94 results about "Density based" patented technology

The embodiment of the invention provides a data processing method and system for an enterprise digital transformation platform, and belongs to the field of data processing. The method comprises the steps that structured field information from all heterogeneous data sources is acquired, and preprocessing operation is executed on the structured field information; constructing the processed structured field information into an embedded input sequence, splicing the embedded input sequence into a natural language fragment according to a preset template, and inputting the natural language fragment into a fine-tuned semantic coding model to obtain a corresponding semantic embedded vector; identifying similar field groups by adopting a clustering algorithm based on density or a hierarchical structure, and classifying each group of structured field information into a semantic cluster; and generating a corresponding standard field identifier for each semantic clustering cluster, and storing the generated standard field identifier in a standard field index database of the platform after digital transformation. According to the scheme, the field unified management and cross-system data alignment capability of the enterprise digital platform is remarkably enhanced.

The invention relates to a terrain real scene modeling method, device and equipment and a storage medium thereof. According to the method, multi-source heterogeneous data related to the terrain in a target area is integrated, a multi-modal data cube is generated through noise filtering and data fusion under a unified coordinate system, then a data cavity area is identified based on density clustering and elevation gradient analysis, and geophysical features are extracted. Innovatively constructing a neural implicit inference model fused with physical constraints to perform joint modeling on the earth surface and underground structures, and finally converting implicit terrain representation into an explicit three-dimensional terrain model; seamless fusion modeling of the earth surface and the underground space under the complex shielding environment is achieved, the physical rationality of the model and the hidden area reconstruction precision are remarkably improved, meanwhile, it is ensured that the generated model strictly conforms to the geological physical law, and the geological engineering safety and the planning reliability are effectively guaranteed.

The invention discloses a target detection method and system based on a millimeter wave radar. The target detection method and system are used for realizing accurate detection, positioning and dynamic and static recognition of multiple targets under a complex background. According to the method, a distance-Doppler spectrogram is generated through the technical means of sliding window construction, spectral analysis, clutter suppression and the like, and candidate target points are detected by adopting an SO-CFAR algorithm. Then, determining a target position through high-resolution direction estimation and coordinate transformation, performing spatial clustering in combination with a density-based DBSCAN algorithm, and extracting a target geometric center and a bounding box; in the aspect of target tracking, Kalman filtering is used for predicting and updating the position and speed of the target, and a beam forming technology is used for enhancing a target signal, so that the target recognition stability is improved. And finally, the system performs robust dynamic and static state recognition on the target through a dynamic and static judgment module, so that high precision and robustness of the target detection process are ensured. The method can effectively cope with static background interference and dynamic target changes, and is suitable for target detection and tracking in a complex environment.

The invention discloses a wind and light output scene generation method based on depth feature mining and adaptive clustering, and the method comprises the steps: cleaning wind and light output data, carrying out the normalization processing of the data, and enabling the data to be mapped to a preset interval, so as to eliminate the dimension influence; constructing a deep convolutional feature extraction network to extract a corresponding feature map from the normalized wind and light output data; a K-Means + + algorithm is adopted to initialize a clustering center, an improved ISODATA clustering algorithm is executed based on a density threshold dynamic splitting mechanism, clustering parameters are optimized through Bayesian optimization, and a typical scene is generated. The accuracy of the wind and light output scene is remarkably improved.

The invention discloses a BIM-based construction engineering visual management method and system, and relates to construction engineering management, and the method comprises the steps: collecting a data set containing geometric information, attribute information and associated information; dividing the data set into a plurality of hierarchical subsets according to the hierarchical relationship of the building engineering BIM model; an improved DBSCAN clustering algorithm is adopted to perform clustering analysis on each hierarchy subset; classifying and archiving the data set according to the data cluster, and generating a semantic association data collaboration array A according to a classifying and archiving result; carrying out dimension reduction processing on the semantically associated data collaboration array A; constructing a spatial index of the dimension reduction data array by adopting a density-based spatial index method; and dividing the building engineering BIM model into a plurality of partitions by using a distributed computing framework, and performing distributed matching on the partitions and the spatial index to obtain a data matching point set. For the problem that the BIM-based building engineering high-dimensional data clustering precision is low in the prior art, the analysis precision is improved.

