基于信息融合的图像分段检测

基于鲁棒控制理论,在视觉信息与机械运动信息的有机融合方面采用了多信息分布检测控制的信息融合方法;在理论上证明了各传感器的决策值经过与或逻辑表决融合后,可以使系统的误警概率减小,检测概率增大,并通过实验验证了上述理论的正确性。

基于链码和多线程的图像分段轮廓跟踪

针对将图像轮廓分段,并对每段轮廓单独进行或对所有段轮廓同时进行轮廓跟踪的需求,提出了基于链码和多线程的图像分段轮廓跟踪方法。对基于链码的图像轮廓跟踪算法进行改进,使之能应用于基于多线程的图像分段轮廓跟踪方法中。以面向对象的观点,将轮廓跟踪操作封装成类。一段轮廓的跟踪对应一个线程和一个轮廓跟踪类的对象。应用表明,该方法很好的解决了图像分段轮廓跟踪的问题,不仅可以灵活的对整个轮廓中所关心的某几段或所有段进行跟踪,而且各段轮廓的跟踪过程能够相互通信,跟踪速度快、效率高。

基于分段图像识别的风电场异常运行数据清洗方法

风电场异常运行数据清洗对风电场功率预测、理论功率计算以及发电性能评估至关重要。针对现有数据清洗方法中基于统计特征的清洗算法无法有效清洗风功率曲线中部的堆积型异常数据,基于图像识别的直接清洗算法会忽略不同风速区间的像素点密度分布差异等问题,文章提出了基于分段图像识别的风电场异常运行数据清洗方法。首先,以切入风速和额定风速为数据集分段点,生成分段二值图像集,构建基于Canny边缘检测的分散型像素点辨识模型;然后,基于边缘内图像集,构建基于数学形态学的堆积型异常像素点辨识模型;最后,以中国西北地区某两个风电场的实际运行数据为算例,与四分位法、基于密度的空间聚类(DBSCAN)法、基于图像的直接清洗算法进行对比分析,验证了所提算法的有效性和适用性。所提方法采用数学形态学分割图像的主要部分和突起部分,从而有效辨识风功率曲线中部的堆积型异常数据;通过设置不同分段图像的清洗参数以改进图像直接清洗算法对局部细节的识别性能,清洗后的正常数据散点分布更为平滑。

UWB SAR图像的非均匀分段PGA算法

在超宽带合成孔径雷达(UWB SAR)成像过程中,由于载机偏航所导致的相位误差具有严重的距离向空变性,因此需要沿距离向分段应用相位梯度自聚焦(PGA)算法以改善PGA算法的聚焦性能。本文分析了载机空间偏航位置与相位误差空变特性之间的关系,提出了图像非均匀分段PGA(NUSPGA)算法,并从理论上对算法进行了推导。文中还提出了参考点对齐法以校正分段聚焦后子图像间的平移失真问题。最后由实测数据验证了本文所提算法的有效性。

基于高级视觉特性的感兴趣区域判定算法

提出了一种快速确定视频场景中感兴趣区域的算法,其基本思想是通过对人类视觉系统HVS的某些高级视觉特性进行分析,对图像分段后获得的区域做出相应的视觉重要性估计,从而确定每个区域在主观意义上的视觉感兴趣值,该值的大小反映了其重要性程度。实验结果表明,利用该算法所获得的感兴趣区域基本符合人眼的主观判断,而且算法的复杂度较低,具有较高的实用价值。

风沙颗粒起动形式的实验研究

应用高速摄影技术,在风洞实验中获得沙粒运动的连续高速摄影图片,通过时序分段图像叠加方法处理,获得单个粒子的运动轨迹.按照沙粒与气流之间、沙粒与沙粒之间相互作用关系,划分沙粒的起动形式.结果表明,在风沙两相流中的沙粒相存在"自发起跳"、"冲击起跳"和"入射-反弹"等3种起动方式,且3种起动形式在沙粒的起动过程中是交替发生的.

