53 results about "Silhouette coefficient" patented technology

The invention belongs to the technical field of stratum recognition, and particularly relates to a shield parameter geological recognition calculation method and system based on an unsupervised clustering algorithm, and the method comprises the steps: selecting a plurality of groups of data containing factors affecting stratum category recognition; pCA converts a group of complex related variables into fewer and irrelevant variables, and the new variables are called as main components and can capture most of information in original data; determining the clustering number of K-means + + by adopting an elbow method; determining the clustering number of a hierarchical clustering method by adopting a hierarchical tree; k-means + + and hierarchical clustering are respectively used for clustering the data, and evaluation indexes such as identification precision or contour coefficients of the data and the data are calculated, so that potential modes in the data are analyzed and mined more comprehensively. The invention aims to deeply explore the internal relevance between shield tunnel parameters and geological types, provides an innovative and robust solution for real-time detection of geological features in shield construction by overcoming the limitation of the prior art, and has wide application potential.

The invention discloses an intrusion detection method and system based on multi-scale feature extraction and variation clustering. The method comprises the following steps: step 1, collecting network flow data; step 2, extracting multi-scale adaptive gating features; step 3, preliminarily judging feature abnormity; 4, updating clustering parameters based on the contour coefficient, the CH index and the k distance map; 5, clustering the multi-scale adaptive gating features corresponding to the network traffic, and analyzing the category of the network traffic according to a clustering result; and step 6, performing final classification decision on the suspicious network traffic by adopting a preset rule. According to the scheme, traffic features under different time scales are captured through a multi-head attention mechanism, effective information is screened by using a gating mechanism, and the feature expression ability is improved; clustering is driven by multi-dimensional indexes, so that the clustering process is more adaptive, and the recognition precision of abnormal traffic is improved; not only is efficient classification of known attacks realized, but also the recognition capability of unknown attacks is remarkably improved.

The invention relates to the technical field of machine learning, and provides a sample generation method and device, computer equipment and a readable storage medium, and the method comprises the steps: obtaining a to-be-generated original data set, and separating the original data set into a minority class sample set and a majority class sample set; for the minority class samples, by calculating a contour coefficient under each clustering number, determining an optimal clustering number enabling the contour coefficient to be maximum, and clustering the minority class samples based on the optimal clustering number to obtain a plurality of minority class sample clusters so as to generate a plurality of candidate samples; and respectively acquiring a first average nearest neighbor distance from each candidate sample to the minority class sample and a second average nearest neighbor distance from each candidate sample to the majority class sample, so as to determine the target sample as a newly generated minority class sample. According to the method, samples conforming to a real fraud mode can be accurately generated in financial fraud detection, rare disease feature distribution can be effectively simulated in medical diagnosis, and the early screening accuracy is improved.

The invention provides a direct path distinguishing method based on adaptive parameter DBSCAN clustering. Firstly, a distance distribution matrix is calculated, and an average nearest neighbor distance matrix is obtained through sorting and averaging operation; on the basis, a Savitzky-Golay filter is used for filtering a normalized distribution probability histogram obtained according to the distance matrix. Secondly, searching by adopting a peak searching algorithm to obtain an interval of an optimal neighborhood radius, and calculating a minimum point number corresponding to the interval of the optimal neighborhood radius by utilizing a mathematical expectation method; thirdly, DBSCAN clustering is carried out according to the neighborhood radius and the minimum point number, a contour coefficient is calculated, and a clustering result corresponding to the maximum value of the contour coefficient is selected; and finally, a maximum likelihood function is calculated according to the point number, the average value, the variance and the ToF features of the clusters, and then the direct path is distinguished. According to the method, under the condition that the number of clusters, the neighborhood radius and the minimum point number in the neighborhood radius do not need to be determined, clustering is carried out according to the characteristics of data, and the direct path distinguishing capacity of a system is improved.

The invention provides an evaluation index data processing and evaluation method, computer equipment, a medium and a product. In a specific embodiment, the evaluation index data processing method comprises the following steps: obtaining evaluation index data of a plurality of to-be-evaluated targets; the number of clustering centers of a K-means clustering algorithm is determined according to the sum of squares of errors corresponding to different numbers of clustering centers, the average contour coefficient corresponding to the number of the partial clustering centers and the number of preset evaluation grade values, and the number of the clustering centers of the K-means clustering algorithm is smaller than or equal to the number of the preset evaluation grade values; according to the number of clustering centers of the K-means clustering algorithm, grouping normalization based on the K-means clustering algorithm is carried out on the evaluation index data of the multiple to-be-evaluated targets, and evaluation grade values corresponding to the evaluation index data of the multiple to-be-evaluated targets are obtained. According to the embodiment, the objectivity of normalization of the evaluation index data and the rationality in physical significance can be improved, so that the reliability of an evaluation result is improved.