The invention belongs to the technical field of electric digital data processing, and more specifically relates to an agricultural machinery behavior identification method fusing interpolation enhancement and depth spatial-temporal feature modeling. The method comprises the following steps: acquiring GNSS trajectory data of an agricultural machine and preprocessing the trajectory data; performing data enhancement on the road points in the data set by adopting a local interpolation method based on DBSCAN (Density Based Spatial Clustering of Applications with Noise) clustering; designing a set of spatial distribution feature extraction method; performing feature extraction on the trajectory data after data enhancement by combining motion feature extraction and spatial distribution feature extraction; introducing a variational auto-encoder VAE to carry out potential feature extraction on the initial features; then, a ResBiLSTM network is adopted to carry out space-time modeling; and finally, an end-to-end classification decision is realized through a linear classifier, so that intelligent identification of the operation behavior of the agricultural machine is realized. According to the method, two key problems in the existing track identification technology are solved, namely, data distribution is seriously unbalanced, and the spatial distribution characteristics of track points are often neglected.

The invention provides a processing method and system for scoliosis detection based on point cloud features. According to the system, a depth camera is used for monitoring the dynamic state of a subject through a point cloud data acquisition module, and back three-dimensional point cloud information is obtained; the scoliosis area acquisition module acquires a clustered two-dimensional diagram by using a density-based spatial clustering algorithm and a point cloud reprojection algorithm, and acquires an area of interest by using an image morphological processing algorithm and a polynomial curve fitting method; the point cloud registration module obtains symmetrical point clouds by using an iterative reweighted least square method and a principal component analysis algorithm, performs coarse registration by using a Rodrigues rotation matrix and then executes a point cloud registration algorithm; and the transverse asymmetry analysis module obtains a symmetry plane by using the registered point cloud, and obtains a transverse depth difference point cloud by using a surface reconstruction algorithm. The invention provides a transverse asymmetry analysis method based on a point cloud registration algorithm and a surface reconstruction algorithm, scoliosis screening is performed through the depth camera, and popularization of scoliosis screening is facilitated.

The invention provides a trajectory clustering method based on space-time weighting and a density peak value, and relates to the technical field of trajectory clustering. The method provided by the invention comprises the following steps: based on a minimum description length criterion, fusing an adaptive space-time weight and a space-time geometric distance to segment trajectory data to obtain sub-trajectory segments; calculating a space-time local density, a relative distance and a decision value of each sub-track segment based on a density peak clustering algorithm; performing noise identification on the sub-track segments based on the average space-time geometric distance and the average local space-time density, iteratively selecting a class cluster center based on the decision values of the non-noise sub-track segments and the time-space factor decision values, and generating a class cluster set; and determining a candidate representative trajectory set from the class cluster set based on the spatio-temporal local density and the maximum density, extracting continuous sub-trajectory segments from the candidate representative trajectory set based on spatio-temporal continuity constraints, and sequentially connecting end points of the continuous sub-trajectory segments to generate a representative trajectory. According to the method, the trajectory clustering effect is improved through space-time weighted segmentation and density peak trajectory clustering.

The invention provides an ultrasonic three-dimensional imaging method and system based on an FPGA, and relates to the technical field of ultrasonic imaging, and the method comprises the steps: transmitting a chirp signal, receiving an echo signal reflected by a target object, and carrying out the preprocessing; carrying out adaptive filtering, constant false alarm rate detection, adaptive beam forming and three-dimensional positioning algorithm processing on the preprocessed signal to generate target position data; and performing three-dimensional point cloud image reconstruction on the target position data by using a density-based spatial clustering algorithm and a deep learning model, and outputting a target three-dimensional image. Through the innovation of a signal processing algorithm and the collaborative design of the system, the technical bottlenecks of a traditional ultrasonic sensor in detection distance, anti-interference, real-time performance and positioning precision are solved, and the three-dimensional sensing capability with high reliability and strong environmental adaptability is provided for intelligent driving.

The invention relates to the technical field of protein prediction, in particular to a protein binding site prediction method based on geometric deep learning. According to the method, protein geometric description and a corresponding geometric diagram learning scheme are introduced, and a feasible way is provided for modeling three-dimensional structural features; in addition, the invention also provides an RSA-guided two-stage hybrid transfer learning strategy, firstly, a model is trained on a larger data set to learn general representation, and then fine tuning is carried out for binding site prediction, so that the overfitting problem in limited data training is relieved; according to the invention, a prediction post-processing module and an integrated learning module based on density clustering are designed, and the high variance of the model in multiple times of training operation is significantly reduced.