Quantitative research of diagenesis:its effect on pore evolution of the Fuyu oil reservoir in the north Qijia region

From observing cores of 18 wells,identifying 175 ordinary thin sections and 61 thin casting sections,energy spectrum analyses of 37 samples,homogenization temperature measurement of fluid inclusions of 11 samples,we determine the types of diagenesis and pores of the Fuyu oil reservoir in the north Qijia region.We classified the pores and measured their plane porosity using CIA 2000,the software of rock image analysis,calculated the effect of different kinds of diagenesis on porosity,studied the controlling actions of diagenesis to pore evolution quantitatively,combined with burial history,thermal history and their diagenetic environments.Our results show that mechanical compaction and carbonate cementation are the major destructive diagenesed,developed during early diagenesis stages.The reduction in porosity by mechanical compaction and carbonate cementation are about 25% and 8%,while the destructive intensity of siliceous cementation and clay mineral cementation is relatively much smaller,i.e.,the reduction of porosity is about 2% and 0.2% Dissolution is constructive diagenesis,the increment of porosity is about 6%.There are four diagenesis evolution stages,during which the porosity reduced from 30%～38% to 2%～20%.Mechanical compaction and early cementation are the main diageneses in the early diagenesis stages,when porosity was reduced to 2%～10%.Dissolution is the main diagenesis of an A I substage of the middle diagenesis stage,when porosity increased 1%～8%.The dissolution of the A Ⅱ substage of the middle diagenesis stage affected by late cementation,raised porosity 1%～5%.The porosity varied slightly during the middle stage B.

A Novel 360°3D Profilometry

A new method of 360° turning 3D shape measurement in which light sectioning and phase shifting techniques are both used is presented in this paper.A sine light field is applied in the projected light stripe,meanwhile phase shifting technique is used to calculate phase of the light slit.Thereafter wrapped phase distribution of the slit is formed and the unwrapping process will be made by means of the height information based on the light sectioning method.Therefore phase measuring results with better precision can be obtained.At last the target 3D shape data can be produced according to geometric relationships between phases and the object heights.The principles of this method are discussed in detail and experimental results are shown in this paper.

Deriving Changjiang coastal zone wind from C-band SAR and its application to salinity simulation

Wind plays an important role in hydrodynamic processes such as the expansion of Changjiang （Yangtze） River Diluted Water （CDW）, and shelf circulation in the Changjiang estuary. Thus, it is essential to include wind in the numerical simulation of these phenomena. Synthetic aperture radar （SAR） with high resolution and wide spatial coverage is valuable for measuring spatially inhomogeneous ocean surface wind fields. We have collected 87 ERS-2 SAR images with wind-induced streaks that cover the Cbangjiang coastal area, to verify and improve the validity of wind direction retrieval using the 2D fast Fourier transform method. We then used these wind directions as inputs to derive SAR wind speeds using the C-band model. To demonstrate the applicability of the algorithms, we validated the SAR-retrieved wind fields using QuikSCAT measurements and the atmospheric Weather Research Forecasting model. In general, we found good agreement between the datasets, indicating the reliability and applicability of SAR- retrieved algorithms under different atmospheric conditions. We investigated the main error sources of this process, and conducted sensitivity analyses to estimate the wind speed errors caused by the effect of speckle, uncertainties in wind direction, and inaccuracies in the normalized radar cross section. Finally, we used the SAR-retrieved wind fields to simulate the salinity distribution off the Changjiang estuary. The findings of this study will be valuable for wind resource assessment and the development of future numerical ocean models based on SAR images.

Clustering based segmentation of text in complex color images

We propose a novel scheme based on clustering analysis in color space to solve text segmentation in complex color images. Text segmentation includes automatic clustering of color space and foreground image generation. Two methods are also proposed for automatic clustering: The first one is to determine the optimal number of clusters and the second one is the fuzzy competitively clustering method based on competitively learning techniques. Essential foreground images obtained from any of the color clusters are combined into foreground images. Further performance analysis reveals the advantages of the proposed methods.

