基于广义背景填充的塔式多相水平集的脑肿瘤分割算法

背景填充技术和Chan-Vese模型的塔式多相水平集算法能够在给定的图像中检测出多个目标,而背景填充技术中采用了固定的填充色彩,故该算法得到的边缘是不可调的.为此提出了一种采用可调填充色的广义背景填充技术,使得轮廓可调以便获得所需边缘.脑肿瘤分割实验表明,采用不同的填充色可以得到不同的脑肿瘤轮廓,有利于临床诊断.     

基于区域气球力Snake模型的运动目标快速跟踪方法

提出了一种包含区域信息的Snake模型用于运动目标检测与跟踪。在通常情况下,基于区域信息的跟踪方法对背景光线的微小变化、位置的微小移动较为敏感,而基于边缘信息的跟踪方法则难以对边缘模糊的图像取得满意的跟踪效果。在算法中同时引入这两种信息,边缘信息使得算法快速而鲁棒性好,区域信息可以对边缘模糊的图像取得正确的跟踪效果。使用双差分图像设计了自动初始化的方法来实现视频的自动跟踪。同时,对目标的下一步运动位置增加了一个预测环节来加快主动轮廓模型的收敛速度。该算法的每帧计算时间一般小于0.1S,能应用于实时系统。     

基于快速均值漂移和水平集方法的图像分割技术

通过加强边缘检测函数的约束作用,对L i提出的无需重新初始化的水平集方法进行了改进,用于提高水平集方法对弱边缘与低对比度图像的分割处理性能。同时,利用快速均值漂移法对图像进行预分割,将得到的轮廓作为水平集方法的初始轮廓,从而降低水平集函数的迭代次数。脑部M R图像的分割实验结果表明,该算法能够准确地提取相应的脑组织,同时大大地提高水平集函数的演化速度。     

基于格式塔完形规则的闭合轮廓提取方法

针对传统基于格式塔完形规则的轮廓提取易受到背景上边缘影响,导致轮廓提取的准确率低问题,提出一种新的基于格式塔完形规则的闭合轮廓提取方法。首先,为减少边缘提取过程中提取的背景边缘,给出了融入显著性约束的Canny边缘检测方法,即在Canny边缘提取的框架中,融入显著性约束;接着,针对Canny边缘检测方法检测到的边缘易受噪声影响问题,提出一种融入边缘信息的折线逼近边缘拟合算法;然后,采用格式塔完形规则中的闭合性对拟合边缘进行度量,得到边缘之间的闭合关系;最后,通过分析边缘间的闭合关系与区域的显著性,设计了一个边缘闭合的代价函数,并利用最小权值最优匹配算法求解代价函数,得到最终的闭合轮廓。实验结果表明,该算法提高了轮廓提取的有效性和准确性。     

基于自适应对抗学习的半监督图像语义分割

现有的自适应对抗学习方法采用固定惩罚因子在不同特征层进行监督学习,并采用FCN(Fully Convolutional Networks)作为判别器的基础框架,模型缺少泛化能力,在分割较复杂场景时易造成类感染和类漂移。针对该问题,提出了一种学习率自适应的对抗学习的图像语义分割方法。该方法设计了一种类似SegNet结构的网络判别器,采用最大池化进行非线性上采样,既继承了FCN的优势,可以输入任一大小的图像,又保留了相对精细化的特征相关性信息。由于提出的模型可以通过自适应学习率调整对抗损失与交叉熵损失的权值,从而更新生成器的分割网络,所以提高了语义分割的精度;此外,提出的模型在判别器中采用了SegNet框架代替FCN框架,克服了暴力池化问题,且能够将未标记目标数据集的边缘信息引入网络结构中,从而能有效纠正网络的边缘区域,较好地保持图像的边缘细节,从而使分割结果更为精细。在PASCAL VOC2012标准数据集进行实验,并与现有的性能较好的弱监督分割模型相比,实验结果表明:本文模型能够更精细地分割出较复杂背景的目标,有效地缓解类感染和类漂移,并且有效地保留了边缘细节。     

停滞压力对超声速二维塞式喷嘴设计的影响（英文）

提出了一种新的计算模型用于研究在出口处存在均匀和平行的气流下，燃烧室的停滞压力对超声速二维塞式喷嘴设计的影响。该模型基于真实气体（Real gas, RG）模型，通过使用Berthelot状态方程，考虑了共体积和分子间的相互作用效应。该模型还考虑了分子振动效应，以评估气体在高温下的行为。停滞压力和停滞温度是模型中的重要参数。在边缘，温度和密度是由2个非线性代数方程的分辨率给出的，这些方程由4个复数函数的积分制定。这是由一个新的、强大的和快速的算法完成的，且其他参数由分析关系决定。喷嘴中的气流扩展是Prandtl Meyer型的。喷嘴轮廓通过将喷嘴边缘的膨胀区离散为几个点来确定。马赫数、流动偏差、压力、温度和密度参数是在反演双变量Prandtl Meyer函数后确定的。计算中提出的函数的积分是由30阶的Gauss Legendre求积获得。由于进口和出口部分的气流是单向的，通过计算临界截面比与理论得到的临界截面比的收敛来进行数值控制并验证结果。在这种情况下，喷嘴的轮廓和流动参数，如喷嘴的质量、长度和推力系数，会自动收敛到精确结果。本文提出的新RG模型是对理想气体（Perfect gas,...     

