视频车辆分类与计数的模型与应用

由于受形态变化、光照变化、视觉碰撞和视觉模糊的影响,基于监控视频的车辆分类和计数一直都是待解决的复杂问题。为了更好地解决这个问题,本文提出新的模型来更好的提取前景。详细来讲,在初次前景提取中,建立模型判断是否存在车辆碰撞,对存在碰撞的车辆通过灰度空间双阀值和YCbCr图像空间处理后,对前景进行更准确的再提取。并在此基础上针对碰撞车辆,定义间隙特征向量将车辆分割问题转换为寻找分割点的优化问题,从而给出高效的车辆分割算法,对发生碰撞的车辆进行准确分割。之后利用神经网络对车辆分类,并设计一种基于已正确对碰撞车辆分割的算法对车辆计数。实验结果表明,本文提出的模型在视频车辆的分类和计数中取得优异的表现,并且数据处理速度能够满足及时性。比起人为计算车流量或建立三维模型等进行分析车辆碰撞情况下的车辆分类与计数,此方法兼顾了准确性与时效性,效率提高,成本减少。     

Fast Surface Mesh Denoising with Regularization and Edge Preservation

We describe a hybrid algorithm that is designed to smooth, but not only smooth, noisy polygonal surface meshes with sharp edges. While denoising, our method simultaneously regularizes triangle meshes on flat regions for further mesh processing and preserves edge sharpness for faithful reconstruction. A clustering technique, which combines K-means and geometric a priori information, is first developed and refined. It is then used to implement vertex classification so that we can subsequently apply different smoothing operators on different vertex groups. This yields a highly efficient robust algorithm that is capable of handling both edge sharpness and mesh sampling irregularity without any significant cost increase. (© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim)     

Towards a more efficient dynamic mesh adaptation methodology for continuum discretization in complex engineering problems

A novel and efficient method of adaptive mesh generation, for dynamically adaptive unstructured grids, is proposed. A locally refined triangulation is constructed on a coarse background mesh, subdividing each triangle in the refinement region R into four congruent sub-triangles iteratively, by connecting edge midpoints, until triangles of a prescribed lengthscale are obtained. The unavoidable propagation outside the refinement region R is restricted to a single triangle in the coarse background mesh. The triangles, in the immediate vicinity of region R, are broken down using the concept of iterated function systems, widely used in fractal modeling, by recursive generation of sub-triangles with a gradation towards the region R triangles. A quantitative assessment of the present algorithm proves its superiority over other comparable models reported in the literature. The time cost of the algorithm is linear, and the method can be easily extended to three dimensions.     

Feature extraction algorithms for constrained global optimization II. Batch process scheduling application

The feature extraction algorithms developed in part I of this series are applied to solve nonconvex mixed integer nonlinear programming problems which arise in the optimal scheduling of multipurpose chemical plants. A general formulation of the multipurpose plant scheduling problem is developed which considers the allocation of plant equipment and secondary, limited resources to recipe operations so as to satisfy given production requirements while minimizing cost. Results obtained with a test example involving 135 binary and 922 continuous variables show that the successive refinement strategy is effective in identifying dominant regions of the solution space. Furthermore it is shown that multiple moment based characterization methods are superior to the interval analysis method reported in the literature. Trials using a second, larger nonlinear test problem involving 356 binary and 2402 continuous variables demonstrate that the focused successive refinement strategy is more efficient than a constant resolution strategy which employs genetic algorithm constructions. Although the conventional genetic algorithm can be significantly improved by introducing a heuristic mutation strategy which increases the likelihood of constant feasibility, the successive refinement strategy remains dominant. These studies demonstrate that the feature extraction strategy employing successive refinements and relatively low order moment based region characterization methods, offers an effective approach to solving an important class of large scale MINLP problems with multiple local optima.     

