Elliptic Object Detection Based on Shape Preserving and Active Contour

本文通过形状约束方程(组)与一般主动轮廓模型结合,将目标形状与主动轮廓模型融合到统一能量泛函模型中,提出了一种形状保持主动轮廓模型即曲线在演化过程中保持为某一类特定形状。模型通过参数化水平集函数的零水平集控制演化曲线形状,不仅达到了分割即目标的目的,而且能够给出特定目标的定量描述。根据形状保持主动轮廓模型,建立了一个用于椭圆状目标检测的统一能量泛函模型,导出了相应的Euler-Lagrange常微分方程并用水平集方法实现了椭圆状目标检测。此模型可以应用于眼底乳头分割,虹膜检测及相机标定。实验结果表明,此模型不仅能够准确的检测出给定图像中的椭圆状目标,而且有很强的抗噪、抗变形及遮挡性能。

改进的基于区域的运动目标分割方法

针对视频监控系统,提出了一种改进的基于区域的运动目标分割方法。与传统方法相比,在运动检测阶段,结合时域差分和背景差分进行运动检测,并通过自适应方法进行背景更新;在差分图像二值化时,采用自适应阈值方法来代替传统的手工确定阈值法;对于区域分割,使用基于加权平方欧式距离的均值聚类算法代替传统的均值聚类算法。实验结果表明该改进方法比传统方法具有更好的实时性、鲁棒性和有效性。

轨迹测量系统中特征点检测技术研究

针对机器人直线轨迹测量系统中的图像处理鲁棒性问题,开展线结构光光条图像的特征点识别技术研究,提出了一种基于种子点拟合和直线生长的直线分割方法,很好地解决了各种条件下各段拟合直线的端点自动获取问题,提高了系统的环境适应性和测量精度.实验表明,该方法具有很好的鲁棒性.

基于变分的主动轮廓图像分割方法

图像分割是图像处理中很重要的一个问题，是计算机视觉的基础。因为它能够简化信息的存储和表示，从而能够对获取的图像内容进行智能解释，所以在很多应用问题中，图像分割是必不可少的过程，如医学图像处理，环境三维重建及自动目标识别等。图像分割的方法有很多种，如边缘检测，阈值，区域融合，分水岭及马尔可夫随机场等。虽然这些方法有其各自特点，但是它们在图象分割过程中不能充分将图像底层信息与高层信息结合，从而无法模拟人类视觉系统智能性。当图像底层信息不足时，这些仅基于数据驱动的分割模型无法达到令人满意的结果。尽管某种具体图像分割方法不可能满足所有图像分割要求，但利用尽可能多的高层与底层信息，将图像分割成有意义和人们所期望的区域始终是研究者所追求的目标。图像分割问题的数学建模和计算中有两个关键因素。第一是建立合适的分割模型将分割边界和分割区域的作用有效结合。第二是利用最有效的方法将分割边界和分割区域的几何特征统一到分割模型中。基于变分原理的主动轮廓图像分割将图像视为连续函数。这就使得研究者可以从连续函数空间角度来研究图像分割问题。这同时也为研究者提供严格的数学工具，如微分几何、泛函分析和微分方程等。为此它能很好的解决上述两个问题。第一，Mumford-Shah(M-S)模型为基于变分的主动轮廓分割模型提供了一完整的数学理论框架，并且Mumford-Shah模型从信息论的角度也能得到合理解释。第二，水平集方法能有效的表示分割边界和分割区域的几何特征。与其它方法相比，变分主动轮廓在理论和实际计算过程中都具有显著的优势。首先它能直接处理和表示各种重要的几何特征，如梯度、切向量、曲率等，并且能有效模拟很多动态过程，如线性和非线性扩散等。再则其可以利用很多已有的丰富数值方法进行分析和计算。本文基于变分原理与偏微分方程方法，利用主动轮廓模型具有结合底层图像信息与高层先验知识的特点，将特定先验知识与主动轮廓分割模型进行有效结合以弥补底层图像信息的不足，从而使主动轮分割廓模型具有更强的智能性。本文主要从两点对变分主动轮廓分割模型展开了研究：1、演化轮廓的形状约束；2、演化轮廓的梯度下降流约束及其滤波实现。其主要工作包括以下四个方面的内容：第一，基于M-S模型和样条曲线的开边界检测。本章通过对演化轮廓引入合理边界条件，利用样条曲线表示待检测的开曲线，将一般开曲线的检测问题变为合理的图像分割问题，从而达到一般开曲线检测目的。此方法称为开扩散蛇模型。一般开曲线的检测具有很多应用领域，如：河流、道路、天际线、焊缝等检测。第二，方差主动轮廓模型。在目标跟踪应用中，跟踪目标的主要运动形式表现为平移。本章将此作为一种先验知识与主动轮廓模型结合，提出了一种方差主动轮廓模型(HV)。此模型的特点是轮廓在演化过程中具有平移优先和快速的良好特性。它比已有的主动轮廓模型更适于自动目标跟踪领域。第三，基于M-S模型和隐式曲面变分方法的一般梯度下降流滤波器。本章为一般梯度下降流求取提供了统一框架及解决方法。首先本章将H0梯度下降流和一般梯度下降流统一到Mumford-Shah模型框架中，从而将一般梯度下降流的求取转换为一个极小化泛函问题，并利用隐式曲面变分方法对此极小化泛函进行求解。另外本章从滤波器设计角度出发，通过对H0梯度下降流滤波可以得到一般梯度下降流。滤波器的实现体现了内嵌于一般梯度下降流的先验属性。根据此思想，本章将对应于HV和H1主动轮廓的內积空间顺序组合，对H0梯度下降流进行顺序滤波，提出了一种既具有全局平移优先性又具有局部光滑速度场的主动轮廓，称为HV1主动轮廓。它将H0，H1和HV主动轮廓统一起来。第四，形状保持主动轮廓模型及其应用。针对某些特定目标的检测，本章提出了形状保持主动轮廓模型。此模型能够达到分割即目标的目的，同时能够给出目标的定量描述。基于此模型，本章实现了具有椭圆、直线和平行四边形轮廓特征目标的检测。椭圆形状约束用于眼底图像分割。直线和平行四边行分别用于自动目标识别中的天际线检测和机场跑道跟踪。