Fusing PLSA model and Markov random fields for automatic image annotation

A novel image auto-annotation method is presented based on probabilistic latent semantic analysis(PLSA) model and multiple Markov random fields(MRF).A PLSA model with asymmetric modalities is first constructed to estimate the joint probability between images and semantic concepts,then a subgraph is extracted served as the corresponding structure of Markov random fields and inference over it is performed by the iterative conditional modes so as to capture the final annotation for the image.The novelty of our method mainly lies in two aspects:exploiting PLSA to estimate the joint probability between images and semantic concepts as well as multiple MRF to further explore the semantic context among keywords for accurate image annotation.To demonstrate the effectiveness of this approach,an experiment on the Corel5 k dataset is conducted and its results are compared favorably with the current state-of-the-art approaches.

Automated Classification of Segmented Cancerous Cells in Multispectral Images

Automatic reading procedures in colon cells biopsies allow a faster and precise reading of microscopic biopsies. These procedures implement automatic image segmentation in order to classify cell types as cancerous or noncancerous. The authors have developed a new approach aiming to detect colon cancer cells derived from the ＂Snake＂ method but using a progressive division of the dimensions of the image to achieve rapid segmentation. The aim of the present paper was to classify different cancerous cell types based on nine morphological parameters and on probabilistic neural network. Three types of cells were used to assess the efficiency of our classifications models, including BH （Benign Hyperplasia）, IN （Intraepithelial Neoplasia） that is a precursor state for cancer, and Ca （Carcinoma） that corresponds to abnormal tissue proliferation （cancer）. Results showed that among the nine parameters used to classify cells, only three morphologic parameters （area, Xor convex and solidity） were found to be effective in distinguishing the three types of cells. In addition, classification of unknown cells was possible using this method.

一种改进拉氏算子的四阶偏微分方程图像去噪方法

建立了一种描述图像平滑度的泛函,并推导出新的四阶偏微分方程图像去噪模型,在有效去噪的同时,较好地保持了图像的特征。由于该方法得到的图像是分段线性图像,避免了二阶偏微分方程处理图像常出现的"阶梯"效应,同时,和同类的四阶偏微分方程去噪模型相比,其处理结果不会出现"斑"点,因此视觉效果更加理想。最后通过实验证明了该方法的有效性。

高斯混合模型、求解算法及视觉应用综述

高斯混合模型(GMMs)是统计学习理论的基本模型,在可视媒体领域应用广泛。近些年来,随着可视媒体信息的增长和分析技术的深入,GMMs在(纹理)图像分割、视频分析、图像配准、聚类等领域有了进一步的发展。从GMMs的基本模型出发,从理论和应用的角度讨论和分析了GMMs的求解算法,包括EM算法、变化形式等,论述了GMMs的模型选择问题:在线学习和模型约简。在视觉应用领域,介绍了GMMs在图像分段、视频分析、图像配准、图像降噪等领域的扩展模型与方法,详细地阐述了一些最新的典型模型的原理与过程,如用于图像分段的空间约束GMMs、图像配准中的关联点漂移算法。最后,讨论了一些潜在的发展方向与存在的困难问题。

视频技术

Y2000-62576 0104197IEEE 第4届图像分析与判读西南部会议录=4thIEEE Southwest symposium on image analysis and inter-pretation[会,英]/IEEE Computer Society TechnicalCommittee on Computational Medicine.—IEEE Comput-er Society,2000.—311P.(ZC)本会议于2000年4月2～4日在美国得克萨斯州的奥斯汀市召开。会议录共收集论文59篇。主要内容包括:视频图像分析,图像分段,立体图像分析,多分辨率、多频谱与多维图像分析,生物医学图像分析与彩色图像分析等。

视频技术

SPIE-Vol.3545 00186881998年 SPIE 会议录,卷3545:国际多频谱图象处理专题讨论会(ISMIP’98)=1998 proceedings of SPIE,Vol.3545:international symposium on multispectral im- age processing(ISMIP’98)[会,英]/SPIE-the Interna-tional Society for Optical Engineering.—1998.—628P.(PC)本会议录收集了于1998年11月21～23日在中国武汉召开的国际多频谱图象处理专题讨论会上发表的134篇论文。内容涵盖计算机视觉,3维模拟与规划,图象生成,雷达、激光与红外成象,图象处理。