119 results about "Density based clustering" patented technology

The invention discloses a multipath ghosting suppression method for a millimeter-wave radar tunnel scene, which is applied to the field of millimeter-wave radar target detection and aims to solve the problem of poor multipath ghosting suppression effect in the tunnel scene in the prior art. The method comprises the following steps: firstly, constructing a confidence criterion by using multi-dimensional information point cloud data generated by radar original echoes, and performing density-based clustering on the point cloud data; performing linear fitting on the zero Doppler point cloud to obtain slope information of the tunnel wall, matching the slope of the perpendicular bisector of the dynamic point cloud pair with the slope set, and judging the target pair with lower confidence in the successfully associated target pair as a multipath ghosting; and finally, in consideration of motion characteristics of targets in the tunnel and a multipath reflection geometric model, performing transverse and longitudinal double-rectangular region division on the dynamic point cloud, performing Doppler information constraint on point cloud data in the region, and screening out the targets which meet conditions and are relatively low in confidence coefficient and judging the targets as multipath ghosts. According to the invention, vehicle targets in a semi-closed environment can be detected, and multipath ghosts generated by reflection of tunnel walls, handrails and the like can be effectively identified and inhibited while real targets are reserved.

The invention discloses a rapid aggregation classification method based on feature similarity calculation, and relates to the technical field of image processing, and the method mainly comprises the steps: in a transfer learning mode, obtaining scene high-dimensional feature information through attention mechanism integration after dimension unification of each training image and corresponding gradient features and frequency domain features; performing reference core feature selection under distance comparison based on infinite norm according to scene high-dimensional feature information obtained by each training image; according to the reference core features, obtaining a feature matrix under noise point elimination by using a density-based clustering algorithm; gradient features and frequency domain features of the target input image are extracted, attention mechanism integration after dimension unification is carried out on the gradient features and the frequency domain features and the target input image, and target high-dimensional feature information is obtained; and performing classification judgment on the target input image according to the distance between the target high-dimensional feature information and the reference core feature in the feature matrix. According to the invention, the adaptability and flexibility of system classification discrimination are greatly improved.

The invention provides a software code defect detection method and system based on program code feature fusion, and the method comprises the steps: carrying out the grammar and structure inspection of a code through a static analysis tool, and extracting defect features related to control flow chaos, exception handling and resource leakage; running a preset test case through a dynamic analysis tool, capturing an execution path and an abnormal behavior in a runtime state, and obtaining defect features which are not covered by static analysis; if the output result of the decision tree model is that defects exist, grouping comprehensive defect feature vectors by adopting a density-based clustering algorithm, determining the types and distribution ranges of the defects, and selecting the density clustering algorithm to process defect features in non-spherical distribution; and according to a clustering result, generating a defect repairing suggestion, combining code logic and execution path information, and determining a specific position and an influence range of the defect. According to the method, the stealth defect in the code can be effectively identified and positioned, the software quality and reliability are improved, and timely and accurate defect repairing guidance is provided for developers.

The invention provides a side slope radar point cloud data slice abnormal value complementing method based on space-time fusion, which comprises the following steps: acquiring point cloud data of a plurality of time slices obtained by scanning of a side slope radar, and mapping the point cloud data into a three-dimensional point cloud; identifying a piece of abnormal regions of the three-dimensional point cloud through statistical filtering and a density-based clustering method; constructing a point cloud processing model based on PointNet + +, generating masked data and a mask index by adopting a mask strategy, and masking off pieces of point clouds; spatial features are extracted in different regions, static space weight matrixes generated by longitude and latitude and time difference global features are fused, deformation values of mask regions are predicted through a full-connection network, and model parameters are optimized; and inputting the trained model into latitude and longitude coordinates of an abnormal region of actual missing point cloud data, performing abnormal value completion, outputting a completed deformation value, and calculating a completion precision index. According to the method, the completion effect is good, all abnormal values of multiple time slices can be completed at a time, and the calculation efficiency is improved.