The invention discloses a slope anchor cable prestress measuring method based on data clustering. The slope anchor cable prestress measuring method comprises the steps that a multi-dimensional database containing environment characteristics, echo energy and counter-pulling method prestress values is constructed; a mixed distance clustering algorithm is designed, numerical features and classification features are jointly processed, and automatic data classification is achieved by iteratively updating a cluster center; the cluster quality is dynamically evaluated based on the contour coefficient, and only when the sample size is gt; 3 and an average contour coefficient gt; constructing an exclusive regression model in the cluster when the contour coefficient is a threshold value; and finally predicting the anchor cable prestress in real time through echo energy. According to the method, the prediction precision under the complex working condition is remarkably improved, the use frequency of the reverse drawing method is reduced, and safe and efficient prestress nondestructive testing is achieved.

The invention provides a multi-modal process fault detection method based on SKDPC-RS-MSDAE, which is used for solving the problems that multi-modal division depends on experience knowledge, the detection precision of tiny faults is low and the like. The method comprises the following steps: firstly, collecting original data of a TE multi-mode process, and mining deep information of the original data by using ResNet; secondly, a DPC modal identification method based on path selection is designed, local density and relative distance are reconstructed through path selection and kNN, an optimal clustering number judgment criterion based on a contour coefficient is constructed, and a DPC based on the contour coefficient and path selection is designed; secondly, establishing an MSDAE detection model for each steady state, and obtaining a corresponding fault probability; and finally, designing a Bayesian reasoning probability index BIP based on multi-probability fusion. Experimental verification is carried out through the TE process, compared with other methods, the overall fault detection rate is improved, and the detection performance of tiny faults is improved.

The invention relates to a river water quality inversion method and system, and specifically, the method comprises the steps: obtaining the hyperspectral data of a river region, carrying out the radiation correction of the hyperspectral data, constructing a plurality of indexes according to the wave band in the hyperspectrum, calculating the correlation between each index and a water quality parameter, and sorting the indexes according to the correlation. Determining an index set corresponding to the water quality parameters based on an index sorting result; clustering the water quality parameters based on the correlation of the water quality parameters and the correlation of the index set of the water quality parameters; in the training process, for each cluster, the importance of indexes is determined according to the correlation between the indexes in the index set of the cluster and water quality parameters in the cluster, the splitting index selection probability during splitting of each tree is determined based on the importance of the indexes, and splitting stop conditions are determined according to the contour coefficient of the cluster; and performing inversion on the new hyperspectral data by using the trained model to obtain water quality parameters.

The invention discloses a label correction crowdsourcing result aggregation method and system based on deep clustering. Firstly, task features are extracted through a pre-training model to serve as input data of the model; secondly, introducing a worker loss item, defining a comprehensive loss function of the model in combination with reconstruction loss and KL divergence in the variational depth embedding model, and training the model through a gradient descent method; secondly, in the training process, clustering the hidden variables by using a Gaussian mixture model to obtain clustering clusters, and mapping a specific class label for each clustering cluster in combination with task labeling information of a worker; and finally, the contour coefficient of each task is calculated, the task with low clustering accuracy is identified, and the precision of the clustering result is further improved through readjustment and correction of worker labels. According to the method, tasks with similar features are clustered, and tasks with low clustering quality are corrected by using worker tags, so that the interference of tag noise on truth value inference is effectively reduced, and the performance of a result convergence model is improved.

The present application relates to the field of artificial intelligence, and discloses a model updating method, including: performing multiple first clustering processes on multiple training images according to multiple target features to obtain multiple first clustering results, each of the multiple first clustering results corresponds to a silhouette coefficient, and the silhouette coefficient represents the clustering quality; determining the first target clustering result according to the silhouette coefficient, and the first target clustering result includes M clustering categories; performing second clustering processes on multiple training images according to multiple target features to obtain second clustering results, and the number of clustering categories included in the second clustering result is M. The present application does not need to verify the model corresponding to the obtained number of different clustering categories based on labeled samples to determine the optimal number of clusters, and can directly select the optimal number of clustering categories through the results of the first clustering.