动态有序变化的阈值曲面方法

对于背景灰度不一致的图像的二值化问题,可采用在图像的不同区域选取不同阈值的动态阈值曲面方法解决。本文提出一种有序变化的动态阈值曲面方法,通过选取三个动态变化调节系数,使曲面动态地逼近阈值分割的最佳值,同时又降低曲面的误分割,能较好地提取二值图像。

移动机器人环境建模的结构基自由网格法

本文给出的结构基自由网格法综合了正规栅格法和自由空间法的基本思想,依环境的结构信息确定地解决自由空间分割过程中构造想象边界的任意性问题,由此消除了路径的不确定性;此外,结构基自由网格模型可以在一定程度上消除规划路径“绕大弯”的现象。

景物的三维信息获取及其运动分析

采用模糊熵函数对图象象素分类作出整体最优分类评价,实现了区域分割.利用矩及其函数做为各区域的特征表达,构成以区域为基元的符号特征集并描述图象内容。根据立体图象对间的几何关系,解出各区域(基元)的相对三维坐标。与象索匹配相比较,它可以获得较高精度的三维信息和可描述的景物信息.通过获取不同时空的各区域(基元)三维信息,确定了它们的空间运动状态。联系这些状态,构造出景物中物体间的空间关系和近似模型,实现了对景物的3-D识别和描述。

一种基于知识的自然景物解释

利用我们研制成的基于知识的区域分析系统,我们对自然景物图象做了分析、解释。本文介绍:(1)图象初始分割;(2)图象特征提取;(3)规则集构造。这里,图象初始分割采用的是模糊域方法,它基于Fuzzy C-means 算法,并在此基础上修改了收敛准则,增加了迭代分割功能。图象特征分为主持征及从特征,它们建立在层次化的区域数据结构上。适合于区域分析的规则集已包括三类规则,它们不仅具有较强的知识表示能力,而且易于控制及利用。本文介绍针对包括天空、道路、树木、建筑物等物体的简单景物图象所做的解释实验,并给出了实验结果.

