基于改进迭代收缩阈值算法的微观3D重建方法

迭代收缩阈值算法（ISTA）求解离焦深度恢复动态优化问题时,采用固定迭代步长,导致算法收敛效率不佳,使得重建的微观3D形貌精度不高。为此,提出一种基于加速算子梯度估计和割线线性搜索的方法优化ISTA——FL-ISTA。首先,在每一次迭代中,由当前点和前一个点的线性组合构成加速算子重新进行梯度估计,更新迭代点;其次,为了改变迭代步长固定的限制,引入割线线性搜索,动态确定每次最优迭代步长;最后,将改进的迭代收缩阈值算法用于求解离焦深度恢复动态优化问题,加快算法的收敛速度、提高微观3D形貌重建的精度。在对标准500 nm尺度栅格的深度信息重建实验中,与ISTA、快速ISTA （FISTA）和单调快速ISTA （MFISTA）相比,FL-ISTA收敛速度均有所提升,重建的深度信息值下降了10个百分点,更接近标准500 nm栅格尺度;与ISTA相比,FL-ISTA重建的微观3D形貌均方差（MSE）和平均误差分别下降了18个百分点和40个百分点。实验结果表明,FL-ISTA有效提升了求解离焦深度恢复动态优化问题的收敛速度,提高了微观3D形貌重建的精度。     

一种对光照具有鲁棒性的图像分割方法

研究了灰度图像的OTSU(最大类间方差)自动阈值分割法。OTSU方法作为一种单一阈值的分割方法,当图像受光照和反射光等的影响明显时,将会出现严重的误分割现象。考虑到OTSU方法的最大类间方差化的思想,根据灰度图像的像素点灰度的直方图分布、空间分布,提出了一个新的分割阈值方法。先根据OTSU方法的特点自设计一个函数,对图像进行变换,以便后面的处理,再对其图像以改进的OTSU方法进行分割。通过对化学实验中两种液体的拍摄图片及数字图像处理中标准图片进行试验,理论分析与实验结果表明：该方法能够对受光照及反射光影响大的图像实现正确的分割,将目标图像清晰地从背景中分割出来。     

SVM图像分割方法的研究

支持向量机分类方法已应用于图像分割。本文以彩色图像分割为例,通过对支持向量机图像分割方法和基于灰度直方图图像分割方法进行比较研究。研究揭示:支持向量机图像分割方法是一种在SVM图像上的全局门限分割,并能自动获得默认门限值。使用支持向量机图像方法的这一特点,很容易使其它的分割方法与支持向量机方法相结合,产生新的混和方法。     

基于多级顺序排除算法(MSEA)的改进

块匹配运动估计是视频编码国际标准中广泛采用的关键技术.许多快速块匹配法通过限制搜索点数来减少运算量,但与全搜索算法相比极易出现匹配误差.该文介绍了一种应用新的判别条件的多级顺序排除算法（MSEA）,并在此基础上提出一种新的算法,该算法引入了尺度化的部分失真消除（PDE）技术,用于尺度化累积部分误差和当前最小误差.实验证明,相对于一脉相承的同为穷举搜索算法的全搜索算法（FS）、顺序排除算法（SEA）、多级顺序排除算法（MSEA）等,该算法大幅度提高了搜索效率.与多级顺序排除算法相比,平均每宏块节省了大约75%的运算次数.该算法在保证图像质量的前提下,使视频编码的速度大大提高.     

基于改进粒子群算法的多阈值图像分割

提出了一种改进的粒子群算法,在初始化种群时采用相对基学习原理,以获得较优的初始候选解;在后期迭代过程中引入扩张模型,使粒子不易陷入局部极小值点,并将其用于多阈值图像分割。由最大熵阈值法得到所要优化的目标函数,用改进的粒子群算法对其进行优化,使其能够准确并迅速的得到分割的最佳阈值组合,并用该阈值组合对图像进行分割。将此分割结果与遗传算法的多阈值分割结果相比较可以看出,该算法可更为准确快速的实现图像分割。     

基于多级推理的玉米病虫害专家系统

详细介绍了玉米病虫害专家系统的设计与开发,包括数据库,知识库以及推理机的设计,针对玉米病虫害多样性增加导致传统的专家系统诊断时出现的同一症状对应多个诊断结果的粗糙诊断问题,提出基于多级推理的玉米病虫害专家系统,根据玉米的发病症状,在病虫害的可能性中进行综合判断,以发病时期,发病条件,发病部位为依据采用多级推理机制,以降低推理结果集元素的数量,同时可视知识参与推理,为推理提供感观信息,提高推理的准确性。     

A power conversion system for PMSG-based WECS operating with fully-controlled current-source converters

We propose a new power conversion system for a permanent magnet synchronous generator （PMSG） based grid-connected wind energy conversion system （WECS） operating with fully-controlled back-to-back current-source converters. On the generator side, two independent current-source rectifiers （CSRs） with space-vector pulse width modulation （SVPWM） are employed to regulate and stabilize DC-link currents. Between DC-link and the electrical grid, a direct-type three-phase five-level current-source inverter （CSI） is inserted as a buffer to regulate real and reactive power fed to the grid and thus adjusts the grid side power-factor. We also present a current-based maximum power point tracking （MPPT） scheme, which helps the generator extract the maximum power through closed-loop regulation of the generator speed. By applying the multilevel modulation and control strategies to the grid-side five-level CSI, a multilevel output current waveform with less distortion is produced, and the bulk requirement of the output capacitor filter to eliminate the harmonic current is reduced. All the proposed concepts are verified by simulation models built in a PSIM environment.     

Multilevel projection algorithm for solving obstacle problems

Obstacle problems are nonlinear free boundary problems and the computation of approximate solutions can be difficult and expensive. Little work has been done on effective numerical methods of such problems. This paper addresses some aspects of this issue. Discretizing the problem in a continuous piecewise linear finite element space gives a quadratic programming problem with inequality constraints. A new method, called the multilevel projection (MP) method, is established in this paper. The MP algorithm extends the multigrid method for linear equations to nonlinear obstacle problems. The convergence theorems of this method are also proved. A numerical example presented shows our error estimate is sharp and the MP algorithm is robust.     

A noise attribute thresholding method for document image binarization

A new thresholding method, called the noise attribute thresholding method (NAT), for document image binarization is presented in this paper. This method utilizes the noise attribute features extracted from the images to make the selection of threshold values for image thresholding. These features are based on the properties of noise in the images and are independent of the strength of the signals (objects and background) in the image. A simple noise model is given to explain these noise properties. The NAT method has been applied to the problem of removing text and figures printed on the back of the paper. Conventional global thresholding methods cannot solve this kind of problem satisfactorily. Experimental results show that the NAT method is very effective. Received July 05, 1999 / Revised July 07, 2000