The invention belongs to the related technical field of data processing, and discloses a hierarchical clustering method integrating distance features, shape features and time sequence features, which comprises the following steps of: in an initial stage, regarding each time sequence process as an independent cluster, and calculating distances of different features among the time sequence processes; according to the contribution degree of each distance, constructing a composite similarity index DST integrating distance, shape and time sequence characteristics; calculating the distance between different clusters according to the constructed composite similarity index and the determined inter-cluster connection mode, and continuously iterating and merging the nearest clusters; when all clusters are combined into a large cluster, clustering is completed, and the optimal clustering number is determined according to the contour coefficient after each time of combination. According to the method, the composite similarity index DST is constructed, the weight is automatically calculated according to the contribution degree of different features to the current time sequence process, the multi-dimensional difference between the time sequence processes can be more comprehensively reflected, and the clustering result is more in line with the actual dynamic evolution rule of the time sequence process.

The invention discloses an industrial park weather typing method and device integrating a self-organizing mapping neural network and a k-means algorithm, and the method comprises the steps: firstly, constructing an initial data set through air quality monitoring data and ERA5 meteorological reanalysis data, converting the data into a numerical value type, and carrying out the standardization, so as to guarantee that the feature data has the distribution with the mean value being 0 and the standard deviation being 1; thirdly, initializing a self-organizing mapping (SOM) model, and performing dimension reduction on the data by setting parameters such as grid size, sigma and learning rate so as to obtain two-dimensional feature representation; then, initializing a k-means clustering model, setting a clustering number and a random seed, and performing clustering analysis on the feature data after dimension reduction by using a fitpredict () method so as to obtain a clustering label of each sample; in the evaluation and visualization stage, evaluation indexes such as SSE, MSE, RMSE, a contour coefficient and a Clinski-Harabasz index are calculated so as to evaluate the clustering effect and the separation degree. And finally, adding a clustering label into the original data, and visually analyzing the distribution of each feature under different clusters by using a violin chart, thereby providing a visual basis for further analysis. According to the method, through integration of the SOM and the k-means model, the precision of ozone concentration weather typing of the small-scale industrial park is remarkably improved, and the high-concentration ozone weather is divided more finely and more accurately.

The invention discloses a daily load curve form mode clustering method considering peak-valley shape features, and the method comprises the steps: firstly building a corresponding daily load curve peak-valley shape feature (Peak-Valley Shape, PVS) model on daily load curve data which is subjected to data preprocessing, then building a dynamic time warping distance (DTWpvs) of the daily load peak-valley shape feature, and carrying out the clustering of the daily load curve form mode. The method comprises the following steps of: establishing a target function of minimizing the sum of distances in clusters between data samples in the clusters and the centers of the clusters, then establishing a target function of minimizing the sum of distances in the clusters (Within Cluster Distance Sum, WCDS) between the data samples in the clusters and the centers of the clusters, carrying out clustering division in a daily load curve mode by adopting a K-media clustering algorithm, and finally, determining the optimal clustering number and a final clustering result by using a silhouette Coefficient Index (SCI). According to the method, the prominent problem that the conventional clustering method only performs classification from the overall shape of the daily load curve and does not consider that the load management technology most needs to pay attention to load peak and valley shape characteristics is solved; and compared with the conventional clustering method, the method can more accurately obtain a daily load curve form mode clustering result with more representative peak and valley characteristics.

The present invention provides a method and a device for training a bearing fault diagnosis model. The training method includes the following steps: performing data segmentation on the fault vibration signals of different types of bearing faults to obtain a training set, where the training set includes a number of training samples; calculating the silhouette coefficients of the training samples in the training set under different parameters to select the optimal parameters of the enhanced composite multi-scale slope entropy; using the enhanced composite multi-scale slope entropy after parameter selection to extract fault features to obtain a training feature set; and using the training feature set to train the honey badger algorithm optimized kernel extreme learning machine to obtain a trained composite multi-scale fault diagnosis model. The present invention can effectively identify bearing faults of different types and damage degrees, and the average recognition rate reaches 99.9%.

