尺度不变特征与几何特征融合的人耳识别方法

要提高人耳的识别率,关键是特征的提取与表达.尺度不变特征变换(SIFT)技术是局部点特征提取算法,在尺度空间寻找极值点,提取对图像的尺度和旋转变化具有不变性,对光照变化和图像变形具有较强的适应性的特征向量.尝试用SIFT技术来提取外耳图像的结构特征点以形成稳定的特征描述子,为了克服一幅图像中有多个局部描述子相似的问题,在SIFT特征描述子中融入一个耳廓几何特征.最后采用特征向量的欧氏距离作为两幅图像相似性度量标准进行人耳识别.在耳图像库七进行实验.结果表明,该方法不仅可以有效地提取人耳特征,通过少量特征可获得较高的识别率,而且对耳图像刚体变化具有较强的稳健性.     

基于矢量夹角的三维物体识别

提出了一种基于局部描述符的物体识别算法.算法根据点云位置信息得到其矢量和曲率信息,根据形状索引提取特征点,在每个特征点根据矢量夹角把点云物体分割成不同的曲面片,每个矢量夹角曲面片通过一个二维直方图描述.该图显示了特征点与邻域之间法线矢量夹角对特征点法线矢量与特征点到邻域矢量之间夹角的出现频率.对于给定的一个物体,通过比对预测物体和模型物体的曲面片集描述,可得到潜在的对应模型物体,再通过迭代最近点算法,得到最终的识别结果.     

Three-dimensional matched interface and boundary (MIB) method for treating geometric singularities

This paper reports the three-dimensional (3D) generalization of our previous 2D higher-order matched interface and boundary (MIB) method for solving elliptic equations with discontinuous coefficients and non-smooth interfaces. New MIB algorithms that make use of two sets of interface jump conditions are proposed to remove the critical acute angle constraint of our earlier MIB scheme for treating interfaces with sharp geometric singularities, such as sharp edges, sharp wedges and sharp tips. The resulting 3D MIB schemes are of second-order accuracy for arbitrarily complex interfaces with sharp geometric singularities, of fourth-order accuracy for complex interfaces with moderate geometric singularities, and of sixth-order accuracy for curved smooth interfaces. A systematical procedure is introduced to make the MIB matrix optimally symmetric and banded by appropriately choosing auxiliary grid points. Consequently, the new MIB linear algebraic equations can be solved with fewer number of iterations. The proposed MIB method makes use of Cartesian grids, standard finite difference schemes, lowest order interface jump conditions and fictitious values. The interface jump conditions are enforced at each intersecting point of the interface and mesh lines to overcome the staircase phenomena in finite difference approximation. While a pair of fictitious values are determined along a mesh at a time, an iterative procedure is proposed to determine all the required fictitious values for higher-order schemes by repeatedly using the lowest order jump conditions. A variety of MIB techniques are developed to overcome geometric constraints. The essential strategy of the MIB method is to locally reduce a 2D or a 3D interface problem into 1D-like ones. The proposed MIB method is extensively validated in terms of the order of accuracy, the speed of convergence, the number of iterations and CPU time. Numerical experiments are carried out to complex interfaces, including the molecular surfaces of a protein, a missile interface, and van der Waals surfaces of intersecting spheres.     

Digital system of invariant correlation to position and rotation

A new correlation digital system invariant to position and rotation is presented. This new algorithm requires low computational cost, because it uses uni-dimensional signatures (vectors). The signature of the target so like the signature of the object to be recognized in the problem image is obtained using a binary ring mask constructed based on the real positive values of the Fourier transform of the corresponding image. In this manner, each image will have one unique binary ring mask, avoiding in this form the relevant information leak. Using linear and non-linear correlations, this methodology is applied first in the identification of the alphabet letters in Arial font style and then in the classification of fossil diatoms images. Also, this system is tested using the diatom images with additive Gaussian noise. The non-linear correlation results were excellent, obtaining in this way a simple but efficient method to achieve rotation and translation invariance pattern recognition.     