Parallel Smoothed Aggregation Multilevel Schwarz Preconditioned Newton-Krylov Algorithms for Poisson-Boltzmann Problems

We study a multilevel Schwarz preconditioned Newton-Krylov algorithm to solve the Poisson-Boltzmann equation with applications in multi-particle colloidal simulation. The smoothed aggregation-type coarse mesh space is introduced in collaboration with the one-level Schwarz method as a composite preconditioner for accelerating the convergence of a Krylov subspace method for solving the Jacobian system at each Newton step. The important feature of the proposed solution algorithm is that the geometric mesh information needed for constructing the multilevel preconditioner is the same as the one-level Schwarz method on the fine mesh. Other components, such as the definition of the coarse mesh, all the mesh transfer operators, and the coarse mesh problem, are taken care of by the Trillinos/ML packages of the Sandia National Laboratories in the United States. After algorithmic parameter tuning, we show that the proposed smoothed aggregation multilevel Newton-Krylov-Schwarz (NKS) algorithm numerically outperforms than smoothed aggregation multigrid method and one-level version of the NKS algorithm with satisfactory parallel performances up to a few thousand cores. Besides, we investigate how the electrostatic forces between particles for the separation distance depend on the radius of spherical colloidal particles and valence ratios of cation and anion in a cubic system.     

An exact feature selection algorithm based on rough set theory

Feature reduction based on rough set theory is an effective feature selection method in pattern recognition applications. Finding a minimal subset of the original features is inherent in rough set approach to feature selection. As feature reduction is a Nondeterministic Polynomial‐time‐hard problem, it is necessary to develop fast optimal or near‐optimal feature selection algorithms. This article aims to propose an exact feature selection algorithm in rough set that is efficient in terms of computation time. The proposed algorithm begins the examination of a solution tree by a breadth‐first strategy. The pruned nodes are held in a version of the trie data structure. Based on the monotonic property of dependency degree, all subsets of the pruned nodes cannot be optimal solutions. Thus, by detecting these subsets in trie, it is not necessary to calculate their dependency degree. The search on the tree continues until the optimal solution is found. This algorithm is improved by selecting an initial search level determined by the hill‐climbing method instead of searching the tree from the level below the root. The length of the minimal reduct and the size of data set can influence which starting search level is more efficient. The experimental results using some of the standard UCI data sets, demonstrate that the proposed algorithm is effective and efficient for data sets with more than 30 features. © 2014 Wiley Periodicals, Inc. Complexity 20: 50–62, 2015     

Generalized fast marching method: applications to image segmentation

In this paper, we propose a segmentation method based on the generalized fast marching method (GFMM) developed by Carlini et al. (submitted). The classical fast marching method (FMM) is a very efficient method for front evolution problems with normal velocity (see also Epstein and Gage, The curve shortening flow. In: Chorin, A., Majda, A. (eds.) Wave Motion: Theory, Modelling and Computation, 1997) of constant sign. The GFMM is an extension of the FMM and removes this sign constraint by authorizing time-dependent velocity with no restriction on the sign. In our modelling, the velocity is borrowed from the Chan–Vese model for segmentation (Chan and Vese, IEEE Trans Image Process 10(2):266–277, 2001). The algorithm is presented and analyzed and some numerical experiments are given, showing in particular that the constraints in the initialization stage can be weakened and that the GFMM offers a powerful and computationally efficient algorithm.     

Image segmentation by level set evolution with region consistency constraint

Image segmentation is a key and fundamental problem in image processing, computer graphics, and computer vision. Level set based method for image segmentation is used widely for its topology flexibility and proper mathematical formulation. However, poor performance of existing level set models on noisy images and weak boundary limit its application in image segmentation. In this paper, we present a region consistency constraint term to measure the regional consistency on both sides of the boundary, this term defines the boundary of the image within a range, and hence increases the stability of the level set model. The term can make existing level set models significantly improve the efficiency of the algorithms on segmenting images with noise and weak boundary. Furthermore, this constraint term can make edge-based level set model overcome the defect of sensitivity to the initial contour. The experimental results show that our algorithm is efficient for image segmentation and outperform the existing state-of-art methods regarding images with noise and weak boundary.     