基于纹理分割恒虚警检测器设计与验证

提出一种基于纹理分割恒虚警(TSB-CFAR)检测器,旨在解决杂波边缘情况下恒虚警检测器性能急剧下降的问题.该检测器由杂波分割模块和变参考单元CFAR模块组成,前级杂波分割模块通过多重分形纹理分割算法对杂波图进行区域分割,跟踪杂波边缘在当前参考单元中的位置,后级CFAR检测模块根据杂波边缘位置,选择合适的参考单元参与目标检测.理论分析表明该检测器不仅具有良好的杂波边缘虚警控制能力,而且在均匀背景下能够达到单元平均恒虚警(CA-CFAR)检测器的性能.IPIX海杂波和MSTAR合成孔径雷达地杂波数据检测实验也证实了本文方法的有效性.     

基于图像处理技术的自动报靶系统设计和实现

提出了一种通过图像处理技术设计和实现的自动报靶系统。针对靶图像背景颜色对比度大的特点，采用一种基于灰度期望值的图像二值化算法，通过逐步分离的方法，将靶环从背景中分割出来；针对影响靶数精度的主要因素（图像的非线性畸变），将靶数分解为整数与小数两部分。整数部分根据靶图像的物理意义确定，对靶数精度不会造成任何损失；小数部分根据弹孔最内侧点在两条靶环之间的相对位置确定。试验结果表明，用该系统确定的靶数可以     

逆向工程中一种新的特征识别算法

从某种角度上说,逆向工程是从已有实物的测量数据点中提取其实体特征再进行模型重建的过程。本文提出了一种新的特征识别算法,其首先采用基于面积和法矢准则的数据分割技术,对测量数据点进行数据分割。然后从特征所包含的分割面(简称特征分割面)中提取能够惟一标识该特征的4种特征编码,分别为:表述特征截面形状的截面编码、描述特征凹凸性的凹凸编码、显示特征二维俯视轮廓形状的轮廓编码以及反映特征二维俯视轮廓是否封闭的开闭编码。最后将这4种编码输入到基于人工神经(BP)网络的自动特征识别系统中,识别出特征类型并提取特征参数,从而实现特征重建。着重研究并实现了从特征分割面中提取特征编码的算法,并验证了算法的有效性。     

基于最小总体偏差和区域信息Snake模型的图像分割

提出基于最小总体偏差和区域信息Snake模型的图像分割方法。根据最小总体偏差准则，用包含区域信息的变力替换在气球力Snake模型中的恒定力，并应用于图像分割。实验结果表明，该模型与初始轮廓曲线位置无关，能自动分割模糊边界。对于带椒盐噪声的图像，应用该模型也能取得满意的分割效果。     

复杂背景下快速移动目标的实时跟踪

提出了适用于跟踪在复杂背景下快速移动目标的实时跟踪系统.利用基于马尔可夫场模型的背景减除算法检测像素变化,以跟踪移动目标.分割掩膜的先验概率用马尔可夫场模型来表示,因此目标跟踪任务即被转化成最大后验问题.实验结果表明,本文算法在简单背景或复杂背景下的离线和在线移动目标跟踪方面均有较好的效果.     

Ground-Based Cloud Using Exponential Entropy/Exponential Gray Entropy and UPSO

Objective and accurate classification model or method of cloud image is a prerequisite for accurate weather monitoring and forecast. Thus safety of aircraft taking off and landing and air flight can be guaranteed. Thresholding is a kind of simple and effective method of cloud classification. It can realize automated ground-based cloud detection and cloudage observation. The existing segmentation methods based on fixed threshold and single threshold cannot achieve good segmentation effect. Thus it is difficult to obtain the accurate result of cloud detection and cloudage observation. In view of the above-mentioned problems, multi-thresholding methods of ground-based cloud based on exponential entropy/exponential gray entropy and uniform searching particle swarm optimization （UPSO） are proposed. Exponential entropy and exponential gray entropy make up for the defects of undefined value and zero value in Shannon entropy. In addition, exponential gray entropy reflects the relative uniformity of gray levels within the cloud cluster and background cluster. Cloud regions and background regions of different gray level ranges can be distinguished more precisely using the multi-thresholding strategy. In order to reduce computational complexity of original exhaustive algorithm for multi-threshold selection, the UPSO algorithm is adopted. It can find the optimal thresholds quickly and accurately. As a result, the real-time processing of segmentation of groundbased cloud image can be realized. The experimental results show that, in comparison with the existing groundbased cloud image segmentation methods and multi-thresholding method based on maximum Shannon entropy, the proposed methods can extract the boundary shape, textures and details feature of cloud more clearly. Therefore, the accuracies of cloudage detection and morphology classification for ground-based cloud are both improved.     

