当前位置：搜档网 › 毕设之英文文献翻译成中文

毕设之英文文献翻译成中文

A Novel Automatic Image

Annotation Method Based on

Multi-instance Learning

Abstract

Automatic image annotation (AIA) is the bridge of high-level semantic information and the low-level feature. AIA is an effective method to resolve the problem of “Se mantic Gap”. According to the intrinsic character of AIA, which is many regions contained in the annotated image, AIA Based on the framework of multi-instance learning (MIL) is proposed in this paper. Each keyword is analyzed hierarchically in low-granularity-level under the framework of MIL.

Through the representative instances are mined, the semantic similarity of images can be effectively expressed and the better annotation results are able to be acquired, which testifies the effectiveness of the proposed annotation method.

1.Introduction

With the development of multimedia and network technology, image data has been becoming more common rapidly. Facing a mass of image resource, content based image retrieval (CBIR), a technology to organize, manage and analyze these resource efficiently, is becoming a hot point. However, under the limitation of “semantic gap”, that is, the underlying vision features, such as color, texture, and shape, can not reflect and match the query attention completely, CBIR confronts the unprecedented challenge.

In recent years, newly proposed automatic image annotation (AIA) keeps focus on erecting a bridge between high-level semantic and low-level features, which is an effective approach to solve the above mentioned semantic gap. Since 1999 co-occurrence model proposed by Morris etc., the research of automatic image annotation was initiated[1]. In [2], translation model was developed to annotate image automatically based on an assumption that keywords and vision features were different language to describe the same image. Similar to [2], literature [3] proposed Cross Media Relevance Model (CMRM) where the vision information of each image was denoted as blob set which is to manifest the semantic information of image. However, blob set in CMRM was erected based on discrete region clustering which produced a loss of vision features so that the annotation results were too perfect. In order to compensate for this problem, a Continuous-space Relevance Model (CRM) was proposed in [4]. Furthermore, in [5] Multiple-Bernoulli Relevance Model was proposed to improve CMRM and CRM.

Despite variable sides in the above mentioned methods, the core idea based on automatic image annotation is identical. The core idea of automatic image annotation applies annotated images to erect a certain model to describe the potential relationship or map between as keywords and image features which is used to predict unknown annotation images. Even if previous literatures achieved some results from variable sides respectively, semantic description of each keyword has not been defined explicitly in them. For this end, on the basis of investigating the characters of the automatic image annotation, i.e. images annotated by keywords comprise multiple regions; automatic image annotation is regarded as a problem of multi instance learning. The proposed method analyzes each keyword in

multi-granularity hierarchy to reflect the semantic similarity so that the method not only characterizes semantic implication accurately but also improves the performance of image annotation which verifies the effectiveness of our proposed method.

This article is organized as follows: section 1 introduces automatic image annotation briefly; automatic image annotation based on multi-instance learning framework is discussed in detail in section 2; and experimental process and results are described in section 3; section 4 summaries and discusses the future research briefly.

2.Automatic Image Annotation in the framework of

Multi-instance Learning

In the previous learning framework, a sample is viewed as an instance, i.e. the relationship between samples and instances is one-to-one, while a sample may contain more instances, this is to say, the relationship between samples and instances is one-to-many. Ambiguities between training samples of multi-instance learning differ from ones of supervised learning, unsupervised learning and reinforcement learning completely so that the previous methods hardly solve the proposed problems. Owing to its characteristic features and wide prospect, multi-instance learning is absorbing more and more attentions in machine learning domain and is referred to as a newly learning framework[7]. The core idea multi- instance learning is that the training sample set consists of concept-annotated bags which contain unannotated instances. The purpose of multi-instance learning is to assign a conceptual annotation to bags beyond training set by learning from training bags. In

general, a bag is annotated a Positive if and only if at least one instance is labeled Positive, otherwise the bag is annotated as Negative.

2.1Framework of Image Annotation of Multi-instance Learning

According to the above-mentioned definition of the multi-instance learning, namely, a Positive bag contain at least a positive instance, we can draw a conclusion that positive instances should be distributed much more than negative instances in Positive bags. This conclusion shares common properties with DD algorithm [8] in multi-instance learning domain. If some point can represent the more semantic of a specified keyword than any other point in the feather space, no less than one instance in positive bags should be close to this point while all instances in negative bags will be far away from this point. In the proposed methods, we take into consideration each semantic keyword independently. Even if a part of useful information will be lost neglecting the relationship between keywords, various keywords from each image are used to computing the similarities between images so that the proposed methods can represent the semantic similarity of image effectively in low- granularity. In the following sections, each keyword will be analyzed and applied in local level so that irrelevant information with keywords will be eliminated to improve the precision of representation of the semantic of keywords. Firstly, keywords w ，including Positive and Negative bags, are collected, and the area surrounded by Positive bags are obtained by clustering adaptively. Secondly, this cluster is viewed as Positive set of w which contains most items than other clusters and is farthest from Negative bags. Thirdly, Gaussian Mixture Model (GMM) is used to learn the semantic of w . Finally, the images can be annotated

automatically based on the posterior probability of each keyword of images according to the probability of image in GMM by using Bayesian estimation. Figure 1 illustrates this process.

Fig.1. The framework of automatic image annotation based multi-instance learning

2.2Automatic Image Annotation

In convenience, we firstly put forward some symbols. w is denoted as a semantic keyword, X={X k｜k=1,…,N} as a set of training samples, where N is the number of training samples; S={x1,L,x n} as a set of representative instances after adaptively clustering,where x n is the nth item in a clusters. Therefore, GMM is constructed to describe semantic concept of w , i.e. GMM is used to estimate the distribution of each keyword of feature space to erect the one-to-one map from keywords to vision feature. Note that the superiority of GMM lies in producing a smooth estimation for any density distribution which can reflect the feature distribution of semantic keywords effectively by non-parameter density estimating.

