103 results about "Weighted distance" patented technology

The invention relates to the technical field of data processing, in particular to an abnormal mode data processing system driven by power marketing big data, which comprises a distributed collaborative acquisition module for constructing a space-time alignment three-dimensional data stream, a multi-modal feature reconstruction module for separating periodic noise and quantizing environmental interference, and a data processing module for processing abnormal mode data. The resistance feature decoupling module generates a purification feature vector set and a noise confidence index through orthogonal projection, and the dynamic algorithm adaptation module dynamically schedules an isolated forest algorithm, a weighted distance measurement algorithm and a sparse self-encoding clustering algorithm according to the noise confidence index. The behavior chain verification module establishes a combined physical rule verification mechanism of an environment temperature threshold value, a load deviation degree and an equipment state, and the closed-loop strategy engine module adaptively adjusts a feature decoupling loss function weight according to a decision boundary offset, so that the accuracy and the environmental adaptability of real electricity consumption abnormity identification in a complex noise environment are effectively improved.

The invention relates to a side cloud collaborative adaptive workflow scheduling method and system, and belongs to the technical field of distributed computing and artificial intelligence. The method comprises the following steps of: firstly, in a macroscopic candidate screening stage, reducing problem granularity through task clustering, and obtaining balance between utilization and exploration based on a weighted distance probabilistic preferential strategy; then, in a collaborative scheduling decision-making stage, a global network state diagram is constructed through a graph neural network, deep spatial features of nodes and neighborhoods of the nodes are extracted, context-aware state representation is formed, and a reinforcement learning agent makes an optimal collaborative decision in multiple options such as local execution, edge migration or cloud unloading according to the state representation; and finally, in a local adaptive optimization stage, performing fine-grained optimization after the task is issued, dynamically adjusting a scheduling frequency and a multi-target weight through an online learning mechanism, realizing balance between a task deadline and a resource utilization rate, and ensuring efficient and robust execution of a node level.

The application discloses a kind of big data-based optimization analysis method and system of territorial space planning, it is related to big data and optimization analysis technical field of territorial space planning, including real-time data acquisition and dynamic preprocessing, dynamic regional division, multi-model integrated space-time prediction, multi-objective dynamic optimization planning, real-time feedback and adaptive adjustment.The big data-based optimization analysis method of territorial space planning provided in the application adopts space-time weighted fuzzy clustering method, divides the territorial space into several sub-regions according to the spatial coordinates and time attributes of data, constructs the weighted distance regular objective function between data points and regional center, and updates the membership and the position of regional center iteratively, realizes the self-adaptation and space-time smoothing of regional segmentation, effectively captures the dynamic change characteristics in region.

The invention discloses a hydroelectric equipment anomaly detection method based on physical mechanism guidance and sequential topological entropy fluctuation characteristics, and belongs to the technical field of hydroelectric equipment monitoring and fault diagnosis. The method comprises the steps that multi-source sensor data of hydroelectric equipment is collected and preprocessed; constructing a physical weighted distance function in combination with an equipment physical mechanism, and embedding time sequence data into a high-dimensional point cloud space; extracting persistent homology features through a sliding window, generating a persistent graph sequence and calculating topological feature indexes; a persistence graph entropy fluctuation index is provided, and anomaly detection is realized by quantifying time sequence fluctuation of topological entropy; and finally, visual output and an alarm mechanism are combined to assist diagnosis. According to the method, equipment physical characteristics and topological data analysis are fused, the problems that a traditional method is insufficient in nonlinear system modeling, insensitive to dynamic evolution and the like are solved, the early warning capacity and detection precision of early faults are improved, and the method is suitable for anomaly detection application of core equipment such as a water turbine and a generator.

The invention provides an automatic test method and system for an intelligent power distribution automation terminal, and relates to the technical field of automatic test, and the method comprises the steps: obtaining the model information and environment information of a to-be-tested terminal; generating test data of the to-be-tested terminal in combination with the model information, the environment information and the Lyapunov index; inputting the test data into the to-be-tested terminal, and outputting operation data; analyzing the operation data by using a dynamic threshold value based on the potential energy field, and outputting abnormal data; fusing and classifying the abnormal data based on a pre-training model and the coupling weighting distance, and determining a fault type; and performing early warning according to the fault type. According to the analysis method based on the Lyapunov index and the dynamic threshold value, the test can comprehensively cover various functions and limit states of the equipment, so that potential anomalies and faults can be effectively found.

