if the underlying strategy is changed, a proper underlying strategy is selected by a high-level strategy. According to the method, a layered reinforcement learning architecture is adopted, the problem of balancing training efficiency and decision performance is solved, decision complexity is reduced, training efficiency is improved, it is ensured that high-quality decisions can be made quickly and accurately in a complex dynamic environment, and the overall performance of the system is improved.

The invention relates to the technical field of internet data processing and resource scheduling, and discloses an intelligent order sending method and system for housekeeping service and a storage medium, and the method comprises the steps that a server side receives original order information containing house investigation data; calling a spatial granularity decomposition model to analyze the order into independent operation area sub-tasks, and calculating the lowest skill level demand in combination with the feature vector; attribute vectors such as comprehensive skill levels and real-time coordinates of service personnel are obtained; constructing a dynamic bipartite graph matching model by taking the subtask set and the attributes of the service personnel as input; an optimal mapping relation list is determined by solving the maximum value of an objective function, and a dispatching instruction is sent. The method further comprises an attendance checking judgment step based on a geofence and a closed-loop iteration step of dynamically updating skill levels according to attendance checking and evaluation data. According to the method, refined disassembly and global optimal scheduling of housekeeping tasks can be realized, and the resource allocation efficiency and the service quality are remarkably improved through a dynamic closed-loop mechanism.

The invention discloses a timing sequence statement positioning model training method and device based on proposal selection and anchor point distribution, and relates to the technical field of timing sequence statement positioning. The method comprises the following steps: initializing a fixed number of learnable queries according to a first static anchor point set; on the basis of the first learnable query, according to the unpruned long video and the natural language description text, performing proposal generation through a time sequence statement positioning model; based on the proposal selection module, redundant proposal filtering is carried out by using a non-maximum suppression algorithm; based on an anchor point distribution module, according to the untrimmed long video, the first static anchor point set and the candidate proposal set, performing bipartite graph matching by using a Hungary algorithm; and carrying out loss function weighting calculation according to the unpruned long video and the optimal proposal set, and carrying out parameter optimization on the time sequence statement positioning model to obtain an optimized time sequence statement positioning model. The method is an efficient and accurate time sequence statement positioning model training method combining proposal selection and anchor point distribution.

The invention discloses a traditional Chinese medicinal material defect detection method based on machine vision, relates to the technical field of intelligent detection, and solves the problems that in the prior art, a threshold cannot be migrated, so that an inter-batch judgment standard is unstable, and high-resolution block reasoning causes boundary pseudo defects and instance wrong affiliation. According to the method, a reference area is constructed in a batch domain, a statistical descriptor is extracted, and monotone mapping calibration is performed on defect scores so as to realize cross-batch unified judgment; meanwhile, a seam continuity field is introduced in the image block splicing process, defect connected domains are fused through a cross-window adjacency relation, and accurate attribution of defects and medicinal material pieces is achieved based on bipartite graph matching; according to the method, the judgment stability and the statistical reliability in the traditional Chinese medicinal material defect detection process are remarkably improved.

The invention discloses a multi-source geographic vector data matching method based on topological relevance, which comprises the following steps: collecting multi-source basic geographic vector data to form a preprocessed data set; constructing multi-index data quality evaluation based on local topological features; extracting and classifying spatial topological relation edges; constructing a weighted adjacent matrix of the target map; a neighborhood topology consistency measurement index is calculated, and neighborhood topology consistency quantization and adaptive weight fusion are carried out; establishing a global optimization objective function by taking the confidence coefficient matrix as input, solving a maximum weight bipartite graph matching problem with topology constraints by taking the topology retention rate as a hard constraint condition, and outputting a global optimal matching scheme; and constructing a quantitative evaluation index topology deviation measure based on topology invariant difference, judging whether a matching scheme has a significant topology conflict, and performing hierarchical repair according to the conflict condition. According to the method, the reliability and logic self-consistency of multi-source vector data fusion are improved, and the key problem of topological contradiction in the prior art is solved.