The embodiment of the invention provides a point cloud denoising method and device, a medium and a product, and relates to the technical field of artificial intelligence. The method comprises the following steps: acquiring target point cloud data subjected to abnormal point elimination; according to the target point cloud data, constructing a density-based hierarchical clustering algorithm and a de-noising network of an operation selection strategy, and generating a de-noising model based on the de-noising network; and inputting the to-be-denoised point cloud data into the denoising model, and outputting the denoised point cloud data. According to the scheme, isolated noise possibly misleading path decision is filtered in advance, and interference is cleared for follow-up path selection; a denoising network based on a density hierarchical clustering algorithm and an operation selection strategy is constructed, a point cloud structure is accurately divided, representative elite points are screened in combination with the operation selection strategy, under extreme conditions, the elite points are preferentially used as core extension paths, noise point dominant decision making is avoided, and the accuracy of the system is improved. The problem that in the prior art, a single path is poor in adaptability in an extreme scene is solved, and the denoising accuracy is improved.

The invention discloses a shallow sea area water depth inversion method based on a satellite-borne laser radar, and particularly relates to the technical field of single-photon water depth information extraction. The denoising method comprises the following steps: S1, data preprocessing: reading data, carrying out visual presentation, dividing photons into offshore photons, sea surface photons and undersea photons, and intercepting an effective photon region; s2, coarse denoising: denoising the area intercepted in the step S1 by adopting a DBSCAN (Density Based Spatial Clustering of Applications with Noise) algorithm; s3, fine denoising: screening effective signals of sea surface photons by adopting a Gaussian fitting algorithm, removing sea photons, separating subsea photons from the sea surface photons, performing density analysis on the subsea photons, extracting a density peak point, and constructing a density trend curve of the subsea photons by adopting a cubic spline interpolation method, eliminating residual noise deviating from normal distribution according to the trend curve; and S4, performing water depth inversion: performing refraction correction on the subsea photons subjected to fine denoising, fusing the corrected subsea photons and the sea surface photons subjected to fine denoising, and generating a water depth distribution diagram of the shallow sea area.

The invention discloses an obstacle sensing method and system based on laser radar data analysis, and relates to the technical field of radar data obstacle analysis, and the method comprises the steps: scanning a surrounding environment through a multi-mode sensor, obtaining sensor data, synchronously collecting the data of the multi-mode sensor, and carrying out the preprocessing, an obstacle target is obtained from a camera image, sensor data are integrated by using a data fusion algorithm, and obstacle features are extracted by using a segmentation threshold function based on density clustering in combination with an obstacle local feature descriptor and global feature description to form complete obstacle feature description. The laser radar can provide high-resolution distance data, so that the system can accurately identify and position obstacles in the surrounding environment, algorithm processing is combined, the method can complete scanning and analysis of the environment in a short time, rapid response to the obstacles in the dynamic environment is achieved, and the system can be widely applied to the field of dynamic environment monitoring. And the accuracy and reliability of obstacle detection are greatly improved.

According to the automatic production line digital model hole filling method and device provided by the embodiment of the invention, the to-be-processed area is accurately identified through multi-level grid analysis and curvature field calculation. A patch generation mechanism based on density clustering and a neural network is innovatively designed, and intelligent filling optimization is realized in combination with local geometric features. According to the system, non-uniform rational basis spline surface reconstruction and a Laplacian grid smoothing algorithm are adopted, continuity and fairness of a filling area are ensured, and the system is closely combined with motion control and path planning of an automatic production line. According to the method, the limitation of traditional model repair is broken through, and an efficient and reliable solution is provided for integrity reconstruction of the industrial digital model.

Disclosed in the present invention is a three-dimensional circle detection method for a freeform surface, comprising: acquiring freeform surface point cloud data, and using a normal vector-guided rolling ball method to extract a freeform surface circular-hole point cloud boundary contour from the point cloud data; performing Euclidean clustering-based boundary segmentation on the point cloud boundary contour to obtain a clustering result of all boundary point clouds, the clustering result consisting of a plurality of circular-hole contours; performing a density-based weighted iterative algorithm on each circular-hole contour to obtain a freeform surface circular-hole normal vector; and projecting a freeform surface circular hole along the freeform surface circular-hole normal vector onto a plane, and then performing circular-hole iterative fitting based on solving an overdetermined equation on the projected circular-hole point cloud to obtain final positioning information. The present invention is applicable to circular-hole positioning on freeform surfaces of different curvatures, has strong robustness, and effectively improves the accuracy of circular-hole positioning on the freeform surfaces.