This invention provides a tailings particle classification method based on multi-dimensional morphological feature compression and clustering, belonging to the fields of geotechnical engineering, mineral processing, and particle classification technology. The method includes: S1-S6: acquiring an STL file database of tailings particles; S7: calculating and extracting morphological parameters to form an original parameter dataset; S8-S10: obtaining a three-dimensional morphological parameter set retaining the original morphological feature information; S11-S12: determining the optimal number of clusters using the elbow rule, silhouette coefficient, and Gap statistic, and using the K-means algorithm to cluster the three-dimensional morphological parameters into particle morphology groups; S13: using the discrete element method to conduct a stacking test simulation on the clustered particle categories, analyzing the stacking density and average coordination number of different particle categories, and verifying the physical meaning of the classification results. This tailings particle classification method can systematically and efficiently integrate multi-dimensional particle morphological information and directly correlate this information with macroscopic physical and mechanical properties. It is data-driven, has a clear mechanism, and the results are interpretable.

The present invention discloses a method for fine division of fracture-vug reservoir rock facies, including: obtaining single-well data of a region to be measured and performing preprocessing; performing depth alignment on required core physical property data; performing principal component analysis on the single-well data to obtain dimension-reduced and noise-reduced fusion characterization data; dividing it into a clustering data set and a test data set; using the elbow method to select the optimal classification cluster for the clustering data set; using a clustering algorithm improved by fuzzy C-means to classify the data and establishing an identification model for fracture-vug reservoir rock facies; using the silhouette coefficient to evaluate the identification model for fracture-vug reservoir rock facies; combining the test data set to test the identification model for fracture-vug reservoir rock facies; using the test identification model for fracture-vug reservoir rock facies to output a test clustering result and performing sensitivity analysis using a rank correlation matrix; using a sensitivity curve to perform crossplotting to obtain an identification chart for fracture-vug reservoir rock facies.

The present invention provides a pavement risk early warning method based on unsupervised learning of FE-FCM multi-dimensional diseases, comprising the following steps: S1. Obtain pavement disease data; S2. Perform missing value and standardization processing on the pavement disease data, and then perform feature extraction to reduce the data dimension; S3. Based on the FCM model of the sample dispersion probability in the cluster, calculate the pavement performance risk level on each road section; S4. Give early warning to high-risk road sections according to the calculation results. The present invention has the characteristics of easy implementation and high compatibility. Multi-feature data under road diseases can be flexibly selected according to the disease characteristics of different test sections, and the calibrated number of categories can be flexibly selected through the optimal target of the silhouette coefficient. Moreover, the model can be calibrated by continuously improving the later data to improve the accuracy of the model.

The present application belongs to the field of near infrared spectroscopy, and relates to a near infrared spectrum FS-SC semi-supervised feature selection method for Taiping Houkui tea. The present application solves the problem of poor prediction ability of the calibration model caused by the fact that the existing technology cannot fully mine key variables of features by relying on a small amount of labeled near infrared spectrum samples. In the method of the present application, supervised Fisher Score score calculation is performed on the labeled samples, and unsupervised Silhouette Coefficient score calculation is performed on the unlabeled samples. A weight factor is added between the supervised Fisher Score score and the unsupervised Silhouette Coefficient score to obtain the final semi-supervised score. The semi-supervised score is sorted, and the highest score is the feature variable to be selected. The feature variable is used to train a partial least squares prediction discrimination model, which simplifies the model complexity, improves the generalization ability, improves the model precision, and improves the robustness of the model.

The invention relates to the technical field of semantic analysis, in particular to a multi-modal-based semantic analysis model construction method and system, and the method specifically comprises the steps: constructing cross-modal semantic similarity through analyzing the global semantic similarity condition and potential semantic impulse probability between each text data and corresponding image data; all text and image feature vectors are clustered, the clustering quality is evaluated according to contour coefficient distribution, and the semantic coverage completeness is constructed in combination with the feature vector direction consistency degree and the cross-modal semantic similarity so as to adjust temperature parameters in contrast learning, so that when incremental training is carried out, the accuracy of the temperature parameters is improved, and the accuracy of the temperature parameters is improved. The model can dynamically select difficult negative samples according to a whole test set, the representation learning ability of the negative samples is enhanced, and overfitting is prevented, so that the whole feature space distribution is more reasonable and ordered, the robustness of the multi-modal semantic analysis model in a cross-modal semantic alignment task is improved, and the accuracy of the cross-modal semantic alignment task is improved. And training and optimization of the multi-modal semantic analysis model are realized.

