A spatio-temporal clustering method using real-time motion analysis on event-based 3D vision

We propose a real-time method for counting pedestrians and bicyclists by classifying bulks of asynchronous events generated upon scene activities by an event-based 3D dynamic vision system. The inherent detection of moving objects offered by the 3D dynamic vision system comprising a pair of dynamic vision sensors allows event-based stereo vision in real-time and a 3D representation of moving objects. A clustering method exploits the sparse spatio-temporal representation of sensor's events for real-time detection and separation between moving objects. The method has been demonstrated for clustering the events and classification of pedestrian and cyclists moving across the sensor field of view based on their dimensions and passage duration. Tests on real scenarios with more than 100 cyclists and pedestrians yield a classification performance above 92%.

Improved partitioning technique for density cube-based spatio-temporal clustering method

Abstract. Considering the critical role of trajectory data in Big Data era for dynamic geographical processes, human behaviour analysis and meteorological prediction, trajectory clustering has attracted growing attention. Many literatures have discussed the spatiot emporal clustering method of simple trajectories (i.e., has no branches, e.g. vehicle trajectories), yet there are few researches for clustering complex trajectories (i.e., has at least one split and/or merger and/or split -merger branch, e.g. ocean eddy trajectories, rainstorm trajectories). For addressing this issue, we propose a Process-Oriented Spatiotemporal Clustering Method (POSCM) for clustering complex trajectory data. The POSCM includes three parts: the first uses the semantic of process-sequence-state to represent the complex trajectories; the second proposes a Hierarchical Similarity Measurement Method (HSMM) to get the similarity between any two complex trajectories; in the last step, the complex trajectories clustering pattern is extracted through density-based clustering algorithm. Experiments on simulated trajectories are used to evaluate the POSCM and demonstrate the advantage by comparing against that of the VF2 algorithm. The POSCM is applied to the sea surface temperature abnormal variations trajectories from January 1950 to December 2017 in the Pacific Ocean. As shown in this case study, some new mined spatiotemporal patterns can provide new references for understanding the behaviours of marine abnormal variations under the background of the global change.

the traditional anomaly detecting methods for power equipment do not consider the spatial information of the state data. This paper proposed a method for anomaly detection of state data of power equipment based on spatiotemporal clustering method, which considers historical big data and visualizes the structures revealed within data. Using a sliding window, the time series are divided into a number of subsequences. The available spatiotemporal structure within each time window is discovered using the FCM method. In the sequel, an anomaly score is assigned to each cluster. By using a fuzzy relation formed between revealed structures, a propagation of anomalies occurring in consecutive time intervals is visualized. At last, the effectiveness of the method is verified by an icing example.

The aim of this study was to quantitatively investigate the impact of stereoscopic three-dimensional (3D) vision on students' performance when compared with that of two-dimensional (2D) vision in a 3D virtual reality (VR) simulator. Twenty-four dental students (second- and fourth-year BDS) were assigned to perform three operative tasks under 3D and 2D viewing conditions on a Virteasy (HRV) simulator. Groups were crossed over and all students performed the same tasks under the alternate viewing conditions. The performance was evaluated by (1) accuracy, (2) outside target area removal, and (3) tooth cutting time, automatically using the generated feedback. Twenty-one participants completed all sessions. The results revealed a statistically significant effect of 3D vision over 2D vision on students' performance in terms of accuracy (p=0.035). Stereoscopic 3D vision showed significant effect on outside target area removal in the first task (p=0.035). Tooth cutting time was the same under both conditions (p=0.766). The findings revealed improvement in accuracy score and reduction in outside target area removal over the course of the experiment under both conditions. Comparing the difference in 3D effect in the early and advanced learning groups revealed no significant difference among the groups (p>0.05). Utilizing stereoscopic 3D vision in the training session improved students' perception of depth which led to more accurate tooth cutting within the target area, and less outside target area removal. However, 3D shows a limited impact on task completion time.

Applying Spatio-temporal Scan Statistics and Spatial Autocorrelation Statistics to identify Covid-19 clusters in the world - A Vaccination Strategy?

