预校正

This paper focus on the study of the probability based extraction method which includingword, text line and text block extraction in our Optical Character Recognition System. Inpre-process of OCR, several methods of the two problems:image binarisation are alsodiscussed, thus corresponding methods are finally selected and determined.

本文主要对字符识别系统中基于概率的提取方法进行了研究，其中包括对文本行的提取、块的提取和词的提取，并对字符识别预处理中的两个问题：图象二值化和倾斜字体校正的几个方法的优劣进行了讨论和试验模拟，并最终为我们的字符识别系统确定了相应的方法。

Tridiagonal block matrix elemination and sweep algorithm is successfully applied to solve the large matrix involved; central difference method is supplemented to Houbolt method to form a time integration method with estimation and correction, to solve the time integration of second order implicit non-linear equation; simplification of system equation by decoupling between the physical coordinates x and y further improves the calculation accuracy and speed.

其中：成功地引入块三对角矩阵追赶法求解上述问题构成的大型矩阵；提出中心差分法+Houbolt法的预估计—校正时间积分法，以解决二阶隐式非线性方程的时间积分问题；系统方程在物理坐标x和y上的"解耦"简化，使计算速度和计算精度得以进一步的提高。

Cavity flows both with and without sound excitation are studied by solving the 2 D time dependent Reynolds averaged Navier Stokes equations using explicit predictor corrector difference algorithm of Mac Cormack.

采用显式MacCorma－ck二步预估校正有限差分格式，求解二维雷诺平均非定常Navier－Stokes方程；并用Cebeci－Smith代数紊流模型作适当修正来模拟紊流效应，对有、无前缘声激励两种情况的空腔流场作了数值模拟，计算的振荡频率及振荡幅值都与实验结果基本符合，并且较好地模拟出了声激励对流场的影响。

The homotopy continuation algorithm can solve non-linear equations without initial values,but lots of computation shall be done.Therefore,a new homotopic path-tracking algorithm based on Matlab and its function"\" is proposed using secant-prediction and Newton-correction.

由于同伦延拓算法不需要初始值就能求出非线性方程组的解，但计算量大，为此利用Matlab平台和它里面的函数"\"，采用割线预估和牛顿校正，提出了一种新的路径跟踪算法基于Matlab环境的同伦路径跟踪算法。

And the spectrum was divided into three regions, which were 430~100 nm, 1 000~1 800 nm, 430~1 800 nm, respectively. Four methods of spectrum data preprocessing were used, which were multiplicative scattering correction, standard normal variate, first derivative and second derivative.

应用多元散射校正，标准正态变量变换，一阶微分和二阶微分四种不同预处理方法，分别在430~1 000 nm、1 000~1 800 nm和430~1 800 nm三个光谱范围内建立识别脐橙污染程度的偏最小二乘法数学模型。

First derivative combined with standard normal variate was used to pretreat spectra, and calibration model for ginsenosides had been built by partial least squares regression method.

标准正则变换结合一阶导数对光谱进行预处理后，采用偏最小二乘回归算法建立在线测量总皂苷浓度的近红外光谱校正模型。

The artificial neural network models of original spectrum and two pretreated spectra (firstderivative and standard normal variate, respectively) of osthol and imperatorin were established.

分别利用萃取物的原始光谱和两种预处理光谱建立了网络的数学校正模型，并把所得结果进行了比较，确定用一阶导数光谱所建立的模型为最佳网络模型。

A DPLS discriminant analysis model was developed based on spectra region from 400 nm to 2 498 nm with mathematic pretreatment 2-8-6-1 and scatter correction treatment the standard normal variate. The accurate discriminant rate of calibration and external validation was 95.7% and 95.6% respectively. For the adulterated samples (MBM mass percentage≥5%), the accurate discriminant rate was 100%.

建立的判别分析模型：数学预处理方法为2-8-6-1，散射校正方法为变量标准化处理，光谱范围为全谱（408～ 2 492 nm），定标模型的正确判断率为95.7%，外部验证正确判断率为95.6%，对于掺入量≥5%MBM时，正确判断率为100%。

The numerical techniques developed here, including the high accuracy method of using expansions in orthogonal functions in normal direction and the effective algebraic mapping to deal with the problem of infinite region, can raise the calculation precision and the convergence velocity greatly.

文中所发展的方法，如，在法向上采用高精度的高阶谱方法和对无限区域数值问题的更为有效的代数变换处理等，提高了计算精度和收敛速度；同时，通过预估校正迭代法，使抛物化稳定性方程中重要的正规化条件得到满足，保证了数值计算的稳定。

The partial least square regression is used to build calibration models.

通过正确的光谱预处理方法、光谱区域的选择和偏最小二乘回归确定定量校正模型，建立了近红外光谱快速测定柴油性质的方法。

John Robbins lives with his family in Santa Cruz, California.

约翰罗宾和家人居住在加利福尼亚州的圣克鲁斯县。

The international background is mainly the economic ties among different nations tented to be widen and deepen at the peace and development times.

国际背景主要是和平与发展时代条件下的世界各国经济贸易联系的日益加深。