The invention discloses a blockchain technology-based switch security management and data integrity method, which comprises the following steps of S1, deploying blockchain nodes in a network, and accessing each switch as a node to the blockchain network; s2, using a chain structure to store configuration data and operation log information of the switch, and generating a digital signature; s3, configuring an adaptive consensus mechanism; s4, training the threat prediction model by using a support vector machine algorithm, and detecting and identifying known abnormal behaviors and potential threats; s5, performing clustering analysis on the data in the switch network in combination with a density-based clustering algorithm and an improved isolated forest algorithm, identifying unknown abnormal behaviors and potential security threats, and triggering an early warning mechanism; s6, deploying a smart contract for switch authority management and configuration recovery operation; and S7, providing a user visual interface. According to the invention, a self-adaptive consensus mechanism and an intelligent contract are utilized, so that switch security management and data integrity are realized.

The invention requests to protect an aerial photography small target detection method based on regional clustering and a feature attention mechanism. The method comprises the following steps: firstly, taking a central coordinate of a candidate target extracted from an input image as an input feature of a clustering algorithm; then clustering the targets by using a density-based DBSCAN clustering algorithm, cutting out the regions from the image according to a clustering result, performing center filling processing on the cut-out regions to keep the scale and the ratio in a reasonable range, and integrating the cut-out regions with original data to obtain a new expanded data set; and putting the expanded data set into the improved single-stage target detection network RetinaNet for training, and performing final target classification and positioning by using a detection head based on an attention mechanism. The method can achieve the effect of data enhancement. In addition, the detection head based on the attention mechanism can remove interference of useless information in classification and regression tasks, so that the network can adaptively pay attention to more appropriate information for classification and regression.

The invention discloses a railway vehicle traction filter screen detection method, device and equipment and a storage medium, and relates to the technical field of railway vehicles. The railway vehicle traction filter screen detection method comprises the steps that working condition parameters of a traction system in a railway train during normal operation are obtained; preprocessing the working condition parameters according to a preset data cleaning sequence to obtain a preprocessed data set; executing a preset normalization processing operation on the data set; and carrying out multiple times of sampling from the processed data set by using a density-based clustering algorithm, and training a density-based clustering model. The method solves the problems that in current actual subway operation and maintenance work, blockage of a filter screen of the traction converter is an important reason for abnormal temperature rise of the traction converter, and due to the fact that different line conditions are greatly different, and different operation environments such as tunnels and viaducts exist, a traditional pressure or temperature threshold value judgment method is poor in effect; and automatic learning cannot be carried out according to changes of line conditions.

The invention provides a virtual power plant resource partitioning method based on a multi-layer perceptron. The method comprises the following steps: acquiring geographical location information of each resource in a virtual power plant; adjusting a density-based clustering algorithm according to the geographical location information of each resource in the virtual power plant, and performing clustering analysis on the geographical location information of each resource by using the adjusted density-based clustering algorithm to obtain a plurality of first clustering results; counting the number of geographical location information contained in each first clustering result, adjusting the first clustering results according to the number to obtain a plurality of second clustering results, and taking a range formed by each second clustering result as a partition; and inputting the geographical location information of each resource into a pre-trained multi-layer perceptron to obtain an importance degree score of each resource. According to the method provided by the invention, each energy resource can be partitioned and effectively regulated and controlled.

The invention discloses an advanced blasting pressure relief method based on rock burst prediction, and the method achieves the precise recognition and efficient intervention of rock burst risks through the combination of a micro-seismic monitoring technology, energy thermodynamic diagram modeling and an intelligent algorithm. According to the method, micro-seismic data is deeply analyzed, an abnormal activity area is identified, an energy thermodynamic diagram is constructed, a stress concentration situation is visually displayed, and risk level prediction is carried out in combination with historical data, so that early warning accuracy and response timeliness are improved, an optimal blast hole position in a high-risk area is automatically identified by adopting a density-based clustering algorithm, and the early warning efficiency is improved. Classification and directional blasting pressure relief are implemented according to risk levels, local stress is effectively released, the occurrence probability of rock burst is reduced, the overall process is highly automatic, monitoring, prediction, decision making and intervention are integrated, and the method is suitable for various complex coal mine geological environments.