Density Peak Clustering (DPC) is a widely used clustering method that automatically identifies cluster centers based on local density and distance metrics. However, DPC tends to converge to local optimal solutions when processing datasets with uneven density distributions. This paper proposes DPC-DIST, an improved algorithm that addresses this limitation through a geometric distribution coefficient δ . By incorporating spatial distribution characteristics into density calculations, DPC-DIST effectively prevents convergence to suboptimal solutions. Comprehensive experiments on six benchmark datasets (Wine, Iris, Seeds, Blobs, R15, R3) demonstrate that DPC-DIST consistently outperforms the original DPC, K-means++, and spectral clustering algorithms. Significant improvements were observed in high-dimensional data processing, with 9.3 % increase in Silhouette Coefficient and 47.7 % improvement in Calinski-Harabasz Score on the Wine dataset. The algorithm shows particular advantages in facility location optimization and logistics planning applications, demonstrating its practical value for real-world scenarios.

Data mining is a collection of methods used to extract useful information from large data bases. Cluster Analysis refers to the grouping of a set of data points into clusters. Most widely used partitioning methods are K-means and Fuzzy c-means (FCM) algorithms. However, they suffer from the difficulties such as random selection of initial centre values and handling outlier data points. Most of the existing clustering methods use the Euclidean distance metric. The modified fuzzy c-means algorithm (MFCM) is efficient in handling outlier data points. In this paper, a new hybrid algorithm is proposed to solve the limitations of the traditional clustering methods. The hybrid K-MFCM algorithm is tested on four real world bench mark data sets from UCI machine learning repository with various distance metrics including Euclidean, City Block and Chessboard. The cluster centroid values of hybrid algorithm are calculated for various data sets. The experimental results show that the hybrid algorithm gives good results in terms of objective function value and better fuzzy cluster validity results for chessboard distance metric than other distance metrics.

The timely analysis of deformation monitoring data and reasonable diagnosis of the structural health are key tasks in dam health monitoring studies. This article presents a spatio-temporal clustering and health diagnosis method for super-high concrete arch dams that uses deformation monitoring data obtained from plumb meters. The spatio-temporal expression of the deformation monitoring data is proposed first by upgrading a punctuated time series to a curved panel time series, including cross-sectional, dam axial, and temporal changing directions. Second, a comprehensive similarity indicator on three aspects, namely, the absolute distance, incremental distance, and growth rate distance, is constructed after a deep discussion on deformation similarity characteristics both temporally and spatially. Next, the temporal clustering method is proposed by keeping the key features, namely, extreme points and turning points, while eliminating extraneous details, namely, noise points. Finally, the optimal spatio-temporal clustering of dam deformation is achieved by designing a multi-scale fuzzy C-means method of data mining and its iterative algorithm. The proposed method is applied to the Jinping-I hydraulic structure, which is the highest concrete arch dam in the world. The clustering results is quite sensitive in different weight coefficients of the comprehensive similarity indicator and clustering numbers of fuzzy C-means method. The dam deformation behaviors on high-water-level, water-falling, and low-water-level periods are analyzed and diagnosed. The advanced version of proposed methods is verified by comparative analysis on dam health diagnosis results obtained from ordinary deformation distribution figures and the spatio-temporal clustering figures. The proposed method will facilitate the recognition of abnormal deformation areas and associated safety diagnoses.

In an effort to achieve improved classifier accuracy, extensive research has been conducted in classifier ensembles. Very recently, cluster ensembles have emerged. It is well known that off-the-shelf clustering methods may discover different structures in a given set of data. This is because each clustering algorithm has its own bias resulting from the optimization of different criteria. Furthermore, there is no ground truth against which the clustering result can be validated. Thus, no cross-validation technique can be carried out to tune input parameters involved in the clustering process. As a consequence, the user is not equipped with any guidelines for choosing the proper clustering method for a given dataset. Cluster ensembles offer a solution to challenges inherent to clustering arising from its ill-posed nature. Cluster ensembles can provide more robust and stable solutions by leveraging the consensus across multiple clustering results, while averaging out emergent spurious structures that arise due to the various biases to which each participating algorithm is tuned. In this chapter, we discuss the problem of combining multiple weighted clusters, discovered by a locally adaptive algorithm (Domeniconi, Papadopoulos, Gunopulos, & Ma, 2004) which detects clusters in different subspaces of the input space. We believe that our approach is the first attempt to design a cluster ensemble for subspace clustering (Al-Razgan & Domeniconi, 2006). Recently, several subspace clustering methods have been proposed (Parsons, Haque, & Liu, 2004). They all attempt to dodge the curse of dimensionality which affects any algorithm in high dimensional spaces. In high dimensional spaces, it is highly likely that, for any given pair of points within the same cluster, there exist at least a few dimensions on which the points are far apart from each other. As a consequence, distance functions that equally use all input features may not be effective. Furthermore, several clusters may exist in different subspaces comprised of different combinations of features. In many real-world problems, some points are correlated with respect to a given set of dimensions, while others are correlated with respect to different dimensions. Each dimension could be relevant to at least one of the clusters. Global dimensionality reduction techniques are unable to capture local correlations of data. Thus, a proper feature selection procedure should operate locally in input space. Local feature selection allows one to embed different distance measures in different regions of the input space; such distance metrics reflect local correlations of data. In (Domeniconi, Papadopoulos, Gunopulos, & Ma, 2004) we proposed a soft feature selection procedure (called LAC) that assigns weights to features according to the local correlations of data along each dimension. Dimensions along which data are loosely correlated receive a small weight, which has the effect of elongating distances along that dimension. Features along which data are strongly correlated receive a large weight, which has the effect of constricting distances along that dimension. Thus the learned weights perform a directional local reshaping of distances which allows a better separation of clusters, and therefore the discovery of different patterns in different subspaces of the original input space.