For a specified keyword w , GMM represents its vision feature distribution, p ( x w ) is defined as follows:

),()(1∑∑==i M

i i x N i w x p μπ Where ),(∑i i x N μ represents the Gussian distribution of i th component, μi and ∑i are the corresponding mean and variance reapectively, π

i is weight of the i th component, reflecting its significance, and 11=∑=M i i π, M is the number of

components. Each component represents a cluster in feature space, reflecting a vision feature of w . In each component, the conditional probability density of low-level vision feature vector x can be computed as follows:

Where d is the dimension of feature vector x . The parameters of GMM are estimated by EM method which is maximum likelihood estimation for distribution parameters from incomplete data. EM consists of two steps, expectation step, E-step, and maximum step, M-step, which are executed alternately until convergence after multiple iteration. Assuming that the keyword w can produce N w representative instances, ),(∑=i i i μθ represents mean and co-variance of the i th

Gussian component. Intuitively, different semantic keywords should represent different vision features and the numbers of components are not identical with each other in general so that an adaptive value of M can be obtained based on

Minimum Description Length (MDL)[9].

The proposed method extracts semantic clustering sets from training images which are used to construct GMM in which each component represents some vision feature of a specified keyword. From the perspective of semantic mapping, the proposed model described the one-to-many relationship between keywords and the corresponding vision features. The extracted semantic clustering set can reflect the semantic similarity between instances and keywords. According to the above methods, a GMM is constructed for each keyword respectively to describe the semantic of the keyword. And then, for a specified image to be annotated X={x 1,…,x m },where x m is denoted as the m th separated region, the probability of keyword w is computed according to formula (3).

∏=∝m

i i w p X w p x 1)()( (3) Finally, the image X is annotated according to 5 keywords of greatest posterior probabilities.

3. Experimental Results and Analysis

For comparison with other image annotation algorithms fairly, COREL[2], a widely used image data set, is selected in our experimental process. This image set consists of 5000 images, 4500 images from which are used as training samples, the rest 500 images as test samples. 1 through 5 keywords is extracted to annotate an image, so in all 371 keywords exists in dataset. In our experiments, each image is divided 10 regions using Normalized Cut segment technology [6]. 42,379 regions

are produced in all for a whole image data set, and then, these regions are clustered to 500 groups each of which is called a blob. For each region, 36-demension features, such as color, shape, location etc. are considered like literature [2].

In order to measure the performances of various image annotation methods, we adopt the same evaluation metrics as literature [5], some popular indicators in automatic image annotation and image retrieval. Precision is referred as the ratio of the times of correct annotation in relation to all the times of annotation, while recall is referred as the ratio of the times of correct annotation in relation to all the positive samples. The detailed definitions are as follows:

A B precision =

(4) C

B recall = (5) Where A is the number of images annotated by some keyword; B is the number of images annotated correctly;

C is the number of images annotated by some keyword in the whole data set. As a tradeoff between the above indicators, the geometric mean of them is adopted widely, namely:

(6)

Moreover, we take a statistics of the number of keywords annotated correctly which are used to annotate an image correctly at least. The statistical value reflects the coverage of keywords in our proposed methods, denoted b y “Nu mWords ”.

3.1 Experimental Results

Figure 2 shows that the annotated results of the proposed method, MIL Annotation, keep rather a high consistent with the ground truth. This fact verifies the

effectiveness of our proposed methods.

Fig.2. Illustrations of annotation results of MIL Annotation

3.2 Annotation Results of MIL Annotation

Table 1 and Table 2 show that compare the average performance between our proposed method and some traditional annotation models such as COM[1], TM[2], CMRM[3], CRM[4] and MBRM[5], on COREL image data set. In experiments, 263 keywords are concerned.

Table 1. The performances of various annotation model on COREL

Table 2. The comparison of F-measure between various models

From Table 1 and Table 2, we can know that the annotation performance of the proposed method outperforms other models in two keyword set, and the proposed method has a significant improvement relation to existing algorithms in average precision, average recall F-measure and “Num Words”. Specifically, MIL annotation can obtain a significant improvement over COM, TM, CMRM and CRM; in existing probability-based image annotation models, MBRM can get a best annotation performance which is equivalent to the performance of MIL annotation.

4. Conclusions

Analyzing the properties of automatic image annotation deeply can know it can be viewed as a multi- instance learning problem so that we proposed a method to annotated images automatically based on multi-instance learning. Each keyword is analyzed independently to guarantee more effective semantic similarity in low-granularity. And then, under the frame of multi-instance learning, each keyword is further analyzed in various hierarchies. Irrelevant information with keywords will be eliminated to improve the precision of representation of the semantic of keywords by mapping keywords to corresponding region. Experimental results demonstrated the effectiveness of MR-MIL.

References

[1] Mori Y, Takahashi H, Oka R. Image-to-word transformation based on dividing and vector quantizing images with words. In: Proc. of Intl. Workshop on Multimedia Intelligent Storage and Retrieval Management (MISRM'99), Orlando, Oct. 1999. [2] Duygulu P, Barnard K, Freitas N, Forsyth D. Object recognition as machine translation: learning a lexicon for a fixed image vocabulary. In: Proc. of European Conf. on Computer Vision (ECCV’02), Copenhagen, Denmark, May. 2002: 97-112.

[3] Jeon J, Lavrenko V, Manmatha R. Automatic image annotation and retrieval using cross-media relevance models. In: Proc. of Int. ACM SIGIR Conf. on Research and Development in Information Retrieval (ACM SIGIR’03), Toronto, Canada, Jul. 2003:119-126.

[4] Lavrenko V, Manmatha R, Jeon J. A model for learning the semantics of pictures. In: Proc. Of Advances in Neural Information Processing Systems (N IPS’03), 2003.