基于DM642的车牌识别系统设计

本文设计与实现了一种基于TMS320DM642的车牌识别系统,详细阐述了该系统的硬件构成、软件流程、检测算法以及针对DSP处理器进行的系统优化。系统通过摄像头获取汽车牌照图像,以TMS320DM642处理器为核心建立硬件平台,完成车牌定位,倾斜角校正,字符分割,字符识别等一系列算法。实验结果表明基于TMS320DM642的车牌识别系统准确、有效,应用前景广泛。

云南岩石圈流变学及多震层孕震环境

The Study on rheology of the lithosphere and the environments of the seismogenic layer is currently the basic project of the international earthquake research. Yunnan is the ideal place for studying this project. Through the multi-disciplinary comprehensive study of petrology, geophysics, seismo-geology, rock mechanics, etc., the depth-strength profiles of the lithosphere have been firstly constructed, and the seismogenic layer and its geophysical and tectonic environments in Yunnan have been systematically expounded in this paper. The related results achieved are of the important significances for further understanding the mechanism of strong earthquake generation, dividing the potential foci and exposing recent geodynamical processes in Yunnan. Through the comprehensive contrast of the metamorphic rocks in early and middle Proterozoic outcropping on the surface, DSS data and experimental data of rock seismic velocity under high temperature and high pressure, the petrological structure of the crust and upper mantle has been studied on Yunnan: the upper, middle and lower crust is composed of the metamorphic rocks of greenschist, amphibolite and granulite facies, respectively or granitoids, diorites and gabbros, respectively, and the upper mantle composed of the peridotites. Through the contrast studies of the heat flow and epicenters of the strong earthquakes, the distribution of the geotemperature and the data of focal depth, the relationship of between seismicity and geothermal structure of the lithosphere in Yunnan has been studied: the strong earthquakes with magnitude M ≥ 6.0 mainly take place at the geothermal gradient zone, and the seismic foci densely distribute between 200～500 ℃ isogeotherms. On the basis of studies of the rock properties and constituents of the crust and upper mantle and geothermal structure of the lithosphere, the structure of the rheological stratification of the lithosphere has been studied, and the corresponding depth-strength profiles have been constructed in Yunnan. The lithosphere in majority region of Yunnan has the structure of the rheological stratification, i.e. the brittle regime in the upper crust or upper part of the upper crust, ductile regime in the middle crust or lower part of the upper crust to middle crust, ductile regime in the lower crust and ductile regime in the subcrustal lithosphere. The rheological stratification has the quite marked lateral variations in the various tectonic units. The distributions of the seismogenic layer have been determined by using the high accurate data of focal depth. Through the contrast of the petrological structure, the structure of seismic velocity, electric structure, geotemperature structure, and rheological structure and the study of the focal mechanism in the seismogenic layer, the geophysical environments of the seismogenic layer in Yunnan have been studied. The seismogenic layer in Yunnan is located at the depths of 3 ～ 20 km; the rocks in the seismogenic layer are composed of the metamorphic rocks of greenschist to amphibolite facies (or granites to diorites); the seismogenic layer and its internal focal regions of strong earthquakes have the structure of medium properties with the relatively high seismic velocity, high density and high resistivity; there exists the intracrustal low seismic velocity and high conductivity layer bellow the seismogenic layer, the geotemperature is generally 100～500 ℃ in the depth range in which the seismogenic layer is located. The horizontal stress field predominates in the seismogenic layer, the seismogenic layer corresponds to the brittle regime of the upper crust or brittle regime of the upper crust to semibrittle regime of the middle crust. The formation of the seismogenic layer, preparedness and occurrence of the strong earthquakes is the result of the comprehensive actions of the source fault, rock constituent, structure of the medium properties, distribution of the geotemperature, rheological structure of the seismogenic layer and its external environments. Through the study of the structure, active nature, slip rate, segmentation of the active faults, and seismogenic faults, the tectonic environments of the seismogenic layer in Yunnan have been studied. The source faults of the seismogenic layer in Yunnan are mainly A-type ones and embody mainly the strike slip faults with high dip angle. the source faults are the right-lateral strike slip ones with NW-NNW trend and left-lateral strike slip ones with NE-NEE trend in Southwestern Yunnan, the right-lateral strike slip ones with NNW trend and left-lateral strike slip ones with NNE trend (partially normal ones) in Northwestern Yunnan, the right-lateral strike slip ones with NWW trend in Central Yunnan and left-lateral strike slip ones with NW-NNW trend in Eastern Yunnan. Taking Lijiang earthquake with Ms = 7.0 for example. The generating environments of the strong earthquake and seismogenic mechanical mechanism have been studied: the source region of the strong earthquake has the media structure with the relatively high seismic velocity and high resistivity, there exists the intracrustal low velocity and high conductivity layer bellow it and the strong earthquakes occur near the transitional zone of the crustal brittle to ductile deformation. These characteristics are the generality of the generating environments of strong earthquakes. However, the specific seismogenic tectonic environments and action of the stress field of the seismic source in the various regions, correspondingly constrains the dislocation and rupture mechanical mechanism of source fault of strong earthquake.