The invention provides a discrete manufacturing intelligent scheduling method based on a graph theory. The discrete manufacturing intelligent scheduling method comprises the steps of 1, constructing a heterogeneous graph model of scheduling elements; step 2, static scheduling optimization based on a critical path method; step 3, resource allocation optimization based on a multi-resource bipartite graph matching method; and 4, dynamic response and rescheduling are carried out. Based on a graph theory method, a complex'process-resource-constraint 'relationship is converted into a visual heterogeneous graph model, and a systematic scheduling solution based on the graph theory is constructed, so that the scheduling solution is suitable for a discrete manufacturing scene with multi-process, multi-equipment, multi-constraint and dynamic disturbance characteristics; the method is used for realizing static planning of production plan scheduling, resource optimization distribution and dynamic adjustment full-process optimization.

The invention belongs to the technical field of energy storage system operation and maintenance and data diagnosis, and particularly relates to a highway zero-carbon service area-oriented energy storage data diagnosis method and device, and the method comprises the steps: carrying out the real-time monitoring of an energy storage system asset SN set, channel-asset binding and channel physical feature data, recognizing an abnormality, and triggering self-verification; identifying a natural or artificial anchor point through a charging session and grid-connected power steady-state constraint, determining a time window, extracting channel fingerprint features, matching an asset SN historical feature template according to an operation mode, and calculating a normalized difference degree verification binding relation; when mismatching is bound, an optimal rematching scheme is solved through bipartite graph matching according to physical topology constraints, independent verification is carried out through a station-level power closing auditing model, a diagnosis baseline is reconstructed in a layered mode after verification is passed, and an alarm and diagnosis conclusion is output by adopting a hierarchical gating strategy in combination with multiple states. The alarm accuracy is improved, misdispatch work orders are reduced, the diagnosis conclusion can be traced and audited, and stable support is provided for zero-carbon operation accounting.

This application belongs to the technical field of data annotation, and in particular, relates to a method, device, and equipment for annotating point cloud data based on rules. The method includes: obtaining annotation data; determining whether there is original data for the annotation data, where the original data includes existing annotation information; if there is no original data, identifying the data type in the annotation data and annotating the annotation data in the form of an annotation box; selecting the annotation box as the father box, obtaining a prediction result according to a preset rule, and annotating the son box according to the prediction result, and outputting it as the task data annotation result, where the prediction result is that other annotation boxes that may have a father-son relationship are automatically associated with the father box by the preset rule; if there is original data, classifying the annotation data according to the original relationship of the annotation boxes in the annotation information in the task data; establishing a bipartite graph for matching according to the classification result; and outputting the annotation result according to the bipartite graph matching result, which can improve the annotation speed and the annotation accuracy.

The present invention proposes a safety helmet detection method based on noise cancellation training and dimensionality reduction attention mechanism. First, a safety helmet detection dataset is constructed by acquiring safety helmet data pictures and preprocessing. Then, a noise cancellation training framework is constructed to replace the bipartite graph matching algorithm, noise is added to the ground-truth information of each target object, an attention mask is introduced, and the noise objects are grouped. Finally, the noise objects are input into the Transformer decoder. In addition, the present invention improves the multi-head attention mechanism in the Transformer architecture, decouples the key features into one-dimensional row features and one-dimensional column features through one-dimensional global average pooling, then performs row attention and column attention in sequence, and finally performs weighted summation. The present invention uses sub-feature fusion and cross-layer perception enhancement convolutional modules to construct a lightweight feature extraction network, and adds it as a backbone network to the DETR network based on the Transformer architecture to obtain a lightweight DETR detection network.

The present invention discloses a method and system for matching multi-sensor perception information of a vehicle based on graph matching. The method includes the following steps: S1. Construct an original tripartite graph of a lidar, a camera, and a millimeter-wave radar, and divide the original tripartite graph into 3 subspaces based on the branch and bound strategy and the Lagrange multiplier; S2. Determine the vertices and edges of the weighted bipartite graph within the subspaces, and construct a cost matrix model; S3. Analyze the perception error from the spatial dimension, and construct a lidar perception error model and a camera perception error model based on a statistical method; S4. Obtain a multi-sensor perception information matching form based on the cost matrix model, the lidar perception error model, and the camera perception error model. The present invention overcomes the limitations of traditional bipartite graph matching algorithms, improves the accuracy and robustness of sequential bipartite graph matching, and thus provides more reliable environmental perception and decision support for intelligent vehicles.