[5] Feng S, Manmatha R, Lavrenko V. Multiple bernoulli relevance models for image and video annotation. In: Proc. of IEEE Int. Conf. on Computer Vision and Pattern Recognition (CVPR’04), Washington DC, USA, Jun. 2004: 1002-1009. [6] Shi J, Malik J. Normalized cuts and image Segmentation. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2000,22(8): 888-905.

[7] Maron O. Learning from ambiguity. Department of Electrical Engineering and Computer Science, MIT, PhD dissertation.1998.

[8] Maron O, Lozano P T. A framework for multiple-instance learning. In: Proc. of Advances in Neural Information Processing Systems (NIPS’98), Pittsburgh, USA, Oct. 1998: 570-576.

[9] Li J, Wang J. Automatic linguistic indexing of pictures by a statistical modeling approach. IEEE Trans. On Pattern Analysis and Machine Intelligence, 2003, 25(9): 1075 – 1088

基于多实例的新型自动图像标注方法研究

Shunle Zhua ,Xiaoqiu Tana

数学物理信息学院，浙江海洋大学，舟山，316000，中国

摘要：

图像自动标注是连接高层语义特征和底层特征的桥梁。图像自动标注是解决“语义鸿沟”的有效的方法。根据图像自动标注固有的特征，即在标注的图像中包含有很多区域，本论文提出了以多实例的框架研究为基础的图像自动标注。每个关键词都在多实例研究的框架下以低粒度级进行逐层分析。通过这些有代表性的示例的挖掘，图像的相似语义可以有效地进行传送，并且能够实现更好的标注，这也验证了本文中提出的标注方法的有效性。

1.介绍

随着多媒体和网络技术的发展，图像数据已经迅速普及。面对着众多图像资源，一种有效地组织、管理和分析这些资源的技术——基于内容的图像检索正成为热点。然而，在“语义鸿沟”即底层视觉特征如颜色、纹理、形状的限制下，基于内容的图像检索不能完全反映和匹配查询关注，面对着前所未有的挑战。

近年来，新提出的自动语义标注集中于建立起图像的高层语义和底层特征之间的一座桥梁，这是解决上面提到的语义鸿沟的一种有效的方法。自从1999

年Morris提出了共生模式，图像自动标注技术的研究便开始了。在[2]中，翻译模型被开发来实现图像自动标注，它建立在关键词和视觉特征是描述同一图像的不同的语言的假设之上。和[2]相似，文学[3]提出了跨媒体关联模型，该模型中每幅图像的视觉信息被记为BLOB集以体现图像的语义信息。然而跨媒体关联模型中的BLOB是建立在离散区域集群上的，该群会产生视觉丧失以便使标注结果更加完美。为了弥补这个缺陷，[4]中提出了一种连续空间关联模型。此外，[5]中提出了多重贝努利关联模型来改善跨媒体关联模型和连续空间关联模型。

尽管上面提到的方法中易变的方面，建立在图形自动标注上的核心理念却是相同的。图像自动标注的核心理念是应用已标注的图像建立某种模型来描述关键词和用来预测未标注图像的图像特征之间潜在的关系。尽管以前的文献在不同方面都有所成就，但都没对各个关键词的语义描述准确的下定义。鉴于此，在调查了图像自动标注的特点——即图像被标注了多区域组成的关键字后，图像自动标注被当做一种多实例问题来学习。该方法分析了多粒度层次中的每个关键字来反映语义相似度，以便不仅能准确给出语义含义特征，还能提高证实我们提出的方法有效性的图像标注的性能。

本文布局如下：第一部分简要介绍了图像自动标注；第二部分具体讨论了以多实例学习框架为基础的图像自动标注；第三部分给出了实验性进程和结果；第四部分总结并简要讨论了未来的研究。

2.多实例学习框架下的图像自动标注

在以前的学习框架里，样品被视为一个详情，即样品和详情之间的关系式一对一的，然而一个样品可能包含更多的详情，也就是说，样品盒详情之间是一对多的关系。训练多实例学习样品集之间的歧义区分于对那些监督学习、未监督学习和完全强化学习，以至于以前的方法很难解决提出的问题。由于它的

典型特征和广阔的应用前景，多实例学习被机器学习领域越来越重视，它也被称为一种新型学习框架。多实例学习的核心理念是训练样本集由包含未注释实例的概念注释袋组成。多实例学习的目的是通过对训练集的学习在训练集以外给集分配一个概念标注。一般来说，一个包当且仅当至少一个实例被标正时才被标正包，否则该包被标负包。

2.1图像多实例学习的框架

根据上面给出的多实例学习的定义，即一个正包至少包含一个正的实例，我们可以得出结论在正包中正实例应该分布的比负实例多。这个结论和DD算法在多实例学习领域有共同属性。如果一些某些点而不是视觉特征空间里的别的任何点能代表一个特定的关键词的更多语义。正包中应该有不少于一个实例接近这点，而负包中所有实例应该远离这点。上面提到的方法中，我们独立考虑各个语义关键词。尽管忽视关键词之间的关系会使一部分有用信息丢失，每幅图像的各个关键字被用来计算图像之间的相似度，以便所提出的方法能在低粒度下有效地代表每幅图像的语义相似度。在以下部分，每个关键词会被分析和应用到局部，以便和关键词无关的信息能被剔除来提高语义关键词代表的精确性。首先，包括正包和负包的关键词被收集，被正包包围的区域被聚类自适应获得。其次，这个簇被当做比别的簇包含更多详情并最原理负包的正组。再者，高斯混合模型被用来学习w的语义。最后，通过运用贝叶斯估计根据高斯混合模型中图像的可能性，以图像的每个关键词的后可能性为基础，图像能够被自动标注。图1列出了这个进程。

