Mean Shift Clustering Algorithm Research Articles

Research on Image Segmentation based on Clustering Algorithm

Hierarchical clustering (HC) algorithm can obtain good clustering results, but it needs large storage and computational complexity for large image processing. Anew color image segmentation algorithm based on mean shift and hierarchical clustering algorithm named MSHC is presented in this paper. MSHC algorithm preprocesses an input image by MS algorithm to form segmented regions that preserve the desirable discontinuity characteristics of image. The number of segmented regions, instead of the number of image pixels, is considered as the input data scale of HC algorithm. The proximity between each cluster is calculated to form the proximity matrix, and then ward algorithm is employed to obtain the final segmentation results. MSHC algorithm is employed on color image and medical image segmentation.

A Novel Method to Predict Genomic Islands Based on Mean Shift Clustering Algorithm.

Genomic Islands (GIs) are regions of bacterial genomes that are acquired from other organisms by the phenomenon of horizontal transfer. These regions are often responsible for many important acquired adaptations of the bacteria, with great impact on their evolution and behavior. Nevertheless, these adaptations are usually associated with pathogenicity, antibiotic resistance, degradation and metabolism. Identification of such regions is of medical and industrial interest. For this reason, different approaches for genomic islands prediction have been proposed. However, none of them are capable of predicting precisely the complete repertory of GIs in a genome. The difficulties arise due to the changes in performance of different algorithms in the face of the variety of nucleotide distribution in different species. In this paper, we present a novel method to predict GIs that is built upon mean shift clustering algorithm. It does not require any information regarding the number of clusters, and the bandwidth parameter is automatically calculated based on a heuristic approach. The method was implemented in a new user-friendly tool named MSGIP—Mean Shift Genomic Island Predictor. Genomes of bacteria with GIs discussed in other papers were used to evaluate the proposed method. The application of this tool revealed the same GIs predicted by other methods and also different novel unpredicted islands. A detailed investigation of the different features related to typical GI elements inserted in these new regions confirmed its effectiveness. Stand-alone and user-friendly versions for this new methodology are available at http://msgip.integrativebioinformatics.me.

Classification of vertebral column disorders and lumbar discs disease using attribute weighting algorithm with mean shift clustering

In this article, a new data pre-processing method has been suggested to detect and classify vertebral column disorders and lumbar disc diseases with a high accuracy level. The suggested pre-processing method is called the Mean Shift Clustering-Based Attribute Weighting (MSCBAW) and is based primarily on mean shift clustering algorithm finding the number of the sets automatically. In this study, we have used two different datasets including lumbar disc diseases (with two classes-our database) and vertebral column disorders datasets (with two or three classes) taken from UCI (University of California at Irvine) machine learning database to test the proposed approach. The MSCBAW method is working as follows: first of all, the centres of the sets automatically for each characteristics in dataset by using the mean shift clustering algorithm are computed. And then, the mean values of each property in dataset are calculated. The weighted datasets by multiplying these mean values by each property value in the dataset that have been obtained by dividing the above mentioned mean values by the centres of the sets belonging to the relevant property are achieved. After the data weighting stage, three different classification algorithms that included the k-NN (k-Nearest Neighbour), RBF–NN (Radial Basis Function–Neural Network) and SVM (Support Vector Machine) classifying algorithms have been used to classify the datasets. In the classification of vertebral column disorders dataset with two classes (normal or abnormal), while the obtained classification accuracies and kappa values were 78.70%±0.455 (the classification accuracy±standard deviation), 81.93%±0.899, and 80.32%±0.56 using SVM, k-NN (for k=1), and RBF–NN classifiers, respectively, the combinations of MSCBAW and SVM, k-NN (for k=1), and RBF–NN classifiers were obtained 99.03%±0.977, 99.67%±0.992, and 99.35%±0.9852, respectively. In the classification of second dataset named vertebral column disorders dataset with three classes (Normal, Disk Hernia, and Spondylolisthesis), while the obtained classification accuracies and kappa values were 74.51%±0.581, 78.70%±0.659, and 83.22%±0.728 using SVM, k-NN (for k=1), and RBF–NN classifiers, respectively, the combinations of MSCBAW and SVM, k-NN (for k=1), and RBF–NN classifiers were obtained 99.35%±0.989, 96.77%±0.948, and 99.67%±0.994, respectively. As for the lumbar disc dataset, while the obtained classification accuracies and kappa values were 94.54%±0.974, 94.54%±0.877, and 93.45%±0.856 using SVM, k-NN (for k=1), and RBF–NN classifiers, respectively, the combinations of MSCBAW and SVM, k-NN (for k=1), and RBF–NN classifiers were obtained 100%±1.00, 99.63%±0.991, and 99.63%±0.991, respectively. The best hybrid models in the classification of vertebral column disorders dataset with two classes, vertebral column disorders dataset with three classes, and lumbar disc dataset were the combination of MSCBAW and k-NN classifier, the combination of MSCBAW and RBF–NN classifier, and the combination of MSCBAW and SVM classifier, respectively.

