基于参数自适应粒子群模糊聚类的图像分割研究

传统模糊聚类算法初始值过多依赖先验知识，易陷入局部最优，而采用粒子群模糊聚类虽有所改善，但粒子群也存在陷入早熟收敛的情况。针对这一问题，提出一种基于参数自适应的粒子群模糊聚类算法APSO-FCM。首先采用自适应参数控制策略对粒子群算法进行改进，然后使用提出算法优化模糊聚类初始数目，接着使用新的模糊C均值聚类方法进行图像分割。实验结果表明，提出算法划分的图像效果得到改善，在划分系数、划分熵等图像分割指标上有进一步提升。     

基于模糊C均值聚类与空间信息相结合的图像分割新算法

针对传统的模糊C均值聚类（FCM）图像分割方法未考虑图像的空间信息,对噪声十分敏感的问题,本文提出了一种结合空间信息的模糊C均值聚类分割新算法;该算法将图像的二维直方图引入传统的模糊C均值聚类算法中,并对隶属函数做了改进;依据平方误差和最小准则,来确定模糊分类矩阵及聚类中心;最后,依据最大隶属度原则,划分图像像素的类别归属,以改善传统的PCM算法的分割质量。实验结果表明,该算法显示了较好的分割效果和较强的抗噪性能。     

基于模糊聚类遗传算法的图像分割方法研究

图像分割是图像处理和图像分析的关键步骤,在图像工程中占据重要地位.模糊C均值聚类(FCM)算法是一种经典的模糊聚类分析方法,但其算法初始聚类原型是随机选取的,从而造成算法性能强烈地依赖聚类原型的初始化,将遗传算法强大的通用性应用于模糊聚类算法,对模糊聚类中心进行编码,然后依据FCM算法的目标函数建立适应度函数,选择适当的交叉率和变异率,最终实现基于模糊聚类遗传算法的图像分割.采用这种方法一方面能较好地解决模糊聚类对初始化敏感的问题,又能在一定程度上提高了分割速度.实验结果表明,该算法具有良好的分割效果.     

基于协方差描述子的彩色图像分割算法

为克服模糊C均值聚类图像分割对噪声较为敏感的缺陷,提出了一种结合协方差描述子的模糊C均值聚类算法。采用协方差描述子的滤波能力以改善传统模糊C均值聚类算法对噪声敏感的缺陷;提取超像素的协方差矩阵作为特征,降低图像识别的特征冗余。并做了仿真实验,对提出的算法与三个图像分割算法进行比较,结果表明该图像分割算法具有较好的噪声鲁棒性和分割准确率。     

基于模糊聚类的国画图像分割研究

国画图像分割是图像理解、计算机视觉等研究领域的重要内容。文章根据国画颜色有限性的特点以及视觉的颜色聚类特性,提出了一种基于模糊C均值聚类(FCM)的国画图像分割算法,实现对国画图像的分割。算法根据中国国画颜色以墨色为主,以墨破色的特点,通过选取国画图像每个像素点的灰度颜色特征,计算各像素点到各个聚类中心之间的欧式距离,并利用模糊C均值聚类的方法更新聚类中心,聚类结束后得到分类后的图像;再利用分类图像标识出分割图像,从而实现国画图像的分割。通过实验得出,算法有效地保证了各区域图像的完整,提高了分割的精确度。     

结合图像特征比的红外与可见光图像融合算法

针对现有图像融合算法得到的融合图像清晰度不高以及边缘模糊等问题,提出了一种结合图像特征比的红外与可见光的图像融合算法,通过小波变换将图像进行分解,并通过计算低频分量中空间频率比与能量比,来对低频分量中的有效信号进行保留,而对高频分量则是通过边缘检测算法计算出高频分量中的边缘信号比,来对高频分量的边缘信号进行保留;实验结果表,该算法能够得到相比其他算法更为清晰的融合图像,具有一定的实用价值。     

基于模糊推理的电气设备红外图像分割

使用模糊理论处理电气设备红外图像分割的不确定性，提出了一种基于模糊推理的电气设备红外图像分割算法。首先分别利用电气设备红外图像故障区域的像素灰度、像素点与图像质心的马氏距离以及图像膨胀操作定义了强度特征、全局故障可能性特征和局部灰度特征；然后根据特征的模糊语言值制定了27条模糊规则，设计了一种模糊推理红外图像分割算法；最后，从主观和客观评价指标上将算法与传统Otsu算法和FCM算法进行了对比。实验表明，该算法的分割精度和误分割率比其他两种算法都有一定的改善，同时该算法能够滤除图像中具有高亮度的干扰区域，对具有小亮度差和小面积故障区域的红外图像有较好的分割效果。     

基于学生t分布的鲁棒分层模糊算法及其在图像分割中的应用

著名的模糊C均值算法（FCM）一直被视为图像分割应用中一个强有力的工具.然而,由于FCM中距离函数选择问题使得其对图像噪声的鲁棒性不足.本文提出了一个新的分层模糊C均值算法,使得传统的模糊C均值算法对于图像噪声和离群点有更好的鲁棒性.在此基础上引入了一个更加灵活的函数,即将距离函数本身看作是一个子学生t分布函数.使分层模型具有更好的通用性和灵活性.本文提出的算法可以扩展到其他基于FCM模型的算法实现,以获得更优的鲁棒性.实验结果表明本文提出新的分层模糊C均值算法的鲁棒性确实有效.     