The invention discloses a monocular vector high-precision map data association method based on geometric context, and relates to the technical field of map data association. Extracting a map landmark set in a current frame view range from the map through the initial vehicle pose, and projecting a point set to a pixel plane; performing structured modeling on each point set in the detection road sign set and the projection road sign set; constructing a cost matrix for the current estimated vehicle pose; introducing a gating mechanism based on map distance and a self-adaptive gating range based on map depth, and eliminating matching pairs with inconsistent geometric structures; complete data association is obtained through cost matrix calculation, optimal bipartite graph matching is performed on the cost matrix by adopting a Hungary algorithm, and a data association relationship meeting global optimum is obtained. According to the method, feature information of road signs in a vector high-precision map is fully utilized, outlier matching relation pairs are eliminated through self-adaptive gating, finally, data association is conducted through a Hungary algorithm, and matching robustness and accuracy are improved.

The invention provides a super-large-scale MIMO system beam focusing method based on an adaptive time delay-phase structure, and relates to the technical field of radio communication, comprising the following steps: constructing an adaptive time delay-phase structure suitable for a super-large-scale MIMO system, and establishing a near-field beam focusing joint optimization mathematical model; solving an all-digital optimal beam focusing matrix through maximum ratio transmission; a switch matrix optimization problem is converted into a bipartite graph matching problem, and an optimal matching matrix is solved by using a Hungary algorithm; solving a phase shift matrix through a Riemann spectrum conjugate gradient method; optimizing the time delay matrix by using a gradient descent method; optimizing the digital precoding matrix by using a least square method; and giving a base station beam focusing matrix through alternate iteration, and completing beam focusing of the super-large-scale MIMO system. According to the structure provided by the invention, dynamic sharing with the antenna array elements can be realized by utilizing the switching network, so that the adaptive capacity of hybrid precoding to a channel environment is effectively enhanced.

The invention discloses a cutting track planning method and device and a storage medium, and the method comprises the steps: building a bipartite graph model of a workpiece graph set and a threading hole set, and building a weight matrix based on an edge set in the bipartite graph model; based on the weight matrix, performing optimal matching on a workpiece graph node set and a threading hole node set by adopting a bipartite graph matching algorithm, and determining a matching pair between the workpiece graph and the threading hole according to nodes which can be matched with each other; performing type division on the workpiece graph according to the matching pair to obtain a male die and female die workpiece graph set, and combining the cutting starting point position of the workpiece graph to obtain a first corresponding relation between the female die workpiece graph and the cutting starting point position and a second corresponding relation between the male die workpiece graph and the cutting starting point position; and according to the obtained first and second corresponding relations, carrying out cutting track planning on the workpiece graphs in the female die and male die workpiece graph set. Therefore, batch programming of a plurality of male dies and female dies is realized, and the efficiency of generating a machining program is improved.

The invention provides a resource target allocation method of a double-layer model based on bipartite graph matching. The method comprises the following steps: step 1, establishing a multi-target static resource allocation problem model; the multi-target static resource sub-problem allocation model comprises a target function and a constraint condition; 2, establishing a bipartite graph matching optimization model; and step 3, performing iterative solution on the multi-target static resource target allocation problem based on the bipartite graph matching double-layer model, and stopping iteration until a preset termination condition is met. According to the method, the joint damage efficiency of the plurality of resources to a certain target can be calculated, and the damage efficiency of the plurality of resources to a certain target is not simply accumulated, so that the solution accuracy is improved.