2.2图像自动标注

为了方便，我们先提出一些标记。w被记为一个语义关键词，X={X k｜k=1,…,N}作为一种样本训练集，N是训练集的个数。自适应聚类后S={x1,L,x n}

作为一种实例代表集，x n 是一簇中第n 个项目。因此，GMM 被构建来描述w 的语义概念，即GMM 被用来评估每个视觉空间关键词的分布，以通过关键词和视觉特征建立一对一的关系。请注意GMM 的优点在于对通过非参数密度估计能有效反映语义关键词的特征分布的任何密度分布产生顺利的估计。对一个特定的关键词，GMM 代表它的视觉特征分布，p(x\w) 被定义如下：N （）代表第i 部分的高斯分布，u 和是各自对应的均值和方差，π是第i 部分的权重，反映它的重要性，而且11M

i i π==∑。M 是构成的个数。每个部分代表视觉空间

的一簇，反映w 的一个视觉特征。在每部分中，底层视觉特征矢量的传统概率密度可计算如下：

其中d 是矢量x 的维数。GMM 的参数通过对不完全数据的分布参数使用EM 方法即最大似然估来估计。EM 由两步组成，期望，E ，最大步，M ，这些被交替执行直到经过多次迭代收敛。假设关键字 w 能产生Nw 代表实例，代表第i 个高斯模块的均值和协方差。直觉上，不同的语义关键字应该代表不同视觉特征，一般来讲组成部分的个数彼此并不一致以便能得到M 的一个以最小描述长度为基础的适应值。

前面提出的方法从用来构建GMM 的训练图像中提取语义聚类，在GMM 中每个部分代表一个特定关键词的一些视觉特征。从语义映射的角度来看，所提出的模型描述了关键字和相应的视觉特征之间一对多的关系。提取出的语义簇集能够反映实例和关键字间的语义相似度。根据上面的方法，一个GMM 为每个关键字各自构造来描述该关键字的语义。进而，对于一个待标注的特殊图像，其中Xm 被记为第m 个分割区域，关键字w 的可能性根据公式（3）来计算。

∏=∝m

i i w p X w p x 1)()( （3）最后，X 图像根据5个关键字的最大后验概率被标注。

3.实验结果与分析

为公平的和别的图像标注算法做比较，实验中选用COREL ［2］这个被广泛使用的图像数据集。该图像集包含5000张图片，其中的4500张图片被用做训练样本，剩下的500张图片用作测试样本。每幅图像标注有1到5个关键字，数据集中共有371个关键字。实验中，每个图像被采用归一化的切段技术分为10个区域。整个图像数据集中产生了42379个区域，然后，这些区域被聚集为500组，每组称为一个blob 。每个区域的36维特征如颜色，形状，位置等像文献［2］中一样被考虑。

为了测量各个图像标注方法的性能，我们采用在图像自动标注和图像恢复中图像常用的，即文献［5］中所用评价指标。精确度被称为正确标注和所有标注的次数的比率，而recall 定义为正确标注次数和所有正样本的比率。详细定义如下：

A B precision =

(4) C

B recall = (5) 其中A 是标注有某些关键字的图像个数；B 是标注正确的图像个数；

C 是整个数据集中被某些关键字标注的图像个数；做为上面指标的一个权衡，它们的几何平均数被广泛采用，即：

(6)

此外，我们采取用做至少正确标注图像的关键字标注正确的个数做为统计。

统计值反映了所提出的方法中关键字的覆盖，被称做“NumWords”。

3.1实验结果

图形２显示了所提出方法的标注结果，MIL标注，和事实保持了高度一致。这个结果证实了所提出方法的有效性。

表https://www.sodocs.net/doc/6b13642155.html,标注方法标注结果插图

3.2MIL标注的标注结果

表1和表2显示了我们提出的方法和一些传统标注模型比如COM［1］，TM［2］，CMRM［3］，CRM［4］和NBRM［5］等在COREL图像数据集之间的平均性能对比。实验中，涉及到了263个关键字。

表1.各种基于COREL的标注方法的性能

表2.各种模型之家F方法的对比

从表1和2我们可以看出我们提出的方法的标注性能超越别的模型2个关键字集，而且和现有的算法的平均精度相比它有一个很重要的改进，平均召回F方法和“Numwords”。尤其是MIL标注比COM，TM，CMRM和CRM有很重要的提高；在现有的基于概率的图像标注模型中，NBRM有和MIL标注同等性能的最好的标注性能。

4.结论

深入分析图像自动标注的性能可知，它能被视为一种多实例学习问题，因此我们提出一种基于多实例学习的自动标注图像的方法。每个关键字被单独分析以确保在低粒度中更多有效的语义相似。进而，在多实例学习的框架下，每个关键字在各个层次被深入分析。和关键字无关的信息会被剔除以提高通过把关键字画到相应区域出现的语义关键字的精确度。实验结果列出了MR—MIT 的有效性。

5.参考资料

[4] Lavrenko V, Manmatha R, Jeon J. A model for learning the semantics of pictures. In: Proc. Of Advances in Neural Information Processing Systems (N IPS’03), 2003.

[7] Maron O. Learning from ambiguity. Department of Electrical Engineering and Computer Science, MIT, PhD dissertation.1998.

[8] Maron O, Lozano P T. A framework for multiple-instance learning. In: Proc. of Advances in Neural Information Processing Systems (NIPS’98), Pittsburgh, USA, Oct. 1998: 570-576.