基于纹理复杂度的含噪模糊图像盲复原算法设计

传统的模糊图像盲复原算法的纹理复杂度高,降低了模糊图像的质量。为此,文中设计了基于纹理复杂度的含噪模糊图像盲复原算法。首先,分析含噪模糊图像的噪声特点,在保留原有图像信息的前提下计算垂直方向和水平方向的扩散系数,得到ADF在含噪模糊图像上的表达,完成含噪模糊图像的去噪预处理。通过建立含噪模糊图像的数学模型描述其退化过程,继而构建含噪模糊图像降质模型。最后,在纹理复杂度分析的基础上,完成奇异值分解检测和alpha通道的计算,通过合成操作实现含噪模糊图像的盲复原算法的设计。实验结果表明,相比于传统盲复原算法,在所提算法下,图像的纹理复杂度低,图像质量得以提升,且复原结果的误差较小,算法整体有效性较高。     

改进FCM算法在医学图像分割的方法研究

本文提出使用改进模糊C均值聚类(MFCM)算法和模糊可能性C均值聚类(FPCM)算法的图像分割方法并应用于医学图像分割过程中。MFCM算法是通过调整FCM算法的测量距离来批准标签像素受到其他图像像素和在切分中抑制噪声效果来约束,从而使得成员变量没有最大约束值。基于真实医学图像的实验表明了MFCM算法和FPCM算法在医学图像中进行分割的实际效果,具体是通过对FCM、MFCM、FPCM进行精度对比来验证算法有效性。     

一种粒子群算法与FCSS相结合的模糊球壳聚类算法

模糊球壳聚类算法（FCSS）广泛应用于模式识别与机器学习领域。由于其采用传统的基于梯度法和交替寻优策略求解模型,对初始值比较敏感,往往只能得到模型的局部极值点,从而影响聚类的效果,甚至使所采用的方法失效。本文将现代全局优化方法之一的粒子群优化算法与模糊球壳聚类算法（FCSS）相结合,利用粒子群算法良好的全局收敛能力来改善传统聚类算法易于陷入局部极值的缺陷,从而得到一种新的球壳聚类算法（PSO-FCSS）,数值实验表明,新方法对球壳形数据有令人满意的聚类效果。     

任意分层手征介质球域中并矢格林函数理论和应用

本文首先应用散射迭加法，给出了含源任意分层手征介质球域中的并矢格林函数，以手征介质球壳为例，分析了位于球中心处点偶极天线的辐射特性，以及归一化辐射阻抗随球壳厚度的变化规律，在平板近似条件下，研究了手征导纳对手征介质球壳传输特性的影响，并且讨论了手征介质球壳内、外表面与空气之间阻抗匹配时，点偶极天线远区辐射场的极化特性。结果表明，通过改变球壳的归一化厚度，可以调节辐射场的极化状态。     

A rigorous coupled-wave analysis and crossed-diffraction gratinganalysis of radiation and scattering from three-dimensionalinhomogeneous objects

A rigorous coupled-wave analysis (RCWA) algorithm useful for studying the radiation and scattering from three-dimensional (3-D) inhomogeneous objects is developed. A spherical electromagnetic (EM) state variable (SV) form of Maxwell's equations is presented. The RCWA method is used to numerically study EM dipole radiation within and exterior to an inhomogeneous spherical shell     

基于共形几何代数的C-球壳聚类方法及其实现

提出一种新的基于共形几何代数的C-球壳聚类算法。首先C-球壳聚类分析中的模式和球壳原型,在共形几何代数中可以表示为矢量,而共形空间中模式与球壳的内积,等价于模式与球壳的距离,从而能够有效地表示球壳聚类分析中的目标函数,并且推导给出求解该目标函数最小化的新方法;介绍该算法基于VB和MatrixVB的实现,该方法将Matlab的强大计算功能与VB的windows用户界面开发方面优势结合起来,充分发挥了各自优势,缩短了软件的开发周期。实验结果表明,该算法能够有效地对球壳形数据进行聚类,计算方法正确、计算速度快、系统资源消耗少,能满足数据聚类的要求。     

金属球壳在激光辐照下热效应的数值计算

利用有限元法对金属球壳在恒定功率密度激光辐照下的热效应进行了数值计算,分析了在激光辐照过程中,金属球壳内外壁的温升、温差的变化规律,并主要以铜、铝材料为例,比较了铜、铝球壳在激光辐照下的温度变化和外壁温度响应情况.数值计算结果表明,激光辐照期间,在经过热响应的过渡过程后,金属球壳内、外壁的升温率几乎为常量,内外壁温差基本保持恒定值;对于壁厚在0.004～0.02 m范围内的铜、铝球壳,激光辐照停止后,内外壁热平衡过程在几十ms到1s的时间内结束.     