Chaotic features for dynamic textures recognition

This paper presents a novel framework for dynamic textures (DTs) modeling and recognition, investigating the use of chaotic features. We propose to extract chaotic features from each pixel intensity series in a video. The chaotic features in each pixel intensity series are concatenated to a feature vector, chaotic feature vector. Then, a video is modeled as a feature vector matrix. Next, two approaches of DTs recognition are investigated. A bag of words approach is used to represent each video as a histogram of chaotic feature vector. The recognition is carried out by 1-nearest neighbor classifier. We also investigate the use of earth mover's distance (EMD) method. Mean shift clustering algorithm is employed to cluster each feature vector matrix. EMD method is used to compare the similarity between two videos. The output of EMD matrix whose entry is the matching score can be used to DTs recognition. We have tested our approach on four datasets and obtained encouraging results which demonstrate the feasibility and validity of our proposed methods.

A segmentation system based on clustering method for pediatric DTI images

ABSTRACTA system is presented for the segmentation of white matter fiber tracts in pediatric diffusion tensor magnetic resonance imaging (DTI) images. DTI is an in vivo method to delineate the connectivity of white matter fiber tracts in human brain by fiber tractography. Fiber tractography is a promising method to visualize the whole bundles of fiber tracts. Fiber tractography is unable to provide a quantitative analysis and description of specific white matter fiber tracts. Obviously, segmenting and clustering the fiber tracts into anatomical bundles play an important role in fiber tracts analysis. Traditional manual segmentation method requires neuroanatomical expertise and significant time. It can not be a standardized and widely used method for segmentation of complicated fiber tracts in pediatric DTI images. Hence, an image segmentation system with an adaptive mean shift (AMS) clustering method is proposed to cluster fiber tracts into bundles automatically in this article. In the image segmentation system, fiber similarity measure based on Euclidean distance is used in the clustering method. Since the increase of children's mental illness in recent years, segmentation of pediatric DTI images by clustering methods is focused in our research. The effectiveness and robustness of adaptive mean shift clustering algorithm for segmentation of fiber tracts are also evaluated by error analysis experiments. In addition, the experiment results show that adaptive mean shift method used in our system is more efficient and effective than K‐means and Fuzzy C‐means (FCM) clustering methods for the segmentation of fiber tracts in real pediatric DTI images. © 2015 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 25, 102–113, 2015

Sensing-Transmission Edifice Using Bayesian Nonparametric Traffic Clustering in Cognitive Radio Networks

In cognitive radio networks, the main objective of spectrum sensing is to exploit spectrum holes left by the primary users (PUs). Different PUs' traffic patterns might provide different opportunities for second user (SU) spectrum access. In this paper, we identify the PUs' traffic patterns and then maximize SU transmission accordingly. First a theoretical framework is developed to cluster PU traffic patterns based on a Bayesian nonparametric inference model, in which the number of traffic types is unknown. Second, in order to exploit the spectrum holes, we study a sensing-transmission structure to optimize the SU transmission strategy. Specifically, we exploit the short and long transmission opportunities based on the PU traffic pattern and channel idle time distribution. Finally, we propose a threshold-based sensing-transmission method that optimizes the SU utility, while protecting PU transmissions. Both sensing and transmission errors are considered for perfect sensing with/without acknowledgement-based transmission and imperfect sensing, respectively. From the simulation results, we show that the proposed technique outperforms the nonparametric mean shift clustering algorithm. Furthermore, we utilize these clustering results to optimize the SU's transmission strategy with perfect and imperfect sensing. We compare our proposed technique with the probabilistic sensing-transmission structure and show the performance gain in terms of throughput.