基于直方图变换的多光谱图像3D SPIHT压缩编码算法

提出了新的多光谱图像压缩方案,直方图变换的三维分层树的集划分(3D SPIHT)压缩编码算法。基于多光谱图像的成像特点,在去相关之前,提出一种可逆的直方图变换方法对多光谱各波段图像灰度值进行调整,来提高各波段间的相关性,然后再对变换后的图像利用K_L和二维小波变换去除谱间冗余和空间冗余。小波编码采用两种编码方案:3D_SPIHT以及对它进行改进后的三维位平面的SPIHT算法,并对两者进行了比较。实验表明,采用直方图变换的这两种方法都获得了良好的效果,比没变换前有更好的图像质量和压缩性能。     

Fluctuation limitations on the voxel minimal size at laser nanopolymerization

Laser (or two-photon) nanopolymerization is an effective way of producing 3D polymeric submicron structures. The mainstream in developing this technology is improvement of the spatial resolution of nano-objects. Fluctuation-induced inhomogeneities are studied as the main physical reason limiting the spatial resolution of polymeric structures obtained by nanopolymerization. Typically, complex polymeric structures have the form of a raster composed of many elements, voxels, about 100 nm across. Monte Carlo simulation of a spherically symmetric polymeric voxel is carried out. It is shown that, when the voxel size becomes less than critical, the position and size of a voxel vary from realization to realization (become irreproducible). This effect is attributed to the disappearance of the voxel??s core??part of a voxel that has macroscopic properties. Irreproducible formation of the single voxels may lead to distortions of the fine features of complex microstructures and, hence, to a deterioration of the spatial resolution. Estimates are made of the minimal size of voxels that can be reproducibly produced in real laser experiments.     

A dynamical approach to symplectic and spectral invariants for billiards

The Lazutkin parameter for curves which are invariant under the billiard ball map is viewed symplectically in a way which makes it analogous to the sum of the values of a generating function over a closed orbit. This leads to relations among lengths of closed geodesics, lengths of invariant curves for the billiard map, rotation numbers, and the Lazutkin parameter. These relations establish the Birkhoff invariant and the expansion for the lengths of invariant curves in terms of the Lazutkin parameter as symplectic and spectral invariants (for the Dirichlet spectrum) and provide invariants which characterize a family of ellipses among smooth curves with positive curvature. Geodesic flow on a bounded planar region gives rise to several geometric objects among which are closed reflected geodesics and invariant curves-closed curves whose tangents are invariant under reflection at the boundary. On a bounded domain, the map that assigns to each geodesic segment its successor after reflection at the boundary is called the billiard ball map and its dual (in the cotangent bundle for the boundary) is called the boundary map.     

Application of invariant grey scale features for analysis of porous minerals

Three-dimensional (3D) texture analysis can be used to differentiate similar materials which have a complex structural nature that is not easily reduced to geometric primitives. A method which extends the concept of invariant grey scale features to non-structured 3D textures is introduced and applied to the study of five processed mineral carbon materials which are characteristically similar but derive from different industrial sources. X-ray microtomography (XRMT) was used to obtain 3D tomographic data with isotropic voxel spacing of 9.8 microm. These data were used to construct invariant features for 3D texture measurement via Monte Carlo based sampling routines and integrals of grey scale relational kernel functions. The procedure produced multi-component texture vectors, which were successfully tested against texture samples as a classification-recognition tool. Identification accuracies ranging from 69% to approximately 84% were achieved for the five material sources examined. This result provides a sound basis for quantitative analysis of these materials which to date have proved very difficult to examine using traditional image analysis tools because of their complex natural structure.     

Flow properties of fast three-dimensional sequences for MR angiography

To reduce the scan time of time of flight or phase contrast angiography sequences, fast three-dimensional k-space trajectories can be employed. The best 3D trajectory depends on tolerable scan time, readout time, geometric flexibility, flow/motion properties and others. A formalism for flow/motion sensitivity comparison based on the velocity k-space behavior is presented. It consists in finding the velocity k-space position as a function of the spatial k-space position. The trajectories are compared graphically by their velocity k-space maps, with simulations and with an objective computed index. The flow/motion properties of various 3D trajectories (cones, spiral-pr hybrid, spherical stack of spirals, 3DFT, 3D echo-planar, and shells) were determined. In terms of flow/motion sensitivity the cones trajectory is the best, however, it is difficult to use it for anisotropic resolutions or fields of view. Tolerating more flow sensitivity, the stack of spirals trajectory offers more geometric flexibility.     