A hybrid trust region algorithm for unconstrained optimization

This paper presents a hybrid trust region algorithm for unconstrained optimization problems. It can be regarded as a combination of ODE-based methods, line search and trust region techniques. A feature of the proposed method is that at each iteration, a system of linear equations is solved only once to obtain a trial step. Further, when the trial step is not accepted, the method performs an inexact line search along it instead of resolving a new linear system. Under reasonable assumptions, the algorithm is proven to be globally and superlinearly convergent. Numerical results are also reported that show the efficiency of this proposed method.     

A variational model and its numerical solution for local,selective and automatic segmentation

Variational region-based segmentation models can serve as effective tools for identifying all features and their boundaries in an image. To adapt such models to identify a local feature defined by geometric constraints, re-initializing iterations towards the feature offers a solution in some simple cases but does not in general lead to a reliable solution. This paper presents a dual level set model that is capable of automatically capturing a local feature of some interested region in three dimensions. An additive operator spitting method is developed for accelerating the solution process. Numerical tests show that the proposed model is robust in locally segmenting complex image structures.     

Combining nonmonotone conic trust region and line search techniques for unconstrained optimization

In this paper, we propose a trust region method for unconstrained optimization that can be regarded as a combination of conic model, nonmonotone and line search techniques. Unlike in traditional trust region methods, the subproblem of our algorithm is the conic minimization subproblem; moreover, our algorithm performs a nonmonotone line search to find the next iteration point when a trial step is not accepted, instead of resolving the subproblem. The global and superlinear convergence results for the algorithm are established under reasonable assumptions. Numerical results show that the new method is efficient for unconstrained optimization problems.     

A new modified panoramic UAV image stitching model based on the GA-SIFT and adaptive threshold method

This paper presents panoramic unmanned aerial vehicle (UAV) image stitching techniques based on an optimal Scale Invariant Feature Transform (SIFT) method. The image stitching representation associates a transformation matrix with each input image. In this study, we formulate stitching as a multi-image matching problem, and use invariant local features to find matches between the images. An improved Geometric Algebra (GA-SIFT) algorithm is proposed to realize fast feature extraction and feature matching work for the scanned images. The proposed GA-SIFT method can locate more feature points with greater accurately than the traditional SIFT method. The adaptive threshold value method proposed solves the limitation problem of high computation load and high cost of stitching time by greater feature points extraction and stitching work. The modified random sample consensus method is proposed to estimate the image transformation parameters and to determine the solution with the best consensus for the data. The experimental results demonstrate that the proposed image stitching method greatly increases the speed of the image alignment process and produces a satisfactory image stitching result. The proposed image stitching model for aerial images has good distinctiveness and robustness, and can save considerable time for large UAV image stitching.     

改进Canny算法在钢印数字特征提取中的应用

边缘检测是实现图像分割、特征提取和图像理解的基础.研究了传统Canny算子的优势与不足.在此基础上,提出了一种快速分块自适应Canny算法.方法首先按字符大小分割图像,然后在每一块上进行自适应边缘检测.自适应边缘检测是在平滑图像的同时得到高斯滤波尺度参数,然后采用Otsu方法的自适应阈值计算Canny算子的高、低门限值.实验结果表明,方法不需人工设定参数就能自动提取不同光照背景下的钢印数字边缘,而且能有效抑制噪声,与传统Canny算子相比,边缘连接程度最佳,噪声敏感程度较低,实时性较强.     

