特征向量

A sequenced method in consideration of the sample range was proposed to rationally array classifiers of a support vector machine with binary tree architecture. A sample distribution radius and a sample distribution distance were introduced to estimate sample range of all classes in high-dimension characteristic space. The classes with bigger sample range were classified earlier in the higher nodal point of the binary tree architecture, and were given wider classificatory areas in the characteristic space.

考虑到样本的分布情况对分类器推广能力具有较大影响，提出一种次序二叉树支持向量机多类算法，采用样本分布半径和样本分布距离估算各个类别的样本在高维特征空间中的分布情况，把分布半径较大的类别或者分布距离较大的类别较早地分出来，并且在特征空间中给其划分较大的分类区域。

The paper includes the constitutes of system hardware, image pretreatment, the achievement of parameter detection artithmetic. The main work is as follow: The pater makes use of vanguard genetic algorithms in image segmentation, uses vanguard genetic algorithms to get the image threshold. Using this method we may get different grey vehicles from the background exactly. Used the support vector machine theory to conduct the vehicles automatic sorting simulation research, the simulation result had proven this method could quite be accurate carries on the vehicles type the recognition; Improved based on the background automatic renewal hypothesized examination region algorithm, enhanced timeliness and the accuracy which the vehicles examines; And to has carried on the specify based on the hypothesized examination region imagery processing algorithm realization; In the paper also introduced withdraws the algorithm based on the gradation continuous movement vehicles characteristic to carry on the vehicles the characteristic to withdraw as well as to use the localization method which colored filter same gradation chart processing unifies to carry on the realization method which the vehicles license plate locates.

本文根据智能交通控制与仿真对于交通流量、车辆到达率、车辆速度等交通信息检测的需求，基于视频图像进行了交通信息处理的算法研究与实现，主要工作包括以下几个方面：将先锋遗传算法应用到图像阈值分割中，利用先锋遗传算法寻求全局最优阈值，可以比较准确的将图像中不同灰度的车辆从背景中分离出来；采用支持向量机理论进行了车辆自动分类的仿真研究，仿真结果证明该方法能够比较准确的将车辆的类型进行识别；改进了基于背景自动更新的虚拟检测区域算法，提高了车辆检测的实时性和准确性；并对基于虚拟检测区域的图像处理算法实现进行了详细说明；论文中还介绍了基于灰度连续性的运动车辆特征提取算法进行车辆的特征提取以及采用彩色过滤器同灰度图处理相结合的定位方法进行车辆牌照定位的实现方法。

In this paper,an improvement is made through selecting a group of normal orthogonal vectors in feature subspace,to generate large amount of virtual training samples.

在模式特征子空间中选取一组标准正交向量，使用这组向量可以生成大量的虚拟训练样本，从而实现对协方差矩阵的优化。

In this paper, an improvement is made through selecting a group of normal orthogonal vectors in feature subspace, to generate large amount of virtual training samples.

摘要在模式特征子空间中选取一组标准正交向量，使用这组向量可以生成大量的虚拟训练样本，从而实现对协方差矩阵的优化。

The existence theorem of generalized weak efficient solutions with respect to variable, cone for a linear G〓teaux differentiable mapping is proved with set valued mapping fixed-point theorem and the relation between a vector optimization and a variational inequality problem. The existence of weak efficient solutions for multi-objective convex vector optimization is characterized.

建立映射在线性G〓teaux可微条件下关于可变锥的广义弱有效解的存在性及多目标凸向量优化问题在G〓teaux可微条件下弱有效解的特征，利用集值映射不动点定理及向量优化与变分不等式的关系证明线性G〓teaux可微锥凸映射关于可变锥的广义弱有效解的存在性定理。

The algebraic and geometric invariance of singular value vector is firstly given,and then singular value vector compressed is extracted as the image feature,then RBF neural network is used to training and recognition.

首先说明了彩色图像的奇异值向量具有代数和几何不变性，再将降维的奇异值向量作为图像的特征，然后应用RBF神经网络进行训练和识别。

After the tested N gas components are led into the spectrograph via sampling probe and pipeline, the spectrograph obtains the transmittivity data in the scanned wave number or wavelength range and transmits the data to the computer, the computer extracts the effective characteristic quantities as the input to the parallel supporting vector machine algorithm module, and the supporting vector machine outputs the multiple component gas concentration values.

被测的N个多组分气体经气体采样探头与连接管路导入光谱仪，由光谱仪获得扫描波数或波长范围内各波数或波长对应的透射率关系给出光谱数据经数据传输线进入计算机。进入计算机内的光谱数据由软件的数据预处理模块提取有效的特征量，作为支持向量机并行算法模块的输入，支持向量机并行算法模块的输出为N个多组分气体浓度。

A non-uniformly windowed pyramidical architecture is proposed, the normalizations of the feature vectors of two extraction algorithms are investgated and the Polak-Ribiere learning algorithm is modified to guarantee the weight vector not to converge to non-stable local minima.

文中提出了一种带非均匀窗形式的金字塔状多层神经网络模型I研究了两种特征提取方法的向量规整问题，改进了Polak—Rjbiere学习算法并证明它能够保证连结权向量不收敛到非稳定局部极小点。

The experimental results indicate that the improved algorithm can describe feature points in a larger region, and can reduce mismatch probability of experimental images from 19% to 11% because global context vectors based on global shape information are induced to the SIFT vectors based local Information.

实验结果表明，由于在基于局部信息的SIFT向量中加入基于全局形状信息的全局向量，使得当特征点的尺度较小时，可以借助更大邻域范围内的信息对其进行描述，从而降低了由于局部信息相似而造成的误匹配的概率。

The traditional classification methods only use one single classifier, which may lead to onesidedness, low accuracy, and that the samples nearby the Support Vector Machine hyperplanes are more easily misclassified. To solve these problems, the multifeature fusion method based on SVM and KNearest Neighbor classifiers was presented in this paper.

针对传统分类方法只采用一种分类器而存在的片面性，分类精度不高，以及支持向量机分类超平面附近点易错分的问题，提出了基于支持向量机和k近邻的多特征融合方法。

However, as the name（read－only memory）implies, CD disks cannot be written onorchanged in any way.

然而，正如其名字所指出的那样，CD盘不能写，也不能用任何方式改变其内容。

Galvanizes steel pallet is mainly export which suits standard packing of European Union, the North America. galvanizes steel pallet is suitable to heavy rack. Pallet surface can design plate type, corrugated and the gap form, satisfies the different requirements.

镀锌钢托盘多用于出口，替代木托盘，免薰蒸，符合欧盟、北美各国对出口货物包装材料的法令要求；喷涂钢托盘适用于重载上货架之用，托盘表面根据需要制作成平板状、波纹状及间隔形式，满足不同的使用要求。