[9] Li J, Wang J. Automatic linguistic indexing of pictures by a statistical modeling approach. IEEE Trans. On Pattern Analysis and Machine Intelligence, 2003, 25(9): 1075 – 1088

关于力的外文文献翻译、中英文翻译、外文翻译

五、外文资料翻译 Stress and Strain 1.Introduction to Mechanics of Materials Mechanics of materials is a branch of applied mechanics that deals with the behavior of solid bodies subjected to various types of loading. It is a field of study that i s known by a variety of names, including “strength of materials” and “mechanics of deformable bodies”. The solid bodies considered in this book include axially-loaded bars, shafts, beams, and columns, as well as structures that are assemblies of these components. Usually the objective of our analysis will be the determination of the stresses, strains, and deformations produced by the loads; if these quantities can be found for all values of load up to the failure load, then we will have obtained a complete picture of the mechanics behavior of the body. Theoretical analyses and experimental results have equally important roles in the study of mechanics of materials . On many occasion we will make logical derivations to obtain formulas and equations for predicting mechanics behavior, but at the same time we must recognize that these formulas cannot be used in a realistic way unless certain properties of the been made in the laboratory. Also , many problems of importance in engineering cannot be handled efficiently by theoretical means, and experimental measurements become a practical necessity. The historical development of mechanics of materials is a fascinating blend of both theory and experiment, with experiments pointing the way to useful results in some instances and with theory doing so in others①. Such famous men as Leonardo da Vinci(1452-1519) and Galileo Galilei (1564-1642) made experiments to adequate to determine the strength of wires , bars , and beams , although they did not develop any adequate theo ries (by today’s standards ) to explain their test results . By contrast , the famous mathematician Leonhard Euler(1707-1783) developed the mathematical theory any of columns and calculated the critical load of a column in 1744 , long before any experimental evidence existed to show the significance of his results ②. Thus , Euler’s theoretical results remained unused for many years, although today they form the basis of column theory. The importance of combining theoretical derivations with experimentally determined properties of materials will be evident theoretical derivations with experimentally determined properties of materials will be evident as we proceed with

七年级下册英语短文带中文翻译(1-12单元)

Unit 1 (sectionB 3a) Dear Student, My name is Bob . I live in Toronto , Canada , and I want a pen pal in China . I think China is a interesting country . I’m 14 years old and my brithday is in November . I can sperk English and a littie Frenth . I have a brother , Paul , and a sister , Sarah . They have pen pals in the United Kingdom and Australia . I like going to the movies with my friends and playing sports . My favorite subject in school is P.E . It’s fun . But I don’t like math. It’s too difficult! Can you wtite to me soon? Bob 亲爱的同学，我的名字叫鲍勃，我住在加拿大的多伦多。我想在中国找个笔友。我认为中国是个非常有趣的国家。我十四岁，我的生日在九月。我能说英语和一点法语。我有一个哥哥 Paul, 和一个姐姐Sarah。他们在英国和澳大利亚有笔友。我喜欢和我的朋友们看电影和做运动。在学校我最喜欢的科目是体育。它很有趣。但我不喜欢数学。它太难了。你能尽快给我写信吗 Bob 3b PEN PAL WANTED My name is Tom King . I’m 14 years old and I’m from Australia . I sperk English . I have a brother , Sam , and a sister , Lisa . I play soccer on weekengs . It’s my favorite sport . I like music at school . It’s fun! My favorite movte is The Long Weekend . Do you know it ? It’s an action movie . Please write and tell me about yoursrlf. 征集笔友我的名字是汤姆~金。我14岁,我来自澳大利亚。我讲英语。我有一个哥哥, 萨姆,和一个妹妹,莉莎。我在周末踢足球，那是我最喜欢的运动。我喜欢学校的音乐课。它是有趣的!我最喜欢的电影是《这个长周末》。你知道它吗?它是一部动作片。请写信给我,告诉我有关你的一些情况。 Unit 2 (sectionB 3a) Welcome to the Garden District Turn left on First Avenue and enjoy the city’s quiet streets and smail parks . Take a walk through the park on Center Avenue . Across from the park is an old hotel . Next to the hotel is a small house with an interesting garden . This is the benginning of the garden tour. 欢迎来到这个花园区在第一大街左转并享受这个城市的安静街道和小公园。在中心大道上散步穿过这个公园。在公园对面是一个老旅馆。紧靠这个旅馆旁边是一个带着有趣花园的小房子。开始游历这个花园。

毕业论文英文参考文献与译文

Inventory management Inventory Control On the so-called "inventory control", many people will interpret it as a "storage management", which is actually a big distortion. The traditional narrow view, mainly for warehouse inventory control of materials for inventory, data processing, storage, distribution, etc., through the implementation of anti-corrosion, temperature and humidity control means, to make the custody of the physical inventory to maintain optimum purposes. This is just a form of inventory control, or can be defined as the physical inventory control. How, then, from a broad perspective to understand inventory control? Inventory control should be related to the company's financial and operational objectives, in particular operating cash flow by optimizing the entire demand and supply chain management processes (DSCM), a reasonable set of ERP control strategy, and supported by appropriate information processing tools, tools to achieved in ensuring the timely delivery of the premise, as far as possible to reduce inventory levels, reducing inventory and obsolescence, the risk of devaluation. In this sense, the physical inventory control to achieve financial goals is just a means to control the entire inventory or just a necessary part; from the perspective of organizational functions, physical inventory control, warehouse management is mainly the responsibility of The broad inventory control is the demand and supply chain management, and the whole company's responsibility. Why until now many people's understanding of inventory control, limited physical inventory control? The following two reasons can not be ignored: First, our enterprises do not attach importance to inventory control. Especially those who benefit relatively good business, as long as there is money on the few people to consider the problem of inventory turnover. Inventory control is simply interpreted as warehouse management, unless the time to spend money, it may have been to see the inventory problem, and see the results are often very simple procurement to buy more, or did not do warehouse departments . Second, ERP misleading. Invoicing software is simple audacity to call it ERP, companies on their so-called ERP can reduce the number of inventory, inventory control, seems to rely on their small software can get. Even as SAP, BAAN ERP world, the field of