Chaotic Features for Traffic Video Classification

This paper proposes a novel framework for traffic video classification based on chaotic features. First, each pixel intensity series in the video is modeled as a time series. Second, the chaos theory is employed to generate chaotic features. Each video is then represented by a feature vector matrix. Third, the mean shift clustering algorithm is used to cluster the feature vectors. Finally, the earth mover’s distance (EMD) is employed to obtain a distance matrix by comparing the similarity based on the segmentation results. The distance matrix is transformed into a matching matrix, which is evaluated in the classification task. Experimental results show good traffic video classification performance, with robustness to environmental conditions, such as occlusions and variable lighting.

Automatic Myonuclear Detection in Isolated Single Muscle Fibers Using Robust Ellipse Fitting and Sparse Representation.

Accurate and robust detection of myonuclei in isolated single muscle fibers is required to calculate myonuclear domain size. However, this task is challenging because: 1) shape and size variations of the nuclei, 2) overlapping nuclear clumps, and 3) multiple z-stack images with out-of-focus regions. In this paper, we have proposed a novel automatic detection algorithm to robustly quantify myonuclei in isolated single skeletal muscle fibers. The original z-stack images are first converted into one all-in-focus image using multi-focus image fusion. A sufficient number of ellipse fitting hypotheses are then generated from the myonuclei contour segments using heteroscedastic errors-in-variables (HEIV) regression. A set of representative training samples and a set of discriminative features are selected by a two-stage sparse model. The selected samples with representative features are utilized to train a classifier to select the best candidates. A modified inner geodesic distance based mean-shift clustering algorithm is used to produce the final nuclei detection results. The proposed method was extensively tested using 42 sets of z-stack images containing over 1,500 myonuclei. The method demonstrates excellent results that are better than current state-of-the-art approaches.

Human Localization based on Distributed Laser Range Finders

Localizing humans inside a home or office environment is vital for various service robot applications. This paper presents a novel method to estimate human locations using distributed laser range finders. First, the Kalman filter is employed on the laser data to produce a statistically optimal scan estimate. Next, Scan points in the foreground are then grouped using mean shift clustering algorithm. An area-divided clustering mode is introduced to ensure enough clusters to represent the contour of human. Finally, the centers of humans are estimated from the obtained clusters using incomplete ellipse fitting. Experiments are conducted to prove the robustness and efficiency of the proposed method.

Automatic Segmentation of Urban Point Clouds Based on the Gaussian Map

AbstractA comprehensive method to segment urban point clouds based on the Gaussian map is presented. The normals of point clouds are firstly mapped to the Gaussian sphere and then partitioned into groups using a mean shift clustering algorithm. Next, a distance‐based clustering method is presented to tackle overlapping surfaces which avoids under‐segmentation. Based on the properties of the Gaussian map and the geometric information of the points, primitive shapes such as planes, cylinders, cones and spheres are recognised. Trees, cars, street lights and other objects are then segmented by using the distance‐based clustering method after removing the planes, cylinders, cones and spheres from the large urban scenes. Finally, a refinement process based on primitive shapes is implemented to improve the segmentation results which effectively avoids over‐segmentation. Experimental results demonstrate that the proposed method can be used as a robust way to segment urban point clouds based on the Gaussian map.

A line-based spectral clustering method for efficient planar structure extraction from LiDAR data

Abstract. Planar structures are essential components of the urban landscape and automated extraction planar structure from LiDAR data is a fundamental step in solving complex mapping tasks such as building recognition and urban modelling. This paper presents a new and effective method for planar structure extraction from airborne LiDAR data based on spectral clustering of straight line segments. The straight line segments are derived from LiDAR scan lines using an Iterative-End-Point-Fit simplification algorithm. Adjacency matrix is then formed based on pair-wise similarity of the extracted line segments, and a symmetric affine matrix is derived which is then decomposed into eigenspace. The planar structures are then detected by mean-shift clustering algorithm in eigenspace. The use of straight line segments facilitates the processing and significantly reduces the computational load. Spectral analysis of straight line segments in eigenspace makes the planar structures more prominent, resulting in a robust extraction of planar surfaces. Experiments are performed on the ISPRS benchmark LiDAR data over three test sites containing a variety of buildings with complex roof structures and varying sizes. The experimental results, which are quantitatively evaluated independently by the ISPRS benchmark test group, are presented. The results show that the proposed method achieves on average 80% of completeness with over 98% of correctness. Better performance is observed over larger size of buildings (>10m2) with over 92% of completeness and nearly 100% of correctness in all test areas, indicating the robustness and high reliability of the proposed algorithm.