This application provides a method, apparatus, and electronic device for evaluating the effectiveness of SNP genotyping based on silhouette coefficients, relating to the field of SNP genotyping technology, and alleviating the technical problems in evaluating the effectiveness of SNP genotyping. The method includes: acquiring endpoint fluorescence intensity data for a target SNP locus; performing clustering based on the endpoint fluorescence intensity data using a clustering algorithm to obtain a first clustering result; filtering samples in the first clustering result using individual silhouette coefficients and a preset silhouette coefficient threshold to obtain a sample filtering result; and clustering the sample filtering result using the clustering algorithm to obtain a second clustering result.

The present invention discloses a track analysis method and system that integrates frequency domain features and multi-dimensional evaluation indicators, including: filtering abnormal flight numbers in the track data collected by a track data acquisition device; obtaining second track data through fast Fourier transform, and extracting high-frequency energy features and peak frequency features; screening out target track data to be used for analysis through a unified standardization processing method; adaptively determining the interval where the optimal number of clustering clusters is located; traversing each number of clustering clusters within this interval, and determining the optimal number of clustering clusters K according to the silhouette coefficient and the Davies-Bouldin index, so as to perform clustering analysis through the K-Means++ algorithm according to the optimal number of clustering clusters K; visually displaying the clustering analysis results of the target track data. It can provide a theoretical reference for the clustering analysis of the arrival and departure tracks of aircraft in the terminal area through an analysis method that integrates frequency domain features and multi-dimensional evaluation indicators, making the optimization results conform to the actual situation and having more practical application potential.

The present application discloses a new intention discovery method, device, equipment and storage medium based on clustering. The present application first pre-trains a classifier according to known intention data, and then selects the number of clusters through the optimized silhouette coefficient, with a better clustering effect. The mode of combining known intention data and unlabeled data to train the classifier uses the known intention data of the previous round as a supervision signal during iteration, continuously updates the known intention data, and stops iterating until no new intention is added, and outputs the alignment label of the discovered new intention. It makes full use of the known intention data, enhances the information exchange between the classification and clustering processes, is more conducive to guiding the clustering process and accurately and fully discovering new intentions, thereby solving the technical problems of the prior art that do not make full use of the data of known intentions, do not consider the differences between new intentions and known intentions, resulting in poor clustering effects and difficulty in accurately and fully discovering new intentions.

A method includes: performing a plurality of times of first clustering processing on a plurality of training images based on a plurality of target features to obtain a plurality of first clustering results, where each of the plurality of first clustering results corresponds to one silhouette coefficient, and the silhouette coefficient indicates cluster quality; determining a first target clustering result based on the silhouette coefficients, where the first target clustering result includes M clustering categories; and performing second clustering processing on the plurality of training images based on the plurality of target features to obtain a second clustering result, where a quantity of clustering categories included in the second clustering result is M.

The invention provides a commercial building energy consumption anomaly detection method based on a graph method, and relates to the technical field of intelligent constructions.The method comprises the steps that a graph structure is constructed based on similarity, nodes are mapped into sub-item and sub-partition sub-tables, and edge weights represent time sequence similarity between the corresponding nodes; generating spectral distribution representing topological characteristics of the graph structures by calculating a Laplacian matrix eigenvalue set of the graph structures, and performing hierarchical clustering on the plurality of graph structures based on a Batta-just distance between the spectral distribution to obtain energy consumption mode clustering results in different time periods; and carrying out contour coefficient analysis on the clustering result, screening a class cluster number corresponding to a maximum contour coefficient value as a corresponding partition, and defining a class cluster set containing historical normal data as a baseline class cluster. According to the invention, the accuracy of building energy consumption anomaly detection can be improved.

The present invention discloses a method for visually extracting high-incidence areas of contact network defects, comprising the following steps: S1, marking the contact network defect records with spatiotemporal information; S2, adaptively clustering the marked contact network defect data, and calculating the total silhouette coefficient for the clustered contact network defect data. When the total silhouette coefficient is the largest, the high-incidence area of ​​defects is output, otherwise the process proceeds to step S3; S3, increasing the number of cluster center points, and returning to step S2 for iteration. The present invention is the first to apply the adaptive K-Means algorithm to contact network defect analysis, and can accurately extract high-risk defect areas without complex processing of the defect records. This method provides effective theoretical support for the formulation of contact network maintenance plans, and is used to optimize the deployment of maintenance resources and ensure the safe operation of the contact network.