The invention discloses a method and device for constructing a three-dimensional fault plane according to an earthquake catalog, and the method comprises the steps: extracting the position information of an earthquake source according to the earthquake catalog of a target region, obtaining the density information of each data point, and generating a density map; based on the density information and the density map, utilizing an OPTICS algorithm to carry out preliminary clustering analysis on the seismic source position information, outputting an accessibility map, and determining a clustering number N; calculating the local density and the center offset distance of each seismic source point based on the Euclidean distance and the truncation distance between every two seismic source points, generating a decision diagram, and selecting the first N seismic source points with the maximum product of the local density and the center offset distance as a clustering center; and PCA analysis and PCA plane determination are carried out on the DPC classification result, and plane features of the fault plane are obtained when set conditions are satisfied. According to the method, an OPTICS algorithm, a DPC algorithm and PCA analysis are combined, and end-to-end parameterization-free three-dimensional fault plane modeling is achieved only by means of seismic source position information.

The invention discloses an open hole vertical well fracturing crack position identification method based on density clustering, which belongs to the technical field of hydraulic fracturing, and comprises the following steps: collecting a water hammer signal at the moment of pump stop by using a high-frequency pressure monitor; performing filtering processing on the signal by using an FIR filter; carrying out deconvolution processing on the filtered signal by adopting a cepstrum analysis method, and extracting crack reflection characteristics; determining the position of a liquid inlet point by combining an impedance identification technology; and performing dynamic analysis by using a density clustering method according to the determined liquid inlet point position, and identifying the crack number and the main body position based on the liquid inlet point space density. By the adoption of the method, the spatial distribution of the liquid inlet points is obtained through water hammer signal processing, the spatial density of the liquid inlet points is analyzed based on the density clustering analysis method, and therefore crack position recognition in the complex liquid inlet process such as vertical well open hole fracturing is achieved.

The invention discloses an industrial key point data cleaning and abnormal point removing method and a related device, and the method comprises the steps: obtaining to-be-processed industrial key point coordinate data, and carrying out the preprocessing of the coordinate data; dynamically determining the optimal clustering number of the preprocessed coordinate data; based on the optimal clustering number, clustering the coordinate data by adopting a first clustering algorithm, and identifying a first type of abnormal points according to distance distribution statistical characteristics in each cluster; performing secondary cleaning on the data from which the first type of abnormal points are removed by adopting a density-based second clustering algorithm so as to identify a second type of abnormal points which are judged as noise points; and integrating the first type of abnormal points and the second type of abnormal points to generate a cleaned data set. According to the method, dynamic clustering, anomaly detection based on intra-cluster statistics and density secondary cleaning are combined, so that the adaptive capacity, accuracy and robustness of the data cleaning process are enhanced, and a purer and more reliable data set can be produced.

The invention provides a three-dimensional steel structure full-coverage path planning inspection method based on topology reconstruction, and relates to the technical field of industrial robots, and the method comprises the steps: obtaining a model file of a to-be-inspected three-dimensional steel structure; constructing an original vertex set and an original edge set according to the model file; performing adaptive clustering on each original vertex in the original vertex set through a density-based spatial clustering algorithm to obtain a cluster; taking the geometric center of the original vertex in each cluster as a topological node, and establishing a mapping relation between the original vertex and the topological node; carrying out topology reconstruction on the original edge set, and constructing an undirected weighted graph; stitching non-connected sub-graphs in the undirected weighted graph through a K-dimensional tree and a minimum spanning tree algorithm to obtain a fully connected graph; generating a full-coverage continuous path sequence through a Chinese postman algorithm; the full-coverage continuous path sequence is converted into a three-dimensional space coordinate sequence, and the three-dimensional steel structure to be inspected is inspected according to the three-dimensional space coordinate sequence.