The application is a kind of double robust causal dose-response curve estimation method based on high-dimensional independent variables, comprising the following steps: 1) constructing a new target function based on the modified adaptive LASSO method to realize dimension reduction; 2) constructing a double weighted distance correlation coefficient DWDC to select the optimal lambda n ; 3) using DR estimator to estimate the dose-response curve. The method mainly aims at the characteristics of the potential confounding variable set contained in the health medical big data, such as high dimension, nonlinear relationship between health outcome and (or) exposure factor, etc. In the framework of GOAL method, a double robust estimation method of causal dose-response curve is provided, which is called GOALDeR method. A large number of statistical simulations show that the estimation accuracy and precision of GOALDeR method are better than those of existing methods, and the method has double robustness and is less affected by the correlation structure between covariates and the n / p (sample size / covariate dimension) ratio.

The embodiment of the invention discloses a weld defect intelligent detection method and device based on multi-modal feature fusion. The method comprises the steps that an excitation unit is controlled to apply standardized mechanical knocking excitation to a target weld joint, and response signals generated by the weld joint are collected; transforming the time domain signals corresponding to the acoustic signal and the vibration signal based on Lie group transformation to obtain an invariant feature vector so as to construct a multi-modal feature vector; the weight of each feature in the multi-modal feature vector is determined based on the physical parameters of the target welding seam, and the weighted distance between the multi-modal feature vector and each clustering center in a standard database is calculated based on the weight, so that a target clustering center is determined; and in response to the fact that the target weighted distance corresponding to the target clustering center is larger than a distance threshold value, determining the defect type of the target weld joint based on the similarity between the multi-modal feature vector and each defect feature vector. In the embodiment of the invention, higher-precision and more reliable weld defect identification can be realized, and the whole process can be automatically carried out without manual operation.

The present invention discloses a system and method for identifying a content of interest in a document. The method includes receiving a plurality of training documents that comprise a plurality of field elements, estimating for each data element present within the training document, a weighted distance of the each data element from a field element that corresponds to the content of interest. A feature vector is created based on the weighted distance and a position of the each data element with respect to the field element. A set of feature vectors are developed for the plurality of field elements and is used for training a field identification model. The field identification model is applied on the document to identify a beginning position of the content of interest.

A natural language processor extracts features from batches of unstructured traffic. A feature weighted distance engine computes a distance matrix between pairs of feature vectors for sessions of unstructured traffic using a weight vector that assigns importance to relative placement of features in feature vectors. The distance function used to compute the distance matrix with the weight vector is conducive to generating high-quality clusters and patterns in unstructured traffic. The sessions of unstructured traffic are clustered according to the pairwise distance matrix. Generated clusters are merged with clusters for previously analyzed sessions of unstructured traffic. A pattern identification engine extracts patterns from the merged clusters that correspond to behavior of applications generating the unstructured traffic.

The application relates to the technical field of big model computing power scheduling, in particular to a computing power cooperative intelligent scheduling method and system based on a big model application, which comprises the following steps: obtaining a global computing power task request containing multiple big model inference tasks, extracting multi-dimensional features according to task constraint conditions and forming a task demand vector. An improved greedy algorithm is adopted to match the task demand vector and a state vector of a dynamic heterogeneous computing power resource pool, a preliminary scheduling instruction set is generated through weighted distance, a multi-step forward-looking evaluation model with computing power resource balance and task communication overhead as joint targets is constructed, the running state of tasks and resources in multiple time slices is simulated, the scheduling instruction is corrected, and finally the instruction is issued to a hardware computing unit to complete task execution. The method can improve the matching accuracy of tasks and computing power resources, optimize the balance of resource allocation, and reduce the data communication overhead between tasks.

This invention discloses a method for optimizing the design of complex equipment based on a hybrid adaptive sampling surrogate model. This method divides the design space into several Voronoi polygons controlled by sample points. Three indices—leave-one-out error, local nonlinearity, and polygon region size—are used to evaluate the importance of each sample point region. An entropy-weighted distance method is used to select the most suitable sensitive region for adding points. A learning function based on error estimation is then combined to determine new sampling points. This iterative update constructs a high-precision surrogate model, establishing a mapping relationship between design parameters and optimization objectives, thereby solving the design optimization problem. This invention innovatively proposes a multi-index adaptive weight allocation method to select sensitive sample regions and proposes a novel learning function to determine the location of added points. This method can construct a high-precision surrogate model using fewer sample points and can be used for the optimization design of complex equipment, including but not limited to tunnel boring machines.