An Elasticity Model Considering Grain Boundaries and Tensile Orientations for Directionally Solidified Superalloys

In order to investigate the elastic properties of directionally solidified(DS)superalloys,an elasticity model called boundaries elastic model(GBE model),considering grain boundaries and tensile orientations,is proposed in this paper. Two assumptions are adopted in the GBE model:(1)The displacement of grains,which moves along the perpendicular direction,is restricted by the grain boundaries;(2) Grain boundaries influence region(GBIR) is formed around the grain boundaries. Based on the single crystal(SC)calculation method of elastic properties,the GBE model can well predict macroscopic equivalent elastic modulus(Young's modulus)of DS superalloys under different tensile orientations effectively. To demonstrate the correctness of the GBE model,3 D finite element simulation is adopted and tensile experiments on a Ni3 Al-base DS superalloy(IC10)along five tensile orientations are carried out. Meanwhile,the grain boundaries are observed by light microscopy and transmission electron microscope(TEM). Therefore,the GBE model is proved to be feasible by comparing the simulated results with the experiments.     

海上油气混输中的分离技术

从发展海上油气混输特殊工艺要求出发,介绍中科院力学所“十五”期间,在油气水分离技术方面开展的研究工作。此工艺用于春晓油田,油中除水,再进行油气混输,防止长距离管输时在高压低温条件下油中产生水化物而阻塞管线。在文献总结、专利调研和国内现场考察的基础上,提出研制集重力、离心、膨胀于一体的复合式分离器,以达到高效快速分离的目的。并对国内外尚未使用过的螺旋管分离装置,开展了油水分离的一些理论及实验研究。     

结合图像局部信息的高斯混合型图像分割框架

针对现有的基于判别型或聚类型的图像,用分割方法无法处理被噪声污染的图像的现状,提出一种新的两步式图像分割框架。该框架首先利用图像的局部信息重塑图像的灰度直方图,增强了像素的类间散布性和类内紧凑性,然后将现有的基于判别型或基于聚类型图像分割方法在重塑图像上执行,从而提高了现有图像分割算法的有效性和鲁棒性。文中用典型的聚类型方法高斯混合模型来说明该框架的可行性。由于框架的两个步骤具有独立性．因此可推广到现有的其他基于像素或直方图的方法。在人工和真实图像上的实验结果证明,这种两步图像分割框架可以获得有效且鲁棒的图像分割结果。     

A New Regularized Minimum Error Thresholding Method

To overcome the shortcoming that the traditional minimum error threshold method can obtain satisfactory image segmentation results only when the object and background of the image strictly obey a certain type of probability distribution,one proposes the regularized minimum error threshold method and treats the traditional minimum error threshold method as its special case.Then one constructs the discrete probability distribution by using the separation between segmentation threshold and the average gray-scale values of the object and background of the image so as to compute the information energy of the probability distribution.The impact of the regularized parameter selection on the optimal segmentation threshold of the regularized minimum error threshold method is investigated.To verify the effectiveness of the proposed regularized minimum error threshold method,one selects typical grey-scale images and performs segmentation tests.The segmentation results obtained by the regularized minimum error threshold method are compared with those obtained with the traditional minimum error threshold method.The segmentation results and their analysis show that the regularized minimum error threshold method is feasible and produces more satisfactory segmentation results than the minimum error threshold method.It does not exert much impact on object acquisition in case of the addition of a certain noise to an image.Therefore,the method can meet the requirements for extracting a real object in the noisy environment.     

Segmentation of Thermographic Sequences in Frequency Modulated Thermal Wave Imaging for NDE of GFRP

Image processing techniques have become highly essential to improve the data usefulness of the raw images obtained for infrared non-destructive evaluation(NDE).The identification of defective regions from a thermogram has been a common problem in the field of NDE.Image segmentation is one of the promising approaches to identify defective regions from thermogram.In the present work,segmentation approach based on clustering is employed.In this image patterns are organized into clusters or groups considering the relationship among these.The method involves finding the centroids of the regions and formation of clusters around each centroid.Thresholding is applied to improve accuracy in formation of clusters.Double thresholding method is adapted to retrieve the shape of defect and to improve the possible diagnosis capabilities of system for NDE applications.The proposed method is investigated with frequency modulated thermal wave imaging(FMTWI)over glass fiber reinforced polymers(GFRP).The GFRP sample consists of square shaped Teflon inserts of different dimensions placed at various depths.A comparative study of the proposed novel segmentation scheme and existing segmentation methods is conducted.