The invention provides a logistics site selection method and system based on combination of density peak clustering and a Gaussian model, and belongs to the field of logistics planning and data analysis. The method comprises the steps of collecting multi-dimensional data in logistics site selection, and performing preprocessing; carrying out clustering analysis on the preprocessed multi-dimensional data by adopting an information entropy-based density peak value clustering algorithm and a Gaussian mixture model combined clustering algorithm, and extracting a clustering center of the data as a potential logistics site selection point; and according to a preset logistics site selection target, constructing a target function and setting a constraint condition, and solving an optimal logistics site selection point from the potential logistics site selection points through an optimization algorithm. Therefore, the internal structure and rule behind the data are disclosed. The limitation that a traditional method depends on a single factor is avoided, the objectivity and reliability of the site selection scheme are improved, the site selection scheme can be rapidly adjusted through data updating and re-clustering, and the optimality is kept.

The invention discloses an AGGAT-based electromagnetic spectrum sensing open set target recognition method, which comprises the following steps: firstly, constructing a knowledge graph representing formation information based on target multi-dimensional features extracted from an electromagnetic spectrum, and converting the knowledge graph into a representation form under a graph model through data coding and normalization; a global attention mechanism is introduced to construct an adaptive global graph attention network, comprehensive utilization of local information and global information in the graph is realized, feature extraction of targets in the formation graph is realized, and AGGAT is applied to feature extraction of the targets; designing a target type recognition loss function comprehensively utilizing supervised learning and comparative learning, and applying the target type recognition loss function to a training process of target type feature updating; and finally, constructing a feature clustering model of a density-based anti-noise clustering method, and constructing a feature comparison model based on Euclidean distance for outputting the type of the open set target. According to the method, the information source can be adaptively selected under different data distributions, so that the flexibility and characterization capability of image attention model feature extraction are improved.

The invention discloses a steel bar bundling distance measuring method and device based on layered recognition and a medium, and relates to the technical field of building construction. The method comprises the following steps: carrying out multi-angle three-dimensional scanning on a construction area by carrying multiple sensors through an unmanned aerial vehicle, and generating a point cloud data set containing a steel bar layer; performing hierarchical identification on the point cloud data set based on a density clustering algorithm to obtain spatial distribution of reinforcing steel bar layers; performing model classification on the reinforcing steel bar layer by using a three-dimensional convolutional neural network model, and extracting design parameters to perform multi-dimensional model comparison to obtain a reinforcing steel bar model and reinforcing steel bar bundling data; and a multi-source data set of the reinforcing steel bars is introduced and aligned, a rule database of the reinforcing steel bars is obtained, and a multi-modal acceptance report of the reinforcing steel bar models and the reinforcing steel bar bundling data is generated according to the rule database. According to the method, the problems of inaccurate acquisition of steel bar bundling data and single comparison dimension in a complex construction scene are solved, and accurate measurement of steel bar spacing in a high-density area and intelligent management of construction quality are realized.

The invention discloses an electric power spot market price abnormity early warning method and device, equipment and a medium, and the method comprises the steps: firstly obtaining historical price data and first multiple space-time sensitive factor data, and generating second multiple space-time sensitive factor data of a future time point through a pre-trained sensitive factor prediction model; inputting the second data into a market price situation awareness model, and predicting price data at each time point in the future; and finally, carrying out clustering analysis on the historical and predicted price data by adopting a density-based spatial clustering algorithm, identifying abnormal price data and outputting early warning information. According to the method and the system, dual modeling of multi-time-space sensitive factor prediction and market price situation awareness is fused, and the clustering algorithm is combined to intelligently identify the abnormal price, so that the adaptability of the early warning system to the complex environment change of the power market is effectively improved, and the accuracy is better. The method can be widely applied to the technical field of new energy.

The invention provides a micro-motion H / V spectrum ratio data processing method and device based on machine learning and a medium, and the method can comprise the steps: obtaining micro-motion H / V spectrum ratio data containing a plurality of curves, recognizing and removing the curves containing abnormal morphological characteristics through a preset curve removing model, and obtaining a candidate data set; classifying peak values of the curves in the candidate data set according to a preset peak value classification model to obtain an effective peak value set; clustering the effective peak value set based on a density clustering algorithm to obtain effective peak value clusters in different frequency intervals; and determining the effective peak cluster with the lowest frequency as a main peak cluster, and rejecting curves of peak values with abnormal frequencies and amplitudes in the main peak cluster to obtain processed micro-motion H / V spectrum ratio data. The method has higher stability and accuracy, the spectrum distribution of the processed curve is more concentrated, the standard deviation curve is more convergent, and the peak value is clearer and more stable.