Spatial Position Detection of Three-Dimensional Objects by Effective Three-Dimensional Correlation Based on Fourier Transform Profilometry

Spatial position detection of a three-dimensional (3D) object by a correlator based on Fourier transform profilometry is discussed. The presence of the 3D object and its precise spatial position can be recognized by the correlator using the position and phase of the correlation peak. The spatial position is obtained by the relative displacement from the position where the reference object is originally placed. The characteristic of the correlator with respect to the moving direction of a 3D object is experimentally verified. Effects of the baseline factor on 3D object recognition are also investigated.     

基于OB-HMAD算法和光谱特征的高分辨率遥感影像变化检测

高空间分辨率遥感影像蕴涵丰富的地物细节信息,针对高分辨率多时相遥感影像的变化检测可以更清楚认识到地理单元的变化情况,传统的遥感变化检测算法面对高分辨率遥感影像时,会出现明显的"椒盐现象"。本文借鉴面向对象图像分析的思想,以高分辨率遥感影像对象的光谱特征为分析对象,在多变量变化检测算法(multivariate alternative detection, MAD)的基础上,提出一种半自动阈值选取的OB-HMAD(object based-hybrid MAD)算法,并利用该算法进行变化检测实验对比分析。首先对高分辨率多时相遥感影像进行多尺度分割,形成多通道的影像对象;其次利用MAD变换,形成差异影像对象,并对其进行MNF变换,提高影像对象的信噪比;然后采用直方图曲率分析(histogram curvature analysis, HCA)进行半自动阈值选取,提取变化区域;最后结合实地样本数据对变化检测结果进行混淆矩阵的精度验证。结合2012年和2013年北京地区Worldview-2影像的实验可知,OB-HMAD算法融合多通道的光谱信息,可以有效的实现多时相高分影像的变化检测,基本消除了基于像元变化检测中"椒盐"现象的干扰,并在一定程度上降低建筑物阴影和几何配准误差的影响,总体精度和kappa系数也较优于其他变化检测算法,但存在较大的漏检误差。MNF变换可以有效的提高影像的信噪比,使差异信息更集中,直方图曲率分析的阈值分割算法相对其他阈值算法,自动化程度更高。     

Application of spherical harmonics derived space rotation invariant indices to the analysis of multiple sclerosis lesions' geometry by MRI

In the longitudinal study of multiple sclerosis (MS) lesions, varying position of the patient inside the MRI scanner is one of the major sources of assessment errors. We propose to use analytical indices that are invariant to spatial orientation to describe the lesions, rather than focus on patient repositioning or image realignment. Studies were made on simulated lesions systematically rotated, from in vitro MS lesions scanned on different days, and from in vivo MS lesions from a patient that was scanned five times the same day with short intervals of time between scans. Each of the lesions' 3D surfaces was approximated using spherical harmonics, from which indices that are invariant to space rotation were derived. From these indices, an accurate and highly reproducible volume estimate can be derived, which is superior to the common approach of 2D slice stacking. The results indicate that the suggested approach is useful in reducing part of the errors that affect the analysis of changes of MS lesions during follow-up studies. In conclusion, our proposed method circumvents the need for precise patient repositioning and can be advantageous in MRI longitudinal studies of MS patients.     