平面设计中英文对照外文翻译文献

(文档含英文原文和中文翻译) 中英文翻译平面设计任何时期平面设计可以参照一些艺术和专业学科侧重于视觉传达和介绍。采用多种方式相结合，创造和符号，图像和语句创建一个代表性的想法和信息。平面设计师可以使用印刷，视觉艺术和排版技术产生的最终结果。平面设计常常提到的进程，其中沟通是创造和产品设计。共同使用的平面设计包括杂志，广告，产品包装和网页设计。例如，可能包括产品包装的标志或其他艺术作品，举办文字和纯粹的设计元素，如形状和颜色统一件。组成的一个最重要的特点，尤其是平面设计在使用前现有材料或不同的元素。平面设计涵盖了人类历史上诸多领域，在此漫长的历史和在相对最近爆炸视觉传达中的第20和21世纪，人们有时是模糊的区别和重叠的广告艺术，平面设计和美术。毕竟，他们有着许多相同的内容，理论，原则，做法和语言，有时同样的客人或客户。广告艺术的最终目标是出售的商品和服务。在平面

设计，“其实质是使以信息，形成以思想，言论和感觉的经验”。在唐朝（ 618-906 ）之间的第4和第7世纪的木块被切断打印纺织品和后重现佛典。阿藏印在868是已知最早的印刷书籍。在19世纪后期欧洲，尤其是在英国，平面设计开始以独立的运动从美术中分离出来。蒙德里安称为父亲的图形设计。他是一个很好的艺术家，但是他在现代广告中利用现代电网系统在广告、印刷和网络布局网格。于1849年，在大不列颠亨利科尔成为的主要力量之一在设计教育界，该国政府通告设计在杂志设计和制造的重要性。他组织了大型的展览作为庆祝现代工业技术和维多利亚式的设计。从1892年至1896年威廉?莫里斯凯尔姆斯科特出版社出版的书籍的一些最重要的平面设计产品和工艺美术运动，并提出了一个非常赚钱的商机就是出版伟大文本论的图书并以高价出售给富人。莫里斯证明了市场的存在使平面设计在他们自己拥有的权利，并帮助开拓者从生产和美术分离设计。这历史相对论是，然而，重要的，因为它为第一次重大的反应对于十九世纪的陈旧的平面设计。莫里斯的工作，以及与其他私营新闻运动，直接影响新艺术风格和间接负责20世纪初非专业性平面设计的事态发展。谁创造了最初的“平面设计”似乎存在争议。这被归因于英国的设计师和大学教授Richard Guyatt，但另一消息来源于20世纪初美国图书设计师William Addison Dwiggins。伦敦地铁的标志设计是爱德华约翰斯顿于1916年设计的一个经典的现代而且使用了系统字体设计。在20世纪20年代，苏联的建构主义应用于“智能生产”在不同领域的生产。个性化的运动艺术在俄罗斯大革命是没有价值的，从而走向以创造物体的功利为目的。他们设计的建筑、剧院集、海报、面料、服装、家具、徽标、菜单等。 Jan Tschichold 在他的1928年书中编纂了新的现代印刷原则，他后来否认他在这本书的法西斯主义哲学主张，但它仍然是非常有影响力。 Tschichold ，包豪斯印刷专家如赫伯特拜耳和拉斯洛莫霍伊一纳吉，和El Lissitzky 是平面设计之父都被我们今天所知。他们首创的生产技术和文体设备，主要用于整个二十世纪。随后的几年看到平面设计在现代风格获得广泛的接受和应用。第二次世界大战结束后，美国经济的建立更需要平面设计，主要是广告和包装等。移居国外的德国包豪斯设计学院于1937年到芝加哥带来了“大规模生产”极简到美国;引发野火的“现代”建筑和设计。值得注意的名称世纪中叶现代设计包括阿德里安Frutiger ，设计师和Frutiger字体大学;保兰德，从20世纪30年代后期，直到他去世于1996年，采取的原则和适用包豪斯他们受欢迎的广告和标志设计，帮助创造一个独特的办法，美国的欧洲简约而成为一个主要的先驱。平面设计称为企业形象;约瑟夫米勒，罗克曼，设计的海报严重尚未获取1950年代和1960年代时代典型。从道路标志到技术图表，从备忘录到参考手册，增强了平面设计的知识转让。可读性增强了文字的视觉效果。设计还可以通过理念或有效的视觉传播帮助销售产品。将它应用到产品和公司识别系统的要素像标志、颜色和文字。连同这些被定义为品牌。品牌已日益成为重要的提供的服务范围，许多平面设计师，企业形象和条件往往是同时交替使用。

英语背诵美文30篇(翻译)