A robust clustering algorithm based on the identification of core points and KNN kernel density estimation

Because of the significant industrial demands towards quality and safety of system, reliability prediction with historical failures data has generated broad interest. Particularly, for system‐oriented failures time‐series data, although the hybridization strategy has been exploited to separately predict the feature components extracted from the original data and achieved noteworthy performance, a convictive method for effectively extracting these feature components has not been explored. In this paper, we introduce weighted shape‐based time‐series clustering to improve the hybrid modeling and prediction, in which a novel distance metric named as w_SBD (ie, weighted shape‐based distance) is devised by fully considering the shapes of time series and the characteristics of failures prediction. Moreover, we further develop a flexible framework to extract and validate the feature components (named as FF_EVFC). In the framework, besides w_SBD, 3 kinds of validations for the extracted feature components are also involved. To demonstrate the robustness of w_SBD and FF_EVFC, we perform extensive experimental evaluations with different clustering and prediction methods. The results show a competitive performance of w_SBD against other common distance metrics and verify the effectiveness of FF_EVFC on the improvement of failures prediction.

Procedure-Specific Simulations (PSS) are 3D realistic simulations that provide a platform to practice complete surgical procedures in a virtual-reality environment. While PSS have the potential to improve surgeons' proficiency, there are no existing standards or guidelines for PSS development in a structured manner. We employ a unique platform inspired by game design to develop virtual reality simulations in three dimensions of urethrovesical anastomosis during radical prostatectomy. 3D visualization is supported by a stereo vision, providing a fully realistic view of the simulation. The software can be executed for any robotic surgery platform. Specifically, we tested the simulation under windows environment on the RobotiX Mentor. Using urethrovesical anastomosis during radical prostatectomy simulation as a representative example, we present a task-based methodological approach to PSS training. The methodology provides tasks in increasing levels of difficulty from a novice level of basic anatomy identification, to an expert level that permits testing new surgical approaches. The modular methodology presented here can be easily extended to support more complex tasks. We foresee this methodology as a tool used to integrate PSS as a complementary training process for surgical procedures.

Machine vision systems become to play an important role in robot based factory automation, because they enable robots to recognize the environment around the workspace like human visual systems and make the manufacturing process flexible. So a large number of 2-dimensional (2D) vision systems have been successfully applied for industrial robots, but there are increasing needs for 3-dimensional (3D) vision systems providing accurate 3D position of work-pieces. For the 3D vision system, the stereo vision based position measurement systems have been applied in various fields. However the stereo vision systems which have been applied were specific for their particular tasks only. This paper introduces a general-purpose stereo vision system applicable to wide range of tasks with industrial robots.

This paper describes a novel clustering algorithm inspired by the humoral-mediated response triggered by the adaptive immune system. The key humoral-mediated features of the algorithm include B-cell antibodies produced through plasma cells and memory B-cell antibodies. Affinity threshold, network threshold, death threshold and negative clonal selection threshold are also used to derive intra-cluster and inter-cluster distance metrics that result in the merging of similar clusters and removal/identification of less significant clusters (outlier detection). The performance of the clustering algorithm is tested on both synthetic and real world datasets and compared with other clustering methods.