利用三角形及其九点圆的摄像机标定

提出基于三角形及其九点圆的摄像机标定方法.利用了三角形九点圆中其九个点的特殊性,并且利用透视投影变换保二次曲线不变性,得到其像点在像平面共椭圆,从而可以通过九点的映射关系将透视投影变换的非线性问题线性化.图像分割和角点提取的误差会直接影响标定的精度,在此三角形及其九点圆中的点特别是算法中的关键点三角形顶点和垂心都是三条直线的交点,减小图像分割与提取时造成的误差.DLT方法的不精确就源于图像分割和角点提取的误差,方法克服了DLT方法的不足.张的方法无法保证单应性矩阵的正交性,因此为了保证正交性和提高精度需要优化.与传统方法相比操作简单,应用九点圆定理,仿射变换的引入将透视投影非线性问题线性化,避免了参数之间的非线性方程求解,降低了参数求解的复杂性,因此其定标过程快捷,准确.模板的构造,减少了图像分割和交点提取误差,算法实现保证旋转矩阵的正交性.综合上述分析,理论上表明方法的有效性.同时实验表明,标定方法操作简单,不需要计算机视觉的专业知识,快速,精度高,鲁棒性好.     

Region merging with topological control

This paper presents a region merging process controlled by topological features on regions in three-dimensional (3D) images. Betti numbers, a well-known topological invariant, are used as criteria. Classical and incremental algorithms to compute the Betti numbers using information represented by the topological map of an image are provided. The region merging algorithm, which merges any number of connected components of regions together, is explained. A topological control of the merging process is implemented using Betti numbers to control the topology of an evolving 3D image partition. The interest in incremental approaches of the computation of Betti numbers is established by providing a processing time comparison. A visual example showing the result of the algorithm and the impact of topological control is also given.     

Accelerated Non-Overlapping Domain Decomposition Method for Total Variation Minimization

We concern with fast domain decomposition methods for solving the total variation minimization problems in image processing. By decomposing the image domain into non-overlapping subdomains and interfaces, we consider the primal-dual problem on the interfaces such that the subdomain problems become independent problems and can be solved in parallel. Suppose both the interfaces and subdomain problems are uniformly convex, we can apply the acceleration method to achieve an $\mathcal{O}(1 / n^2)$ convergent domain decomposition algorithm. The convergence analysis is provided as well. Numerical results on image denoising, inpainting, deblurring, and segmentation are provided and comparison results with existing methods are discussed, which not only demonstrate the advantages of our method but also support the theoretical convergence rate.     

Unreliable Transfer Lines: Decomposition/Aggregation and Optimization

Complexity has been a long standing obstacle to efficient buffer assignment in transfer lines. For fixed buffer sizes, an approximate transfer line decomposition/aggregation algorithm is developed and its ability to predict line performance is demonstrated via Monte-Carlo simulations. It equates the line with a collection of isolated, unreliable multi-state machines with recursively related statistical parameters. For scalability enhancement, state merging is used to reduce the number of machine states from up to 6 down to 2. An efficient dynamic programming based buffering optimization algorithm which minimizes a combined measure of storage and backlog costs in the transfer line is then presented. Numerical results and comparisons with alternative algorithms are reported.     

Restoration of images based on subspace optimization accelerating augmented Lagrangian approach

We propose a new fast algorithm for solving a TV-based image restoration problem. Our approach is based on merging subspace optimization methods into an augmented Lagrangian method. The proposed algorithm can be seen as a variant of the ALM (Augmented Lagrangian Method), and the convergence properties are analyzed from a DRS (Douglas-Rachford splitting) viewpoint. Experiments on a set of image restoration benchmark problems show that the proposed algorithm is a strong contender for the current state of the art methods.     

快速均值漂移图像分割算法研究

Mean shift算法是一种搜索与样本点分布最接近模式的非参数统计方法.但它是一种迭代统计方法,要保证较高的数值计算精度需要较多的迭代次数,耗费较长的计算时间.为克服这一缺点,提出快速均值漂移图像分割算法.该算法在每次迭代时以前一次的聚类中心集合T动态地更新样本集S,并通过使用直方图缩小样本点的搜索范围进一步加快算法的收敛速度.实验结果表明该方法在保证图像分割质量的同时具有较快的收敛速度.