The invention relates to the technical field of customer management and data analysis, in particular to a density information entropy-based adaptive clustering optimization method, which comprises the following steps of: acquiring and preprocessing original transaction data of a customer; constructing a TFA customer subdivision model, calculating three types of indexes of customer development space, purchase frequency and average purchase amount, performing standardization, performing weighting through an analytic hierarchy process to form a weighted feature vector, and constructing a data set; based on a density information entropy optimization K-means algorithm, adaptively determining an optimal cutoff distance, selecting an initial cluster center according to a density and distance criterion, and iteratively completing clustering; and determining an optimal cluster center number and outputting customer grouping and value interpretation results by adopting contour coefficient and Clinski-Harabasz index evaluation. According to the method, the static value and the dynamic potential of the customer can be scientifically described, the clustering stability and accuracy are improved, the business interpretation of a subdivision result is high, accurate marketing and resource optimization configuration of an enterprise can be supported, and the method is adaptive to the customer subdivision requirements of multiple industries.

The invention discloses a clustering and ensemble learning-based power load prediction method, system and equipment and a medium. The method comprises the following steps of: acquiring sample data from historical power load data by using a sliding window method, clustering the sample data by using a clustering algorithm, calculating contour coefficients under different clustering numbers, and selecting an optimal cluster; training a plurality of pre-selected prediction algorithms to obtain corresponding prediction errors, analyzing differences among different prediction algorithms, and selecting an optimal base learner combination; in each cluster, the selected prediction algorithms serve as element learners one by one and are trained, the prediction algorithm with the minimum prediction error serves as the element learner of the cluster, and an integrated model is established for each cluster; and acquiring a power load data sequence acquired in real time, determining the cluster to which the power load data sequence belongs, and calling the corresponding integrated model to carry out power load prediction. According to the method, load classification characteristics and multi-model advantages are fused, and the prediction precision and generalization ability in a complex scene are improved.

The present invention relates to a user segmentation method based on dimensional iterative segmentation and improved silhouette coefficient clustering, and belongs to the technical field of cluster analysis application. The present invention uses various basic information of users as features, first iteratively segments the data set through the kernel probability density function based on feature sorting, and processes the outliers generated in the segmentation to obtain multiple compact subclusters; secondly, each subcluster is used as a node, and the centroid distance between subclusters is used as the edge weight to construct a completely undirected graph, and the edge weight is scaled by the neighborhood interaction degree between subclusters. The Dijkstra algorithm is used to find the shortest path between subclusters to generate a distance matrix, and the subcluster density is calculated to obtain the density difference matrix between subclusters; finally, a subcluster aggregation method based on an improved silhouette coefficient measurement model is used to merge the final cluster structure, and the outliers are marked using a voting mechanism to complete clustering. The present invention effectively improves the accuracy of user segmentation results.

The application relates to an Internet of Things data monitoring method and system, which uses the best cluster number to divide the data of Internet of Things equipment into different clusters through a K-Means clustering algorithm, calculates the average value and standard deviation of each electrical variable in each cluster, and determines the normal range and abnormal threshold of each electrical variable in each cluster; the method can comprehensively consider the connection between the electrical variables to determine the normal range and abnormal threshold of the electrical variables in each cluster, thereby better realizing the monitoring and analysis of the state and performance of the equipment, and realizing the optimization and fault prevention of the equipment; the application also uses a silhouette coefficient as an evaluation index of the clustering effect to determine the best cluster number, avoids the interference of artificial experience and rules, and improves the intelligence and self-adaptive capability of the clustering; the application uses a visual tool to display the data and clustering results of the equipment, so that the user can intuitively understand the condition of the equipment.

The present invention proposes a direct path resolution method based on adaptive parameter DBSCAN clustering. First, the distance distribution matrix is calculated and the average nearest neighbor distance matrix is obtained by sorting and averaging operations. On this basis, the normalized distribution probability histogram obtained according to the distance matrix is filtered using a Savitzky‑Golay filter. Secondly, the peak search algorithm is used to search for the interval of the optimal neighborhood radius, and the mathematical expectation method is used to calculate the minimum number of points corresponding to the interval of the optimal neighborhood radius. Thirdly, DBSCAN clustering is performed based on the neighborhood radius and the minimum number of points and the silhouette coefficient is calculated, and the clustering result corresponding to the maximum value is selected. Finally, the maximum likelihood function is calculated based on the number of points, average value, variance and ToF characteristics of the cluster, and then the direct path is identified. The present invention performs clustering according to the characteristics of the data itself without having to determine the number of clusters, the neighborhood radius and the minimum number of points within the neighborhood radius, thereby improving the system's ability to resolve direct paths.