An adaptive method for computing invariant manifolds in non-autonomous, three-dimensional dynamical systems

We present a computational method for determining the geometry of a class of three-dimensional invariant manifolds in non-autonomous (aperiodically time-dependent) dynamical systems. The presented approach can be also applied to analyse the geometry of 3D invariant manifolds in three-dimensional, time-dependent fluid flows. The invariance property of such manifolds requires that, at any fixed time, they are given by surfaces in R³. We focus on a class of manifolds whose instantaneous geometry is given by orientable surfaces embedded in R³. The presented technique can be employed, in particular, to compute codimension one (invariant) stable and unstable manifolds of hyperbolic trajectories in 3D non-autonomous dynamical systems which are crucial in the Lagrangian transport analysis. The same approach can also be used to determine evolution of an orientable ‘material surface’ in a fluid flow. These developments represent the first step towards a non-trivial 3D extension of the so-called lobe dynamics — a geometric, invariant-manifold-based framework which has been very successful in the analysis of Lagrangian transport in unsteady, two-dimensional fluid flows. In the developed algorithm, the instantaneous geometry of an invariant manifold is represented by an adaptively evolving triangular mesh with piecewise C² interpolating functions. The method employs an automatic mesh refinement which is coupled with adaptive vertex redistribution. A variant of the advancing front technique is used for remeshing, whenever necessary. Such an approach allows for computationally efficient determination of highly convoluted, evolving geometry of codimension one invariant manifolds in unsteady three-dimensional flows. We show that the developed method is capable of providing detailed information on the evolving Lagrangian flow structure in three dimensions over long periods of time, which is crucial for a meaningful 3D transport analysis.     

The one-dimensional Boltzmann gas: The ergodic hypothesis and the phase portrait of small systems

The concept of ergodicity and its application to microcanonical systems composed of few particles of different mases are clarified. The distribution functions in position and velocity are theoretically derived and numerically verified. Moreover, we deal with a one-dimensional Boltzmann gas where the order relation (connected to the one dimensionality) brings constraints depending on the two classes of boundary conditions enforced (reflecting, periodic). The numerical simulations on a one-dimensional Boltzmann gas act as real experiments and allow us to play on the constraints to which the system is subjected.     

基于快速轮廓转动力矩特征的激光主动成像目标识别

针对激光主动成像的图像特性,提出一种基于快速轮廓转动力矩的目标识别方法。将转动力矩的概念引入目标识别中,提出的快速轮廓转动力矩特征(FCTF)不仅包含了轮廓的尺寸、位置、规则度以及目标的亮暗等信息,同时对于旋转、尺度缩放等变换具有不变性。采用转动力矩的快速计算方法,提高了识别算法的计算效率。识别算法首先使用最大稳定极值区域(MSER)算法检测出目标特征区域,并将其变换为圆形区域,然后结合快速转动力矩特征算法提取出目标区域的局部不变特征,最后输入训练好的支持向量机分类器进行识别。实验结果表明相比于已有的激光主动成像目标识别方法,所提算法对于旋转、仿射变换均具有更高的识别率,同时单帧平均运算时间为9.68 ms,满足激光主动成像目标识别系统实时性的要求。     

An assessment of the sharpness of carotid artery tissue boundaries with acquisition voxel size and field strength

A measure of the sharpness of vessel wall interfaces in carotid artery MRI may be useful for assessing the conspicuity of the wall's features. An edge detection technique was used to measure the signal intensity gradients in 2D time-of-flight (2D-TOF) and double-inversion recovery black-blood (DIR-BB) carotid artery images of normal subjects that were acquired at 1.5 T with 0.55 x 0.55 x 2.0-mm (0.6 mm3) acquisition voxels and zero filled to reduce the in-plane reconstructed voxel size by one half in each dimension as well as with 0.27 x 0.27 x 2.0-mm (0.15 mm3) acquisition voxels and at 3.0 T with 0.27 x 0.27 x 2.0-mm (0.15 mm3) acquisition voxels using surface coils. The gradient intensities of the lumen-to-background interface varied closely with the contrast-to-noise ratio of the 2D-TOF imaging. For the DIR-BB imaging, in which higher spatial frequency artery structures are visible, the gradient intensities at the interfaces were higher than theoretically predicted at both field strengths with smaller acquisition voxels. The use of acquisition voxels smaller than those previously used at 1.5 T can improve the visualization of carotid artery structures at 1.5 and 3.0 T with surface coil reception.