The invention discloses an auxiliary management method and system for rattan plaiting skill data based on image analysis, and relates to the technical field of rattan plaiting image processing.The method comprises the steps that an initial plaiting topological graph is constructed by obtaining an original rattan plaiting image, and high-confidence and interference topology first-level sub-graphs are divided through pixel quality evaluation; taking the adjacent high-confidence topology sub-graphs as reference, calculating the similarity between the adjacent boundary length and the topological structure and coding a structure prior vector, and selecting weighted fusion according to vector dispersion or dividing into interference topology secondary sub-graphs according to weighted distance so as to determine generation prior; topological sub-graphs are repaired through a conditional generation model, a complete woven topological graph is output through fusion, and precision can be improved through optimization of time sequence, symmetry and boundary consistency. According to the method, the topological structure of the interference area can be accurately repaired, and the integrity and reliability of the rattan-plaited woven topological graph are remarkably improved.

The invention relates to the technical field of mobile terminal safety monitoring, and discloses a risk signal screening monitoring method and system for a mobile terminal. The method comprises the following steps: collecting intensity change sequences of various risk signals of the mobile terminal, calculating cosine similarity and propagation momentum factors of deviation rate vectors, counting time sequence mutual information of the risk signals to determine a causal propagation relationship, constructing a directed propagation graph to predict a future risk evolution value, and calculating the risk evolution value of the mobile terminal. And calculating a weighted distance, dynamically adjusting a judgment threshold and judging a multi-step progressive attack. The technical problems that in the prior art, staged progressive attacks cannot be recognized, risk evolution cannot be pre-judged in advance, the detection sensitivity cannot be adaptively adjusted, and the response strategy is extensive and single are solved. According to the method and the device, the recognition accuracy and the response timeliness of the multi-step progressive Trojan horse attack are improved.

The invention relates to the technical field of biodiversity research, and discloses a mountain tree species diversity remote sensing estimation method, which comprises the following steps: acquiring mountain tree species distribution data, acquiring a mountain remote sensing image, preprocessing the data, and calculating a tree species diversity index according to tree species distribution. Calculating the correlation between the spectral heterogeneity index and the tree species diversity index, constructing a mountain tree species diversity remote sensing mapping model according to the influence factor calculated by the spectral heterogeneity index, extracting the optimal parameter of mountain tree species diversity remote sensing estimation, and making a tree species diversity index spatial distribution map according to the spectral heterogeneity index to realize tree species diversity estimation. A terrain correction mechanism is introduced into the spectral heterogeneity index constructed by the method, the stability and reliability of the spectral heterogeneity index are improved, spectral response of the model to different tree species is improved after a weighted distance construction mechanism between pixels is combined, and the method is suitable for mountain forest areas with obvious topographic relief and has good application prospects. And multi-source input of multi-sensor remote sensing images is supported.

The invention relates to the technical field of hydrology and water quality on-line monitoring, in particular to a hydrology and water quality on-line monitoring data interpolation method based on double-circulation weighting, which comprises the following steps of: 1) determining the interpolation sequence of each parameter and missing value of hydrology and water quality data by using information entropy; and 2) restoring a missing value by using a complete parameter and instance pool, establishing a parameter weighting matrix, weighting the Euclidean distance, and calculating a local fluctuation error. The method has the advantages that the interpolation method based on the K-nearest neighbor algorithm (KNN) is designed, the interpolation performance is improved by adopting weighted distance measurement and increasing fluctuation errors, increment loop learning is performed by utilizing variable and instance information, original data features are reserved as far as possible, meanwhile, a more accurate interpolation result is provided, and the method is suitable for large-scale popularization and application. The method is an effective tool for processing hydrology and water quality data with multiple missing modes.