Fusion of ALS Point Cloud and Optical Imagery for 3D Reconstruction of Building's Roof

Abstract. Three-dimensional building models are important in various applications such as disaster management and urban planning. In this paper a method based on fusion of LiDAR point cloud and aerial image data sources has been proposed. Firstly using 2D map, the point set relevant to each building separated from the overall LiDAR point cloud. In the next step, the mean shift clustering algorithm applied to the points of different buildings in the feature space. Finally the segmentation stage ended with the separation of parallel and coplanar segments. Then using the adjacency matrix, adjacent segments are intersected and inner vertices are determined. In the other space, the area of any building cropped in the image space and the mean shift algorithm applied to it. Then, the lines of roof’s outline edge extracted by the Hough transform algorithm and the points obtained from the intersection of these lines transformed to the ground space. Finally, by integration of structural points of intersected adjacent facets and the transformed points from image space, reconstruction performed. In order to evaluate the efficiency of proposed method, buildings with different shapes and different level of complexity selected and the results of the 3D model reconstruction evaluated. The results showed credible efficiency of method for different buildings.

Characterization of lunar soils through spectral features extraction in the NIR

Recently launched hyper-spectral instrumentation with ever-increasing data return capabilities deliver the remote-sensing data to characterize planetary soils with increased precision, thus generating the need to classify the returned data in an efficient way for further specialized analysis and detection of features of interest. This paper investigates how lunar near-infrared spectra generated by the SIR-2 on Chandrayaan-1 can be classified into distinctive groups of similar spectra with automated feature extraction algorithms. As common spectral parameters for the SIR-2 spectra, two absorption features near 1300nm and 2000 and their characteristics provide 10 variables which are used in two different unsupervised clustering methods, the mean-shift clustering algorithm and the recently developed graph cut-based clustering algorithm by Müller et al. (2012). The spectra used in this paper were taken on the lunar near side centering around the Imbrium region of the Moon. More than 100,000 spectra were analyzed.

Phase Classification by Mean Shift Clustering of Multispectral Materials Images

A mean-shift clustering (MSC) algorithm is introduced as a valuable alternative to perform materials phase classification from multispectral images. As opposed to other multivariate statistical techniques, such as factor analysis or principal component analysis (PCA), clustering techniques directly assign a class label to each pixel, so that their outputs are phase segmented images, i.e., there is no need for an additional segmentation algorithm. On the other hand, as compared to other clustering procedures and classification methods, such as segmentation by thresholding of multiple spectral components, MSC has the advantages of not requiring previous knowledge of the number of data clusters and not assuming any shape for these clusters, i.e., neither the number nor the composition of the phases must be previously known. This makes MSC a particularly useful tool for exploratory research, assisting phase identification of unknown samples. Visualization and interpretation of the results are also simplified, since the information content of the output image does not depend on the particular choice of the content of the color channels.We applied MSC to the analysis of two sets of X-ray maps acquired in scanning electron microscopes equipped with energy-dispersive detection systems. Our results indicate that MSC is capable of detecting additional phases, not clearly identified through PCA or multiple thresholding, with a very low empirical reject rate.

Support Vector Machines for Object Based Building Extraction in Suburban Area using Very High Resolution Satellite Images, a Case Study: Tetuan, Morocco

Many fields of artificial intelligence have been developed such as computational intelligence and machine learning involving neural networks, fuzzy systems, genetic algorithms, intelligent agents and Support Vector Machines (SVM). SVM is a machine learning methodology with great results in image classification. In this paper, we present the potential of SVMs to automatically extract buildings in suburban area using Very High Resolution Satellite (VHRS) images. To achieve this goal, we use object based approach: Segmentation before classification in order to create meaningful image objects using color features. In the first step, we form objects with the aid of mean shift clustering algorithm. Then, SVM classifier was used to extract buildings. The proposed method has been applied on a suburban area in Tetuan city (Morocco) and 83.76% of existing buildings have been extracted by only using color features. This result can be improved by adding other features (e.g., spectral, texture, morphology and context). DOI: http://dx.doi.org/10.11591/ij-ai.v2i1.1781

一种基于数据竞争的高分辨率图像的聚类分割算法

In recent years, robot technology has been rapidly developed and applied to every walk of life. With the promotion and popularization of the robot technology, requirement of using robot is becoming high and the demand of intelligent robots is particularly urgent. Machine vision is an important research direction in the field of intelligent robotics. In the robot vision system, the core problem is the targets extraction, and the image segmentation is the key technique of extracting targets accurately, rapidly and in real-time. Since the environment is complex and the targets are diverse, the amount of images data perceived by robots is large and the images are unpredictable. Thus, it is very important to extract and segment targets accurately. Aimed at the segmentation processing of high-resolution images, a novel clustering algorithm is proposed in this paper. According to the sum of the energy of data and their sizes, it recognizes the clustering representatives and members, and identifies the most probability members by the competition among data points. Then we apply the algorithm by combining the novel clustering algorithm into Mean Shift clustering algorithm in color image segmentation problem. The algorithm can quickly and efficiently achieve the targets segmentation of high-resolution images, and has good segmentation effect. The experiments show that the proposed approach has better clustering quality and is faster than the traditional clustering algorithm.