The application discloses a kind of multi-path ghost suppression method of millimeter wave radar tunnel scene, it is applied to millimeter wave radar target detection field, for the poor multi-path ghost suppression effect in tunnel scene of prior art;The application first constructs confidence criterion using the multidimensional information point cloud data generated by radar original echo, and carries out density-based clustering to point cloud data;Then linear fitting is carried out to zero doppler point cloud to obtain the slope information of tunnel wall, simultaneously, the median line slope of dynamic point cloud pair is matched with the slope set, and the target pair associated successfully is judged as multi-path ghost with lower confidence;Finally, considering the motion characteristics of target in tunnel, as well as multi-path reflection geometric model, the double rectangular region division of transverse and longitudinal is carried out to dynamic point cloud, and the point cloud data in region is subjected to Doppler information constraint, and the target satisfying the condition and with lower confidence is screened out and judged as multi-path ghost.The application can detect vehicle target in semi-closed environment, while retaining real target, can effectively identify and suppress the multi-path ghost generated by reflection of tunnel wall and railing etc..

Disclosed in the present application are an anomaly recognition method and an anomaly recognition system for tracks of trucks. The method includes: obtaining a running track T according to global positioning system (GPS) data of running of a truck to be recognized; employing a track compression algorithm for the running track T to obtain a compressed track set C; employing a density-based clustering algorithm and performing grouping according to set time periods to obtain a network graph representing a motion track of each time period; inputting the network graph into a track embedding model established and trained in advance to obtain an explicit embedding vector corresponding to each network graph; and determining stability according to a distance between the vectors, and classifying points with the stability lower than a set threshold as abnormal tracks. The track embedding model is realized by employing a Skipgram model on the basis of a graph2vec algorithm.

The invention relates to the technical field of medical health monitoring and neural networks, and provides a chronic obstructive pulmonary disease patient leaving hospital monitoring system which comprises the steps that multi-source data of a patient is acquired, and the multi-source data at least comprises daily health monitoring data and electronic medical record data; preprocessing and feature extraction are conducted on the multi-source data through a data processing module, health feature data of the patient are obtained, and the health feature data are features related to the chronic obstructive pulmonary disease condition selected from the multi-source data and comprise features based on respiratory functions, features based on lifestyles and behaviors and features based on complications and complications; and features based on drug use; integrating and updating the health feature data through a patient health condition evaluation module, and performing clustering analysis on the integrated and updated real-time health data by adopting a density-based clustering algorithm DBSCAN to obtain a health portrait of the patient; the health portrait is analyzed through a medical large model analysis module, an illness state evaluation result and a health management suggestion are obtained, and a deep learning model designed by combining a convolutional neural network CNN with a multi-modal model architecture of an attention mechanism Transform is used in the medical large model analysis module; remote monitoring, interactive communication and personalized intervention operation are executed through the remote medical intervention module according to the illness state evaluation result.

The invention discloses a land reclamation planning surveying and mapping collaborative operation method based on geographic information, and particularly relates to the technical field of planning surveying and mapping. A time sequence analysis method is introduced, dynamic changes of land reclamation areas are analyzed, in combination with a density-based DBSCAN clustering algorithm, time sequence dynamic clustering is performed on the areas, areas with similar characteristics are identified, a risk assessment model is constructed through a clustering result, and risk levels of different areas are assessed. The cloud platform is utilized to support collaborative operation of planning surveying and mapping personnel, and the risk assessment result of the region is updated in real time, so that the land reclamation scheme is automatically adjusted, matching of planning and actual conditions is ensured, the problem that time dynamic characteristics are neglected in the prior art is effectively solved, the timeliness and accuracy of the land reclamation scheme are improved, and the land reclamation method is suitable for large-scale popularization and application. And the management efficiency and effect of overall land reclamation are improved.

The invention discloses a method and device for detecting the uniformity of non-woven fabric through a flash evaporation method, relates to the field of fabric production and detection, and can solve the technical problem that an existing method for detecting the uniformity of the non-woven fabric through the flash evaporation method is large in delay performance. Performing time sequence analysis processing on the rolling pressure data and the light transmission intensity data to obtain a pressure change curve and a light transmission change curve; obtaining a uniformity curve of the detection sample based on an interpolation algorithm according to the pressure change curve and the light transmission change curve; marking outliers of the uniformity curve of the detection sample based on a maximum likelihood estimation method to obtain an outlier set; and according to a density-based clustering algorithm, obtaining a clustering analysis result of the outlier set, and according to the clustering analysis result, obtaining a uniformity detection result of the detection sample so as to adjust process parameters according to the uniformity detection result.