生而为赢（中文翻译） ——新东方英语背诵美文30篇目录： ·第一篇：Y outh 青春 ·第二篇：Three Days to See(Excerpts)假如给我三天光明（节选） ·第三篇：Companionship of Books 以书为伴（节选） ·第四篇：If I Rest, I Rust 如果我休息，我就会生锈 ·第五篇：Ambition 抱负 ·第六篇：What I have Lived for 我为何而生 ·第七篇：When Love Beckons Y ou 爱的召唤 ·第八篇：The Road to Success 成功之道 ·第九篇：On Meeting the Celebrated 论见名人 ·第十篇：The 50-Percent Theory of Life 生活理论半对半 ·第十一篇What is Y our Recovery Rate? 你的恢复速率是多少？ ·第十二篇：Clear Y our Mental Space 清理心灵的空间 ·第十三篇：Be Happy 快乐 ·第十四篇：The Goodness of life 生命的美好 ·第十五篇：Facing the Enemies Within 直面内在的敌人 ·第十六篇：Abundance is a Life Style 富足的生活方式 ·第十七篇：Human Life a Poem 人生如诗 ·第十八篇：Solitude 独处 ·第十九篇：Giving Life Meaning 给生命以意义 ·第二十篇：Relish the Moment 品位现在 ·第二十一篇：The Love of Beauty 爱美 ·第二十二篇：The Happy Door 快乐之门 ·第二十三篇：Born to Win 生而为赢 ·第二十四篇：W ork and Pleasure 工作和娱乐 ·第二十五篇：Mirror, Mirror--What do I see镜子,镜子,告诉我 ·第二十六篇：On Motes and Beams 微尘与栋梁 ·第二十七篇：An October Sunrise 十月的日出 ·第二十八篇：To Be or Not to Be 生存还是毁灭 ·第二十九篇：Gettysburg Address 葛底斯堡演说 ·第三十篇：First Inaugural Address(Excerpts) 就职演讲（节选） 1．青春-------------------------------------------------------------------------- 青春不是人生的一个阶段，而是一种心境；青春不是指粉红的面颊、鲜艳的嘴唇、富有弹性的膝盖，而是指坚定的意志、丰富的想象、充沛的情感；青春，它是清新的生命之泉。青春是一种气质，勇敢胜过怯弱，渴求冒险而不贪图安逸。这样的气息60老者常常有，20青年恰恰无。年岁增添，未必使人垂老；理想不再，终于步入暮年。岁月悠悠，衰微只及肌肤；热忱抛却，颓废必致灵魂。忧烦、惶恐、自卑，使人心灵扭曲，心灰意冷。无论60还是16岁，人人心中都怀着对新奇事物的向往，象孩童般对未来充满憧憬，此情永不消退，在生活的游戏中汲取快乐。在你我的内心深处都有一座无线电台，只要它接收到人间和上帝发出的美好、希望、欢乐、勇气和力量的信号，你就会青春永驻。

10kV小区供配电英文文献及中文翻译

在广州甚至广东的住宅小区电气设计中,一般都会涉及到小区的高低压供配电系统的设计.如10kV高压配电系统图,低压配电系统图等等图纸一大堆.然而在真正实施过程中,供电部门(尤其是供电公司指定的所谓电力设计小公司)根本将这些图纸作为一回事,按其电脑里原有的电子档图纸将数据稍作改动以及断路器按其所好换个厂家名称便美其名曰设计(可笑不?),拿出来的图纸根本无法满足电气设计的设计意图,致使严重存在以下问题:(也不知道是职业道德问题还是根本一窍不通) 1.跟原设计的电气系统货不对板,存在与低压开关柜后出线回路严重冲突,对实际施工造成严重阻碍,经常要求设计单位改动原有电气系统图才能满足它的要求(垄断的没话说). 2.对消防负荷和非消防负荷的供电(主要在高层建筑里)应严格分回路(从母线段)都不清楚,将消防负荷和非消防负荷按一个回路出线(尤其是将电梯和消防电梯,地下室的动力合在一起等等,有的甚至将楼顶消防风机和梯间照明合在一个回路,以一个表计量). 3.系统接地保护接地型式由原设计的TN-S系统竟曲解成"TN-S-C-S"系统(室内的还需要做TN-C,好玩吧?),严格的按照所谓的"三相四线制"再做重复接地来实施,导致后续施工中存在重复浪费资源以及安全隐患等等问题.. ............................(违反建筑电气设计规范等等问题实在不好意思一一例举,给那帮人留点混饭吃的面子算了) 总之吧,在通过图纸审查后的电气设计图纸在这帮人的眼里根本不知何物,经常是完工后的高低压供配电系统已是面目全非了,能有百分之五十的保留已经是谢天谢地了. 所以.我觉得:住宅建筑电气设计,让供电部门走!大不了留点位置,让他供几个必需回路的电,爱怎么折腾让他自个怎么折腾去.. Guangzhou, Guangdong, even in the electrical design of residential quarters, generally involving high-low cell power supply system design. 10kV power distribution systems, such as maps, drawings, etc. low-voltage distribution system map a lot. But in the real implementation of the process, the power sector (especially the so-called power supply design company appointed a small company) did these drawings for one thing, according to computer drawings of the original electronic file data to make a little change, and circuit breakers by their the name of another manufacturer will be sounding good design (ridiculously?), drawing out the design simply can not meet the electrical design intent, resulting in a serious following problems: (do not know or not know nothing about ethical issues) 1. With the original design of the electrical system not meeting board, the existence and low voltage switchgear circuit after qualifying serious conflicts seriously hinder the actual construction, often require changes to the original design unit plans to meet its electrical system requirements (monopoly impress ). 2. On the fire load and fire load of non-supply (mainly in high-rise building in) should be strictly sub-loop (from the bus segment) are not clear, the fire load and fire load of non-qualifying press of a circuit (especially the elevator and fire elevator, basement, etc.

外文文献及翻译

文献翻译原文 Combining JSP and Servlets The technology of JSP and Servlet is the most important technology which use Java technology to exploit request of server, and it is also the standard which exploit business application .Java developers prefer to use it for a variety of reasons, one of which is already familiar with the Java language for the development of this technology are easy to learn Java to the other is "a preparation, run everywhere" to bring the concept of Web applications, To achieve a "one-prepared everywhere realized." And more importantly, if followed some of the principles of good design, it can be said of separating and content to create high-quality, reusable, easy to maintain and modify the application. For example, if the document in HTML embedded Java code too much (script), will lead the developed application is extremely complex, difficult to read, it is not easy reuse, but also for future maintenance and modification will also cause difficulties. In fact, CSDN the JSP / Servlet forum, can often see some questions, the code is very long, can logic is not very clear, a large number of HTML and Java code mixed together. This is the random development of the defects. Early dynamic pages mainly CGI (Common Gateway Interface, public Gateway Interface) technology, you can use different languages of the CGI programs, such as VB, C / C + + or Delphi, and so on. Though the technology of CGI is developed and powerful, because of difficulties in programming, and low efficiency, modify complex shortcomings, it is gradually being replaced by the trend. Of all the new technology, JSP / Servlet with more efficient and easy to program, more powerful, more secure and has a good portability, they have been many people believe that the future is the most dynamic site of the future development of technology. Similar to CGI, Servlet support request / response model. When a customer submit a request to the server, the server presented the request Servlet, Servlet responsible for handling requests and generate a response, and then gave the server, and then from the server sent to