The application relates to the technical field of power distribution network protection, and provides a waveform self-similarity fault type identification method and system based on a Tanimoto coefficient. The method comprises the following steps: acquiring multi-channel time sequence waveform data, pre-processing the waveform data and performing multi-scale division, and generating a multi-scale waveform feature primitive set containing time domain, frequency domain and time-frequency domain feature vectors; based on the multi-scale waveform feature primitive set, a generalized Tanimoto coefficient between different physical channels and different analysis scales is calculated, and a cross-scale and cross-channel self-similarity tensor is constructed; the self-similarity tensor is subjected to tensor decomposition, a time intensity vector of an atomic self-similarity mode corresponding to a fault mode is extracted, and is fused into a decoupled self-similarity feature vector; the decoupled self-similarity feature vector and a pre-stored standard fault feature template are subjected to weighted distance calculation and hierarchical comparison, and a fault type identification conclusion is output according to a comparison result. The method is helpful to improve the accuracy of fault type identification under complex working conditions.

The application belongs to the technical field of equipment monitoring, and particularly relates to an automatic stereoscopic warehouse equipment fault diagnosis method and system, which comprises the following steps: collecting time sequence data segments of the automatic stereoscopic warehouse equipment; extracting a mixed feature vector containing regular statistical features and time sequence shape features from the data segments; preliminarily judging the fault nature of the time sequence data segments based on the time sequence shape features to obtain a shape category; determining the diagnosis weight of the regular statistical features according to the shape category, and calculating the adaptive weighted distance between the regular statistical features of the current data segment and a preset fault prototype to determine the final fault diagnosis result. By introducing time sequence shape analysis and dynamic feature weighting, the application can effectively distinguish fault modes that are similar in statistics but different in physical causes, and significantly improves the accuracy of fault diagnosis.

The invention discloses a frequency band switching method and device, a server, a system, a medium and a product. The frequency band switching method comprises the steps of cleaning to-be-identified data based on a scene adaptive random forest model to determine features of each scene; performing standardization processing on the to-be-identified data through a scene adaptive principal component analysis algorithm to determine weights of features in each scene; determining the scene corresponding to the to-be-identified data according to the weighted distance of the features between the to-be-identified data and the sample data of each scene; and performing frequency band switching based on a scene corresponding to the to-be-identified data. According to the technical scheme, the features and weights in the scene are determined through the adaptive random forest model improved for the scene and the principal component analysis algorithm, the scene to be recognized is determined by calculating the weighting distance, the scene where the terminal is located can be accurately recognized in real time, the corresponding frequency band is automatically switched, and the real-time performance and reliability of frequency band switching are improved.

The application relates to the technical field of agricultural automation control, in particular to a seedling tray falling speed self-adaptive adjusting method based on a laser radar, which comprises the following steps: obtaining point cloud data obtained by scanning a seedling tray bottom surface by the laser radar; wherein the point cloud data comprises three-dimensional coordinates, distance values and echo intensities of a plurality of measuring points; performing confidence degree screening on the measuring points based on the point cloud data to obtain an effective measuring point set; according to the echo intensities of the measuring points in the effective measuring point set, calibrating single-point confidence degree weights of the measuring points, and performing confidence degree weighting on the distance values of the measuring points to obtain confidence degree weighted distance values of the measuring points. The application can automatically distinguish three regions of a convex edge, a concave groove and a flat bottom surface through partition identification of the seedling tray bottom surface, and can perform self-checking and error correction on the partition result, so that the partition result is closer to the actual structure of the bottom surface, lays a foundation for subsequent differential processing of the regions, and improves the accuracy of speed adjustment.

The application discloses a method and device for optimizing a transformation matrix in point cloud registration, a storage medium and equipment, and belongs to the technical field of image processing. A reference point cloud, a target point cloud and an initial transformation matrix are obtained. The target point cloud is transformed by using the initial transformation matrix to obtain a transformed point cloud. A matching point pair is searched for in the reference point cloud and the transformed point cloud by using a normal-weighted ICP algorithm. A weighted distance is calculated according to the normal correlation and the spatial distance of the matching point pair. An error weight of the matching point pair is calculated according to the weighted distance. Whether the matching point pair is abnormal is judged according to the normal correlation to obtain a detection result. A residual sum of squares is calculated for each group of matching point pairs according to the error weight and the detection result. The transformation matrix is updated according to the residual sum of squares until the transformation matrix satisfies a preset condition to obtain an optimal transformation matrix. The application can improve the matching accuracy and convergence. The reliability of the matching result is improved. It is ensured that the global optimal solution can be converged to, and the optimal transformation matrix is obtained.