The invention discloses an analysis method and system based on AI video recognition behavior monitoring, and relates to the technical field of video recognition, and the method comprises the steps: collecting video data, calculating the overall pixel gradient change of adjacent frames, marking preliminary candidate frames, calculating the edge saliency intensity of each frame, and analyzing a salient region point set; clustering and calculating the geometric center point of each cluster as an anchor point value, and forming an anchor point set through the center points of each cluster; according to the method, concentrated extraction of salient points is realized through a density-based clustering algorithm DBSCAN, so that an extraction result has logic uniqueness and local integrity of a salient target, a local feature map is extracted by using a shallow convolutional network, target texture and edge information is effectively captured, bounding box optimization is realized by rotating an IoU loss function, and the robustness of the target is improved. And the fitting precision of the boundary of the salient region is improved.

The invention discloses a road abnormal event detection and early warning method and system based on unsupervised lifelong learning. The method comprises the following steps: acquiring traffic scene data by using various sensors along a road and at a vehicle end, and preprocessing the traffic scene data; performing unsupervised feature extraction on the image data by using an auto-encoder, and performing feature dimension reduction on non-image data by using a principal component analysis method to extract main features; an anomaly detection model is initialized, a K-Means clustering algorithm is adopted to divide the feature space into clusters, and a normal cluster set of an initial normal traffic mode is formed; receiving new traffic data, and continuously performing online learning and updating on the model by adopting an incremental learning method, a density-based clustering algorithm and an elastic weight integration method; newly collected data are input into the updated model for anomaly detection and road anomaly event alarm; according to the method, efficient and accurate detection and timely early warning of road abnormal events are realized through unsupervised feature extraction, model initialization, lifelong learning and a model updating mechanism.

The invention provides a concrete pumping pressure loss prediction method and system based on machine learning, and belongs to the technical field of machine learning. According to the scheme, the method comprises the steps that data samples are collected, and a high-dimensional feature data set containing concrete raw material characteristic parameters and pumping structure parameters is constructed; carrying out unsupervised hierarchical processing on the data samples by adopting a density-based clustering algorithm, taking a cluster number to which each sample belongs in an unsupervised clustering result as a new feature, and supplementing the new feature into the high-dimensional feature data set to obtain an expanded high-dimensional feature data set; and based on a machine learning algorithm, establishing a concrete pumping pressure loss machine learning prediction model, training the concrete pumping pressure loss machine learning prediction model, and performing pumping pressure loss prediction on the to-be-predicted data by using the trained pumping pressure loss prediction model to obtain a pumping pressure loss prediction result. Therefore, the nonlinear and non-uniform density distribution sample structure is identified through the density-based clustering algorithm, and the pumping pressure loss prediction precision is improved.

The invention discloses an intelligent electric vehicle stability control method based on a stability probability spectrum. A data acquisition module, a density-based clustering algorithm module, a probability spectrum generation module based on a convolutional neural network, a self-adaptive model prediction control module and an actuator cooperative distribution control module are included. According to the invention, the data acquisition module is used for acquiring data to calculate the lateral load transfer rate, and the density-based clustering algorithm is started to classify the stability state of the vehicle; and then, generating a stability probability spectrum through a convolutional neural network, dynamically adjusting a weight factor in a minimum cost function in a control target and model prediction control based on the stability probability spectrum, and coordinating response priorities of a plurality of actuators by judging the stability state of the vehicle so as to enhance the transverse and longitudinal coupling control capability. The method can achieve the dynamic recognition and hierarchical control of the stability state of the vehicle, and effectively improves the operation stability and control robustness of the whole vehicle under a complex working condition.

The invention relates to energy supply of load nodes in a wireless power transmission network, in particular to a method for constructing and optimizing an energy transmission path in the wireless power transmission network. The invention discloses a method for constructing and optimizing an energy transmission path in a wireless power transmission network. The method comprises the steps of constructing a sub-network based on a DBSCAN clustering algorithm, constructing an inter-cluster energy transmission path based on a Prim algorithm and optimizing an intra-cluster energy transmission path based on a breadth-first search algorithm. The method comprises the following steps: firstly, dividing a whole wireless power transmission network into a plurality of clusters by adopting a density-based clustering algorithm DBSCAN, constructing a shortest energy transmission path from a cluster head to each load node in the cluster through a breadth-first search algorithm, and constructing a shortest path between the clusters on the basis of a Prim minimum spanning tree algorithm; therefore, the optimal energy transmission of the whole network is realized.