英文文献及中文翻译

毕业设计说明书英文文献及中文翻译学院：专 2011年6月电子与计算机科学技术软件工程

https://www.sodocs.net/doc/6b13642155.html, Overview https://www.sodocs.net/doc/6b13642155.html, is a unified Web development model that includes the services necessary for you to build enterprise-class Web applications with a minimum of https://www.sodocs.net/doc/6b13642155.html, is part of https://www.sodocs.net/doc/6b13642155.html, Framework,and when coding https://www.sodocs.net/doc/6b13642155.html, applications you have access to classes in https://www.sodocs.net/doc/6b13642155.html, Framework.You can code your applications in any language compatible with the common language runtime(CLR), including Microsoft Visual Basic and C#.These languages enable you to develop https://www.sodocs.net/doc/6b13642155.html, applications that benefit from the common language runtime,type safety, inheritance,and so on. If you want to try https://www.sodocs.net/doc/6b13642155.html,,you can install Visual Web Developer Express using the Microsoft Web Platform Installer,which is a free tool that makes it simple to download,install,and service components of the Microsoft Web Platform.These components include Visual Web Developer Express,Internet Information Services (IIS),SQL Server Express,and https://www.sodocs.net/doc/6b13642155.html, Framework.All of these are tools that you use to create https://www.sodocs.net/doc/6b13642155.html, Web applications.You can also use the Microsoft Web Platform Installer to install open-source https://www.sodocs.net/doc/6b13642155.html, and PHP Web applications. Visual Web Developer Visual Web Developer is a full-featured development environment for creating https://www.sodocs.net/doc/6b13642155.html, Web applications.Visual Web Developer provides an ideal environment in which to build Web sites and then publish them to a hosting https://www.sodocs.net/doc/6b13642155.html,ing the development tools in Visual Web Developer,you can develop https://www.sodocs.net/doc/6b13642155.html, Web pages on your own computer.Visual Web Developer includes a local Web server that provides all the features you need to test and debug https://www.sodocs.net/doc/6b13642155.html, Web pages,without requiring Internet Information Services(IIS)to be installed. Visual Web Developer provides an ideal environment in which to build Web sites and then publish them to a hosting https://www.sodocs.net/doc/6b13642155.html,ing the development tools in Visual Web Developer,you can develop https://www.sodocs.net/doc/6b13642155.html, Web pages on your own computer.

英语短文中英文翻译

my friend and I are taking a , we are seeing a boy sit on the chair,he is crying,we go and ask him.“what’s the matter with you” he tell us“I can’t find my dog can you help me”.“yes,I can”.And we help him find his dong .oh it stay under the big tree! 今天我和我的朋友一起去散步。突然我们看见一个男孩坐在椅子上，他哭的很伤心。我们走过去问他：“你怎么了”。他告诉我们：“我的狗不见了，你们能帮我找到它吗”。“是的，我们能帮你找到你的狗”然后我们帮助他找到了他的狗，原来是它呆在一棵大树下。 day an old man siselling a big young man comes to the elephant and begins to look at it old man goes up to him and says inhis ear,“Don't sa y anything about the elephant before I sell it,then i'll give you some money.”“All right,”says the young the old man slles the elephant,he gives the young man some money and says,“Now,can you tell me how you find the bad ears of theelephant?”“I don't find the bad ears,”says the young man.“Then why do you look at the elephant slowly?”asks the old young man answers,“Because I never see an elephant before,and I want to know what it looks like.” 一天，一个老的男人正在卖一头大象。一个年轻的男人走向大象然后开始慢慢看着它（大象），这个老的男人走向他对着他的耳朵说，“不要在我卖出它（大象）之前说关于它（大象）的事，然后我会给你一些钱。”“好的”，这个年轻的男人说。在这个老的男人卖出大象后，他给了年轻的男人一些钱并且说，“现在，你可以告诉我你是怎样知道大象的坏的耳朵了吧？”“我不知道坏的耳朵”，这个年轻的男人说。“然后为什么你慢慢的看着大象？”这个老的男人问。这个年轻的男人回答，“因为我在这之前从来没有见过大象，还有我想知道它（大象）是什么样子的。” 3.An old woman had a cat. The cat was very old; she could not run quickly, and she could not bite, because she was so old. One day the old cat saw a mouse; she jumped and caught the mouse. But she could not bite it; so the mouse got out of her mouth and ran away, because the cat could not bite it.? Then the old woman became very angry because the cat had not killed the mouse. She began to hit the cat. The cat said, "Do not hit your old servant. I have worked for you for many years, and I would work for you still, but I am too old. Do not be unkind to the old, but remember what good work the old did when they were young."? 一位老妇有只猫，这只猫很老，它跑不快了，也咬不了东西，因为它年纪太大了。一天，老猫发现一只老鼠，它跳过去抓这只老鼠，然而，它咬不住这只老鼠。因此，老鼠从它的嘴边溜掉了，因为老猫咬不了它。? 于是，老妇很生气，因为老猫没有把老鼠咬死。她开始打这只猫，猫说：“不要打你的老仆人，我已经为你服务了很多年，而且还愿意为你效劳，但是，我实在太老了，对年纪大的不