The invention discloses an agricultural waste recycling layout simulation method and system based on big data analysis, and belongs to the field of agricultural waste treatment, and the method comprises the steps: according to dynamic distribution characteristics, employing a geographic information system to analyze topographic relief and road conditions, constructing a spatial constraint model, and determining the feasibility range of a transportation path; if the topographic relief exceeds a preset threshold value, adjusting the transportation path through a weighted distance algorithm to obtain an optimized path cost set; dividing a service radius range by adopting a K-means clustering algorithm through the candidate facility site selection point set, and determining a service coverage area of each facility; according to the service coverage area and the dynamic distribution characteristics, calculating the waste treatment capacity demand of each area to obtain a facility scale configuration scheme; and if the processing capability demand exceeds a preset threshold, adjusting the site selection point set and the coverage area through an iterative optimization algorithm to obtain a final resource facility layout scheme. According to the invention, reasonable agricultural waste recovery layout can be carried out.

The invention discloses an industrial big data outlier cleaning method based on cooperation of multiple KNN rules, and belongs to the technical field of industrial big data preprocessing. The core of the method is that a double-layer detection framework is adopted to consider both efficiency and precision: firstly, the relative difference degree between samples is calculated based on variable coefficient weighted distance measurement, a suspicious sample set with abnormal density is quickly screened out, and the calculation scale is remarkably reduced; and secondly, for a suspicious sample set, innovatively fusing and transmitting three types of rules of K neighbor distance, reachable distance and mutual neighbor number, and constructing a mixed factor index to carry out accurate identification. The method effectively solves the problems that a traditional method is high in calculation complexity when coping with massive high-dimensional industrial data and unbalanced in local and global benefit group point detection capacity, has the advantages of being high in precision, high in efficiency and high in universality, and is suitable for improving the data quality of an industrial monitoring system.

The invention belongs to the technical field of data processing, and particularly relates to an online analysis instrument nonlinear error correction method based on manifold learning, and the method comprises the steps: obtaining a high-dimensional feature vector set containing original physical quantity readings and auxiliary environment parameters; and calculating a signal transient response entropy and an environmental coupling stress index of the data of each dimension, and further deducing a manifold tangent space distortion rate representing the bending degree of a data structure. On this basis, distortion weighted distance measurement is constructed to replace a traditional Euclidean distance, high-dimensional features are mapped to a low-dimensional space by using an improved local linear embedding algorithm, and finally an error prediction model is established through a least square support vector machine. According to the method, physical perception measurement is introduced, so that the problems of data manifold curling and Euclidean distance failure caused by sudden change of the environment are reduced, neighborhood selection errors are avoided, and the measurement precision and stability of the instrument in the multi-physics coupling environment are improved.

The application discloses an automatic viewpoint planning method and device for motor end cover screw hole group inner wall detection. In view of the problem of missing depth values in the center area of the screw hole, multi-ray back projection and CAD surface intersection are adopted to realize stable three-dimensional hole positioning. Based on the geometric parameters of the screw hole, a combination of front view points and inclined view points is automatically generated, and the combination is converted into an executable joint angle configuration through inverse kinematics solving and collision detection. The multi-screw hole viewpoint access problem is modeled as a clustering traveling salesman problem, and a genetic algorithm combined with dynamic programming is used for hierarchical path optimization, taking the joint space weighted distance as the cost, so that the overall detection efficiency of the mechanical arm is improved.

The invention relates to the technical field of teaching data processing, in particular to a teaching data management system and method based on cloud service. The method comprises the following steps: preprocessing learning behavior data of each student to obtain a data point corresponding to each student; obtaining a weighted distance between every two data points; clustering is carried out to obtain different clusters and an average contour coefficient characteristic value of each cluster; obtaining the data point distribution closeness of the cluster according to the distribution characteristics of the data points in the cluster; obtaining an inter-cluster separation degree of the clustering; obtaining a clustering result evaluation index by using the average contour coefficient characteristic value, the data point distribution compactness and the inter-cluster separation degree of each cluster after the clustering; clustering the data points by using different K values and obtaining a clustering result evaluation index of each clustering; and taking the clustering result with the maximum clustering result evaluation index as a basis for dividing student groups. According to the invention, the student group can be accurately divided according to the learning behavior data of the students.