This invention relates to the fields of data processing and talent matching technology, and in particular to a two-way matching method for university-enterprise talent based on financial capability profiling. The method includes: constructing a corporate financial capability profiling indicator system, including indicators of corporate financial stability and job financial capacity; constructing a talent financial capability profiling indicator system, including indicators of financial knowledge reserves, financial practice skills, financial risk awareness, and financial digitalization capabilities; collecting and preprocessing data from both the corporate and talent sides; calculating a comprehensive evaluation value based on a combined weighting method using an improved analytic hierarchy process (AHP) and entropy weighting method; constructing a corporate job demand vector and a talent supply vector; solving for the optimal matching scheme based on a bipartite graph matching algorithm, with the objective function of maximizing two-way matching satisfaction; and outputting the matching results. By constructing a two-dimensional evaluation system of corporate and talent financial capabilities, and using an improved weighting method and matching algorithm, a precise two-way matching of university and enterprise talent is achieved, improving the scientific nature of the matching.

The application provides a kind of based on adaptive time delay-phase structure's super large scale MIMO system beam focusing method, it is related to radio communication technical field, including the following steps: constructing adaptive time delay-phase structure suitable for super large scale MIMO system, and establishing near-field beam focusing joint optimization mathematical model;Through maximum ratio transmission solving all digital optimal beam focusing matrix;Switch matrix optimization problem is converted into two-part graph matching problem, and optimal matching matrix is solved using hungarian algorithm;Through riemann spectrum conjugate gradient method solving phase shift matrix;Utilize gradient descent method to optimize time delay matrix;Least square method is used to optimize digital precoding matrix;Through alternate iteration, base station beam focusing matrix is given, and the beam focusing of super large scale MIMO system is completed.The structure provided in the application can be realized with switch network and the dynamic sharing of antenna array element, so as to effectively enhance the adaptive ability of hybrid precoding to channel environment.

The application belongs to but is not limited to the technical field of automatic student behavior analysis, and discloses a classroom student behavior recognition method and system based on a Transformer detection and hierarchical soft routing decision. Frame images are intercepted from classroom monitoring videos and input into a pre-trained RT-DETR model; multi-scale features of the images are extracted by a hybrid encoder of the RT-DETR and input into a Transformer decoder, global context information of the images is captured by using a self-attention mechanism, and a set of boundary boxes of N student targets is predicted by using a bipartite graph matching mechanism; each boundary box detected is dynamically expanded and cropped; a pre-trained visual-language large model CLIP is used to construct a hierarchical decision system with a logical tree structure to perform hierarchical feature extraction; and an output result is obtained by reasoning based on a low-threshold parallel activation soft routing decision mechanism; and the application significantly improves the recognition accuracy of fine-grained actions.

The invention relates to the technical field of vehicle task scheduling, in particular to a dynamic vehicle task scheduling method and system based on two-stage differentiation and multiple targets. The method comprises the following steps: acquiring vehicle data and task data; performing time slice static modeling based on the task data; carrying out vehicle classification and attribute modeling based on the vehicle data; according to a modeling result, performing two-stage differentiation and multi-target dynamic vehicle task allocation, including a first stage of sparse bipartite graph matching allocation and a second stage of CMOEA-EG multi-target optimization; performing fairness evaluation and feedback based on a distribution result; and updating and optimizing based on a feedback result. The task execution result is fed back in real time through the dynamic updating module, and multi-round iterative optimization is supported.

The application discloses a speaker recognition method based on bipartite graph matching and an electronic device, and belongs to the technical field of audio recognition. The method comprises the following steps: acquiring an audio stream, and splitting the audio stream into continuous target time length audio segments in real time; determining a voiceprint embedding vector corresponding to each target time length audio segment; determining an embedding matrix based on the voiceprint embedding vector; determining a target voiceprint feature corresponding to each target time length audio segment according to a bipartite graph method; calculating the similarity between each voiceprint embedding vector; grouping each voiceprint embedding vector based on the similarity between each voiceprint embedding vector, clustering the voiceprint embedding vectors corresponding to the same speaker, obtaining a clustering result, and generating a target recognition report. The application can reduce the calculation complexity while improving the recognition accuracy.