A transformer partial discharge positioning detection method based on ultrahigh frequency signals is characterized by comprising the following steps: S1, estimating time delay of signals obtained by an ultrahigh frequency (UHF) sensor array based on a generalized cross-correlation method; and S2, arranging and combining signals of the UHF sensor array, and solving coordinates of a positioning point based on an improved multi-moth optimization (MFO) algorithm according to the time delay obtained in the step S1. And S3, clustering the coordinates of the plurality of positioning points obtained in the step S2 based on a density clustering algorithm (DBSCAN), and selecting the geometric center position of the category with the maximum sample number as the final local discharge source coordinate.

Methods for detecting power system events in an electrical power system. An example method comprises determining a signal envelope for each of at least first and second power system signals and performing density-based spatial clustering of a series of points formed by combining respective values of the signal envelopes for at least the first and second power system signals. The example method further comprises detecting one or more power system events by identifying outlier points or groups of points in the spatial clustering. Some methods may further comprise automatically classifying power system events by collecting a data set comprising, for each detected power system event, power system signal features corresponding to the detected power system event, and classifying each of the detected power system events using the data set and a machine-learning-based classification algorithm, where said classifying comprises determining a classification label from among two or more predetermined classification labels.

According to an aspect, there is provided a computer-implemented method comprising the following. Initially, information on a plurality of anomaly events relating to operation of a target system is obtained. The information comprises one or more time series of anomaly score data. Segments satisfying one or more pre-defined criteria for anomalous operation are detected from the one or more time series. The one or more pre-defined criteria are defined to exclude fully non-anomalous anomaly score data. Anomaly durations and standardized anomaly scores are determined for the segments. Partition or density based clustering is performed in a two-dimensional space formed by the standardized anomaly scores and the anomaly durations to form n clusters, and m smallest clusters of the n clusters are identified. At least one of the following is performed: outputting information on the m smallest clusters or causing adjusting of operation of the target system based on the m smallest clusters.

Techniques are described for providing a ML data analytics application including guided ML workflows that facilitate the end-to-end training and use of various types of ML models, where such guided workflows may also be referred to as ML “experiments.” For example, the ML data analytics application may enable users to create experiments related to prediction of numeric fields (for example, using linear regression techniques), predicting categorical fields (for example, using logistic regression), detecting numerical outliers (for example, using various distribution statistics), detecting categorical outliers (for example, using probabilistic statistics), forecasting time series data, and clustering numeric events (for example, using k-means, density-based spatial clustering of applications with noise (DBSCAN), spectral clustering, or other techniques), among other possible uses of various types of ML models to analyze data.

The invention belongs to the technical field of disaster recognition, and particularly relates to a density-based parameter adaptive movable landslide space clustering recognition method. The method comprises the steps of firstly obtaining deformation field data of a to-be-detected area through an InSAR technology, then obtaining gradient matrixes of an input image in eight directions through an edge detection algorithm on the basis of an InSAR deformation rate result, removing isolated pixels in the gradient matrixes through Gaussian filtering, calculating a mean value and a standard deviation of the gradient matrixes, and finally obtaining the deformation field data of the to-be-detected area according to the mean value and the standard deviation. Taking the maximum value of the two as a mask threshold value of the input image to obtain a refined deformation rate field; secondly, a fine deformation rate field is used for positioning motion pixels which are possibly movable landslides; after all moving pixels possibly being movable landslides are positioned, a snow ablation optimizer is introduced to optimize core parameters of the DBSCAN, a parameter-adaptive SAO-DBSCAN spatial clustering method is formed to automatically cluster the moving pixels into landslide masses, and the landslide automatic clustering performance and precision are greatly improved.

The present invention relates to a method, apparatus, computer device, and storage medium for dividing stop points. The method comprises: determining the track points of a truck at a stop point based on its running track points; clustering the track points based on a density clustering algorithm to obtain multiple clusters; establishing an adjacency matrix between clusters based on proximity search; segmenting the adjacency matrix according to the method, and selecting a segmentation method that satisfies a preset condition for the multiple new clusters formed after segmentation as a stop point segmentation rule, wherein the preset condition is that the multiple new clusters formed after segmentation have the highest intra-class similarity and the lowest inter-class similarity. The above method can effectively restore the true regional boundaries.