A method may comprise receiving airspace surveillance data from a plurality of sensors, the airspace surveillance data comprising tracks associated with one or more targets, aggregating the airspace surveillance data to obtain aggregated data, performing density-based clustering of the aggregated data to obtain a plurality of clusters, determining one or more candidate associations between the tracks and the clusters, associating each of the tracks with one of the one or more targets based on the candidate associations, and estimating a track for each of the one or more targets based on the associations between the tracks and the one or more targets.

The invention discloses a household optical storage all-in-one machine dynamic clustering method and system considering space-time factors, and relates to the technical field of power system operation, and the method comprises the steps: receiving the node information of a power distribution network, including the number of nodes in the power distribution network, the distance between the nodes, and the branch unit impedance; light storage all-in-one machine user related data is obtained based on the power distribution network node information, the light storage all-in-one machine user related data is preprocessed, a feature matrix is generated, and the light storage all-in-one machine user related data comprises household light storage all-in-one machine equipment information, user position information, household power consumption data and photovoltaic power generation data; on the basis of the feature matrix, unknown parameter values of a density-based clustering method DBSCAN with noise points are obtained in a self-adaptive mode by applying a KNN algorithm, and the unknown parameter values comprise a neighborhood distance threshold value Eps and a neighborhood sample number threshold value MinPts with the distance of a certain sample being Eps; and clustering all household optical storage all-in-one machines according to eps and MinPts, classifying the clustered noise points, repeating the steps, and carrying out secondary clustering to obtain a final clustering result.

The present invention relates to a data processing method and device based on a distributed computing model, belonging to the field of data processing technology. The method comprises the following steps: using a hash algorithm to preliminarily partition data, identifying and adjusting abnormal partitions, and refining abnormal partitions through fine-grained partitioning; identifying nodes with the same time window and generating an access policy set, selecting the optimal access set for task allocation; selecting core objects, expanding cluster clusters through a density-based clustering algorithm, and merging the clustering results of all Map tasks. The present invention improves the uniformity of the amount of data processed by each Reducer through a two-stage partitioning strategy, and then uses a cluster analysis method to find the optimal access set of the current distributed processing server node for task allocation, thereby improving data processing efficiency and resource utilization. Finally, a density-based clustering algorithm is used to remove noise points, thereby improving the accuracy of the clustering results.

The application discloses an environmental monitoring data analysis method and related device, relates to the technical field of environmental monitoring, obtains the standard sample data consistent with the attribute of the to-be-analyzed environmental monitoring data, calls a density-based clustering algorithm to process the two kinds of data, obtains a clustering result, and in the case that a turning point in a K-distance graph of the to-be-analyzed environmental monitoring data does not satisfy a preset turning point condition, determines a reference turning point by using a sliding window algorithm, so that the clustering result is more accurate; when the clustering result satisfies a condition, multi-dimensional quality analysis is performed on the to-be-analyzed environmental monitoring data, and the efficiency and accuracy of environmental monitoring data analysis are further improved; when the to-be-analyzed environmental monitoring data is subjected to multi-dimensional quality analysis, at least data stability and correlation of the to-be-analyzed environmental monitoring data are analyzed, so that the accuracy of an analysis result obtained by multi-dimensional analysis is greatly improved, and the accuracy of application of the analysis result to other scenes, such as data quality evaluation, is further improved.

The invention discloses a synchronous incremental modeling method based on slice scanning. The method comprises the following steps: step 1, obtaining object edge information according to hierarchical scanning data; step 2, using a density-based clustering algorithm to perform clustering analysis on the single-layer contour points so as to distinguish inner contours and outer contours of different areas, the clustering result including a plurality of sub-contours, and performing subsequent data matching and grid generation by taking each sub-contour as a unit; step 3, performing operation by using two adjacent layers of data, and generating a grid between the two layers in a mode of matching a nearest point; 4, optimizing the efficiency of a density-based clustering algorithm by using a parallel operation framework, and ensuring the synchronism with a data acquisition process; and 5, performing grid simplification and smoothing operation on the grid model. The method supports synchronous modeling of hierarchical data and adapts to anisotropic sampling, so that the modeling efficiency and the structure reduction precision are effectively improved.