The embodiment of the invention provides a risk recovery processing method for a communication network, a storage medium and an electronic device, and the method comprises the steps: carrying out the multi-dimensional sampling of network features of the communication network, and obtaining a plurality of first sampling points, obtaining a plurality of distance values between a single first sampling point and a single sample point in the sample point set according to a plurality of distance algorithms, and carrying out weighted summation on the plurality of distance values to obtain a first weighted distance value; determining a first shortest weighted distance value according to the first weighted distance values of the single first sampling point and all the sample points, and determining a load level or fault level label of the communication network of the single first sampling point according to the first shortest weighted distance value; and carrying out risk recovery processing on the communication network according to the communication network load level or fault level label. According to the embodiment of the invention, the problem of poor risk recovery processing effect caused by low accuracy of a calculation result due to single-dimensional sampling and distance calculation by using a distance algorithm is solved.

This invention discloses a mapping-guided fuzzing method for loop code, applied in the field of software testing. The method includes: continuously mutating test cases, inputting them into a test case queue, sending them to the program under test (CUPT) with instrumented loop code structures, and outputting a coverage statistics table when the fuzzing termination condition is met; training a deep learning model and back-calculating the weighted distance sum to construct and output a mapping relationship between the byte sequences of test cases and the coverage of the loop code structure; based on the mapping relationship, assigning mutation probabilities to the corresponding byte sequences in the test cases corresponding to the loop code structure coverage and performing mutations to generate child test cases, which are then used as input to the CUPT for fuzzing until the fuzzing termination condition is met, and outputting a fuzzing report. This method mutates the byte sequences of test cases based on mapping relationships, enabling the targeted generation of test cases that improve the coverage of loop code results.

The application discloses a new node position prediction method based on hybrid particle swarm optimization and grey wolf optimization, comprising the following steps: determining a room or an area where a target wireless sensor node is located; measuring received signal strength at different positions away from the target wireless sensor node; filtering the signal strength obtained by using a Gaussian filter; fitting the obtained signal strength with distance, constructing a signal strength-distance relationship model and evaluating the model to select a model with the best fitting degree; dividing the room or the area to be positioned by using a dichotomy method, determining and narrowing a reference node selection area; limiting the position of a reference node deployment; selecting a reference node with strong and stable signal strength by using a weighted distance strategy; and positioning and optimizing the position of the target wireless sensor node by using a hybrid PSOGWO algorithm based on the obtained data. The application improves the positioning accuracy of the target node and also improves the positioning efficiency.

The present application relates to the technical field of automobile parts detection, in particular to a kind of automobile air conditioner compressor valve group air tightness detection method, comprising: acquisition is detected valve group multiple detection cycle in original air tightness detection data set collected by intelligent sensing system, data set includes pressure decay sequence, leakage acoustic wave sequence and mounting seat vibration sequence.Pressure decay sequence extracts pressure leakage feature vector, and acoustic fingerprint feature vector is obtained by analyzing leakage acoustic wave sequence.Improved nearest neighbor classification algorithm with feature dimension weighted distance measurement is used to determine sample similarity, two types of feature vectors are classified jointly, and leakage mode classification result is output.Air tightness detection conclusion is generated in combination with classification result and vibration sequence.The method can accurately determine the leakage condition of the valve group and locate the specific position of the leaking valve piece, improve the detection dimension and fault positioning accuracy, and adapt to the fine air tightness detection application of the valve group.

The application discloses an image segmentation method and system of a plurality of interaction fuzzy clustering validity index, comprising: 1, using the improved fuzzy C-means clustering algorithm based on kernel to classify the pixel points of the given image; 2, establishing a target function, checking whether the termination condition is met or the maximum iteration number is reached; 3, initializing and updating the membership matrix and the clustering center; 4, calculating the intra-class compactness value through the ratio of the fuzzy cardinality and the fuzzy weighted distance to the sample variance and the disturbance factor, and calculating the inter-class separation value from four aspects of separation relation, using the product of two factors and the iteration number to obtain the index value; 5, comparing the validity index of all classes, selecting the clustering number corresponding to the maximum validity index and the corresponding membership matrix to perform image segmentation. The application can effectively segment the image, cluster the pixel points and obtain effective image clustering results, and is suitable for unbalanced, complex, overlapped and noisy pixel sets.