知觉组织在拥挤现象中的作用

Crowding, generally defined as the deleterious influence of nearby contours on visual discrimination, is ubiquitous in spatial vision. Specifically, long-range effects of non-overlapping distracters can alter the appearance of an object, making it unrecognizable. Theories in many domains, including vision computation and high-level attention, have been proposed to account for crowding. However, neither compulsory averaging model nor insufficient spatial esolution of attention provides an adequate explanation for crowding. The present study examined the effects of perceptual organization on crowding. We hypothesize that target-distractor segmentation in crowding is analogous to figure-ground segregation in Gestalt. When distractors can be grouped as a whole or when they are similar to each other but different from the target, the target can be distinguished from distractors. However, grouping target and distractors together by Gestalt principles may interfere with target-distractor separation. Six experiments were carried out to assess our theory. In experiments 1, 2, and 3, we manipulated the similarity between target and distractor as well as the configuration of distractors to investigate the effects of stimuli-driven grouping on target-distractor segmentation. In experiments 4, 5, and 6, we focused on the interaction between bottom-up and top-down processes of grouping, and their influences on target-distractor segmentation. Our results demonstrated that: (a) when distractors were similar to each other but different from target, crowding was eased; (b) when distractors formed a subjective contour or were placed regularly, crowding was also reduced; (c) both bottom-up and top-down processes could influence target-distractor grouping, mediating the effects of crowding. These results support our hypothesis that the figure-ground segregation and target-distractor segmentation in crowding may share similar processes. The present study not only provides a novel explanation for crowding, but also examines the processing bottleneck in object recognition. These findings have significant implications on computer vision and interface design as well as on clinical practice in amblyopia and dyslexia.

认知电位时空模式研究

This report mainly focused on methodology of spatiotemporal patterns (STP) of cognitive potentials or event-related potentials (ERP). The representation of STP of brain wave is an important issue in the research of neural assemblies. This paper described methods of parametric 3D head or brain modeling and its corresponding interpolation for functional imaging based on brain waves. The 3D interpolation method is an extension of cortical imaging technique. It can be used with transformed domain features of brain wave on realistic head or brain models. The simulating results suggests that it is a better method in comparison with the global nearest neighbor technique. A stable and definite STP of brainwave referred as microstate may become basic element for comprehending sophisticated cognitive processes. Fuzzy c-mean algorithm was applied to segmentation STPs of ERP into microstates and corresponding membership functions. The optimal microstate number was estimated with both the trends of objective function against increasing clustering number and the decorrelation technique base don microstate shape similarity. Comparable spatial patterns may occur at different moments in time with fuzzy indices and thus the serial processing limit generated from behavioral methods has been break through. High-resolution frequency domain analysis was carried out with multivariate autoregressive model. Bases on a 3D interpolation mentioned above, visualization of dynamical coordination of cerebral network was realized with magnitude-squared partial coherence. Those technique illustrated with multichannel ERP of 9 subjects when they undertook Strop task. Stroop effects involves several regions during post-perception stage with technique of statistical parameter mapping based F-test [SPM(F)]. As SPM(F) suggested task effects occurred within 100 ms after stimuli presentation involved several sensory regions, it may reflect the top-down processing effect.