The present subject matter relates to a system (100) and a method (400) for demand-based optimization of airline ticket pricing. The disclosed system (100) includes is a user interface (101) that facilitates the input of flight information, subsequently processed by an integrated memory (203) and processor (201). The machine learning-driven price optimization module (204) involves fetching price range and contextual information, extracting features, and further utilizing a classification model (205) for identifying demand cluster probabilities, and calculating a demand score. The system (100) further refines this demand score to account for market fluctuations. Notably, it employs advanced techniques like density-based clustering for demand segmentation and ML explainability through the Airline Experience Quotient (AEQ). This ensures transparency and enhances the user experience. Additionally, the system's capability extends to efficient model deployment, leveraging MLOps, and presenting the optimal ticket price to users for informed decision-making.

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.

The present application belongs to the technical field of trajectory smoothing method, and particularly relates to a personnel trajectory smoothing method, comprising the following steps: acquiring satellite positioning data; and performing deviation correction treatment on the drift points in the data; adopting a density-based clustering analysis algorithm to cluster the knotted points in the trajectory line due to trajectory return into one point, while ensuring the time sequence of the trajectory point set. The present application performs deviation correction treatment on the drift points in the satellite positioning data by applying Kalman filtering algorithm; the present application adopts the density-based clustering algorithm to cluster the knotted points in the trajectory line, eliminates the problem of trajectory line confusion caused by knotting, and makes the presented trajectory more clear and smooth. The present application ensures the time sequence of the trajectory points in the set after executing the clustering algorithm.

A kind of federal learning model attack defense method based on neural network feature extraction, before carrying out federal learning global aggregation in each round, neural network feature extraction model is constructed in advance in safe environment, and it is trained using public data set, trust of local parameter is guided by server itself;In online stage, the local parameters uploaded by each client are extracted and output to the server using the trained neural network feature extraction model, and the server uses the density-based clustering method (DBSCAN) with noise to classify the feature vectors and eliminate the corresponding malicious local parameters, to realize safe aggregation.The present application trains neural network in safe environment before each iteration, so that it can resist various model poisoning attacks in online stage, thereby effectively preventing the damage of malicious client to global model.

This invention discloses a method and system for identifying cross-connection issues in low-voltage distribution areas. It acquires voltage and current time-series data of users in the low-voltage distribution area, cleans the data using a mask interpolation strategy, and adaptively extracts voltage drop and current surge features based on the median absolute deviation algorithm. It constructs a voltage trend similarity model based on local regularization kernels and a mutation co-occurrence similarity model based on event level weights, employing a dual-track strategy combining daily rigorous assessment and feature revival to generate suspected voltage correlation pairs. A density-based clustering algorithm is used to generate an initial population, and a voltage similarity-weighted centroid screening mechanism is introduced to remove loose outliers in the feature space. A pairwise cross-validation model is constructed based on electrical causality, and the population is topologically refined using relative voltage response criteria and a maximum fully connected clique extraction algorithm to determine the final cross-connection population. This invention significantly improves the accuracy and robustness of cross-connection identification.

Monitoring device for anomaly detection from sensor data (102, z) determined in time succession of a technical system (100), comprising: an input interface (201) configured to receive a new sensor data point (102, z) of the technical system (100); an analysis unit (202) configured to determine, for the new sensor data point (102, z), a minimum reachable distance (minRD) with respect to a selection of neighboring training data points (x1, x2, x3) and to determine, for the new sensor data point (102, z), a minimum position (xi) in a sequence of training data points in a reachable graph (312) created by means of a density-based clustering function using predetermined training data points as input values, to insert the new sensor data point (102, z) in the reachable graph after the minimum position (xi) and before a training data point having a greater reachable distance than the determined minimum reachable distance (minRD), to assign the new sensor data point (102, z) to a determined cluster according to its position in the supplemented reachable graph and as a normal or abnormal condition according to the minimum reachable distance (minRD); and an output interface (203) configured to output the determined assignment for the new sensor data point (102, z) as a result of anomaly detection.