3-D reconstruction of microcalcification clusters using stereo imaging: algorithm and mammographic unit calibration

The three-dimensional (3-D) shape of microcalcification clusters is an important indicator in early breast cancer detection. In fact, there is a relationship between the cluster topology and the type of lesion (malignant or benign). This paper presents a 3-D reconstruction method for such clusters using two 2-D views acquired during standard mammographic examinations. For this purpose, the mammographic unit was modeled using a camera with virtual optics. This model was used to calibrate the acquisition unit and then to reconstruct the clusters in the 3-D space after microcalcification segmentation and matching. The proposed model is hardware independent since it is suitable for digital mammographic units with different geometries and with various physical acquisition principles. Three-dimensional reconstruction results are presented here to prove the validity of the method. Tests were first performed using a phantom with a well-known geometry. The latter contained X-ray opaque glass balls representing microcalcifications. The positions of these balls were reconstructed with a 16.25-microm mean accuracy. This very high inherent algorithm accuracy is more than enough for a precise 3-D cluster representation. Further validation tests were carried out using a second phantom including a spherical cluster. This phantom was built with materials simulating the behavior of both mammary tissue and microcalcifications toward Xrays. The reconstructed shape was effectively spherical. Finally, reconstructions were carried out for real clusters and their results are also presented.     

Insulated loop antenna in a conducting spherical shell

Wait has calculated the impedance seen by a small circular loop of wire placed in an insulating spherical cavity in an infinite homogeneous conducting dielectric. The same technique may be used to compute the impedance of a loop inside a spherical shell. Attention is focused on the change in impedance of an electrically small wire loop when it is placed inside a spherical shell of lossy dielectric material whose radius is small compared to the free-space wavelength. The self-impedance in the absence of the shell can be calculated separately. Simple formulas are developed for the case where the conduction currents in the shell dominate over displacement currents and the shell is thin compared to the skin depth.     

Weighted Graph Based Clustering and Local Mobility Management for Dense Small Cell Network with X2 Interface

Along with the surge in mobile data, dense small cell network has become an effective method to improve system capacity and spectrum efficiency. However, because more small cells are deployed, the interference among dense small cells exacerbates. It also makes frequent handover for mobile users (UEs), which brings a great deal of signaling overhead to the core network. In order to solve the problems of interference and frequent handover, a novel clustering scheme for dense small cell network is proposed in this paper. The scheme is based on the weighted graph. First, we present a dense small cell clustering model based on X2 interface to minimize core network signaling overhead. To improve the usability of the model, we model the system as an undirected weighted graph. Then we propose the maximum benefit merging algorithm to reduce the complexity. This method enables adjacent small cells to cooperate and form virtual cellular cluster according to handover statistics information. Then we select cluster head (CH) according to certain rule in each cluster. Cluster head acts as the mobility anchor, managing the handovers between cluster members. This can reduce core network signaling overhead and the interference among small cells effectively. Compared with the 3GPP handover algorithm, the proposed clustering model in this paper can reduce the signaling overhead more than 70%. The simulation results show that the proposed clustering model can effectively cluster the dense small cell.     

Experimental characterization of the effective propagation constantof dense random media

A new technique for measuring the effective propagation constant of dense random media is presented. This technique involves two major steps: (1) measurement of the mean bistatic scattered field of a cluster of the random medium confined in a spherical boundary and (2) characterization of the complex permittivity for a homogeneous dielectric sphere having identical radius and bistatic scattered field as those of the spherical cluster of the random medium. Using this measurement technique, not only the effective propagation constant of complex dense random media for which an analytical solution does not exist can be characterized, but it can also be used to establish the validity region of the existing models. The sensitivity analyses of the proposed algorithm show that the imaginary part of the effective propagation constant can be measured very accurately. It is also shown that the effective complex permittivity of media with very low dielectric contrast or volume fractions can be characterized accurately. Measurements of the effective propagation constant of different dense random media comprised of homogeneous spherical particles of different packing densities are reported and compared with the existing analytical models     

Rectangular and zonal slots on a sphere with a backing shell: theory and experiment

A rectangular slot antenna on a spherical cavity with a conducting backing shell is investigated rigorously. The Green's function approach is used to formulate an integral equation for the magnetic slot current, which is solved using the moment method. Two previously developed methods to evaluate the admittance integrals are used and compared with each other. Both methods use recurrence formulas and analytical integration to speed the computation. The slot and cavity resonances of the structure are studied. In particular, two simple formulas for the first two natural resonances of the cavity structure are given. The effects of the shell size, slot length, and cavity dielectric constant on the input impedance are discussed. Measurements were carried out to verify the theory, and very good agreement between theory and experiment is obtained. In addition, the zonal slot antenna on a spherical cavity with a conducting backing shell is studied. The effect of the shell radius on the input impedance is investigated and, again, very good agreement between theory and experiment is found.