An adaptive mean shift clustering algorithm based on locality-sensitive hashing

The time complexity of the adaptive mean shift is related to the dimension of data and the number of iterations. The amount of computation will increase prohibitively with the increase of the data dimension. An approximate neighborhood queries method is presented for the computation of high dimensional data, in which, the locality-sensitive hashing (LSH) is used to reduce the computational complexity of the adaptive mean shift algorithm. The data-driven bandwidth selection for multivariate data is used in mean shift procedure, and an adaptive mean shift based on LSH with bandwidth estimation (LSH-PE-AMS) algorithm is proposed. Experimental results show that the proposed algorithm can reduce the complexity of the adaptive mean shift algorithm, and can produce a more accurate classification than the fixed bandwidth mean shift algorithm.

Deformable Bayesian Network: A Robust Framework for Underwater Sensor Fusion

The dynamic tree (DT) graphical model is a popular analytical tool for image segmentation and object classification tasks. A DT is a useful model in this context because its hierarchical property enables the user to examine information in multiple scales and its flexible structure can more easily fit complex region boundaries compared to rigid quadtree structures such as tree-structured Bayesian networks. This paper proposes a novel framework for data fusion called a deformable Bayesian network (DFBN) by using a DT model to fuse measurements from multiple sensing platforms into a nonredundant representation. The structural flexibility of the DFBN will be used to fuse common information across different sensor measurements. The appropriate structure update strategies for the DFBN and its parameters for the data fusion application are discussed. A real-world example application using sonar images collected from a survey mission is presented. The fusion results using the presented DFBN framework are shown to outperform state-of-the-art approaches such as the Gaussian mean shift and spectral clustering algorithms. The DFBN's complexity and scalability are discussed to address its potential for a larger data set.

Fast and effective integration of multiple overlapping range images

We present a simple and fast integration method, which can fuse registered multiple, overlapping range images effectively while preserving the detailed features. First, normal vectors of the nodes in each range image are estimated and the registered multiple overlapping range images are combined into one point set. Due to registration errors, the normal vectors of homogeneous nodes in the overlapping areas would differ from each other. Thus, the normal vectors of the nodes are corrected iteratively by weighted averaging their neighboring normal vectors. Second, the combined point set is partitioned into a number of small clusters using the k-means clustering algorithm based on user-defined space interval. The clusters subsequently are subdivided into smaller subclusters by using the standard normal vector deviation as a measure of cluster scatter. Finally, in the cluster, the local maxima model, obtained using the mean shift clustering algorithm, is employed to represent the cluster. The proposed method is fast because it does not need to detect iteratively the overlapping areas, which is usually time consuming. In addition, because of the noisy filtering property of the mean shift clustering algorithm, the novel method can produce a smooth and watertight point surface while preserving the detailed features.

Fast hierarchical animated object decomposition using approximately invariant signature

In this paper, we introduce a novel method to hierarchically decompose the animated 3d object efficiently by utilizing high-dimensional and multi-scale geometric information. The key idea is to treat the animated surface sequences as a whole and extract the near-rigid components from it. Our approach firstly detects a set of the multi-scale feature points on the animated object and computes approximately invariant signature vectors for these points. Then, exploiting both the geometric attributes and the local signature vector of each point (vertex) of the animated object, all the points (vertices) of the animated object can be clustered efficiently using a GPU-accelerated mean shift clustering algorithm. To refine the decomposition boundaries, the initially-generated boundaries of the animated object can be further improved by applying a boundary refinement technique based on Gaussian Mixture Models (GMMs). Furthermore, we propose a hierarchical decomposition technique using a topology merging strategy without introducing additional computations. Our animated object decomposition approach does not require the topological connectivity of the animated object, thus it can be applied for both triangle mesh and point-sampled geometry sequences. The experimental results demonstrate that our method achieves both good quality results and high performance for the decomposition of animated object.