开放的社交平台，为用户的自由标注提供了条件；图片交易和共享网站的出现，为用户的图片获取提供了方便。与此同时，不断涌现的各种图片处理工具方便了大众用户组织和管理图片。但是，用户在组织管理甚至是查找图片的过程中仍然存在诸多困难，如图片管理软件的功能各不相同，用户查找不到自己最想要的图片，这就造成了图片资源利用率不够，用户工作效率低下等等一系列问题。 用户标注的结果是标签。社会标签在揭示信息资源内容、提高信息资源利用率方面的有着重要的作用。通过标签挖掘用户的标注行为，既可以揭示用户的行为规律，也能服务于信息平台的优化和改进。因此，本研究在充分调研基于大众标签探讨用户标注行为及其影响因素现状的基础上，综合运用了问卷调查法、实验法、内容分析法等多种研究方法，通过对图像标签的收集、编码和分析，挖掘用户在图像标注过程中的标注特点，总结出用户图像标注行为模式，在此基础上探讨影响用户标注行为的因素，以提升图像资源服务、完善图像检索系统功能。 本研究从昵图网上收集了动物、人物、人文景观、人工制品和自然风景五类共150幅常见图片，利用自主设计的图像标注系统，随机招募了49名被试进行图像标注实验，每位被试为每幅图片至少给出5个客观词标签和5个主观词标签，从标签基本特征、图像特征以及用户标注行为模式三个方面来探讨用户的标注行为，同时从图片类型、用户属性、图片中用户感兴趣的区域（Area of Interest，简称AOI）数量来分析用户图片标注行为的影响因素。 研究发现，用户的标注行为主要通过标注基本行为和标注模式行为来表现。其中，标注基本行为包括标签基本特征和图像元数据特征。第一，标签基本特征上，用户使用主观词和客观词标签的数量差别不大，而在使用主观词和客观词标签的长度上差异显著。第二，图像元数据特征上，从标注频次来看，用户在标注图片的客观特征时，关注图片的语义层最多，其次是语法层，非视觉层最少；用户描述图片的一般特征多于专指特征，局部特征多于全局特征，颜色多于形状和纹理。在标注图片的主观特征时，用户更多地描述图片的风格层特征，其次是物理层特征和情绪层特征，审美层特征关注最少。从描述顺序来看，对于客观特征，先描述语义层再描述语法层，先专指语义再一般语义，先局部特征后全局特征，先颜色后形状；对于主观特征，先描述物理层和审美层，后描述风格层和情绪层，物理层中，先冷暖、饱和度、色调后亮暗、形状、纹理，情绪层中，先负面情绪后正面情绪。第三，根据用户标注的特点，总结出两种图像标注行为模式：描述型/分类型的标注动机模式和渐宽型/渐窄型/平行型/跳跃型的标注顺序模式。结果发现，用户更倾向于使用描述型的标注动机模式和平行型的标注顺序模式，同时使用描述型模式的用户更多的使用平行型模式，使用分类型模式的用户更多的使用跳跃型模式。 同时，本研究从用户属性、图片类型、图片中用户感兴趣的区域（AOI）数量三个方面探讨了影响因素与标注基本行为、标注模式行为的关系，结果发现，第一，对标注基本行为有显著影响的因素是图片类型，具体表现为不同主题和内容的图片与主观特征出现频次显著相关，不同表现方式的图片与客观词标签数量、客观词标签长度以及客观特征出现顺序显著相关；第二，对标注模式行为有显著影响的因素是图片中用户感兴趣的区域（AOI）数量，具体表现为图片中AOI区域的数量与渐宽型、渐窄型、跳跃型标注模式显著相关，且AOI区域与用户添加的标签高度一致；用户属性与标注基本行为、标注模式行为均不显著相关。

The open social platform provides conditions for users to freely annotate, and the emergence of photo trading and sharing website provides users with convenient for users image needs.At the same time, various image management tools are convenient for users to organize and manage pictures. However, users still have many difficulties in the process of organizing and searching for pictures. For example, the functions of the image management tools are different， and users can not find the most desired pictures. This leads to a series of problems, such as the insufficient utilization of picture resources, and the low efficiency of the user's work and so on. The result of the annotation is the tag. Social taggs plays an important role in revealing the contents and improving the utilization of information resources. By Analyzing users' tagging behavior, we can not only reveal the users regular behavior, but also serve the optimization and improvement of information platforms. Therefore, based on the full investigation of user tagging behavior and its influencing factors, this study uses a variety of research methods, such as questionnaire, experiment and content analysis, to dig out users’ tagging behavior pattern in the process of image tagging ,to improve the service of image resource and the function of image retrieval system. In this study, 150 images of five categories of animals, characters, human landscape, artificial products and natural scenery were collected from the HuiTu Website. Then 49 subjects were randomly recruited to conduct image tagging experiment in self-designed image tagging system. Each person was given at least 5 objective tagging and 5 subjective tagging for each picture. This paper discusses user's tagging behavior from 3 aspects include basic features of tags, the image features of tags and users’ tagging behavior pattern. At the same time, This paper analyzes the influencing factors from three aspects include image type, user attribute and the number of users’ Area of Interest(AOI) in the picture. The findings is that users tagging behavior is mainly manifested in three aspects: basic tag characteristics, image metadata characteristics and tagging behavior pattern. First, In the field of basic tag characteristics, the number of subjective words and objective words is hardly different, but the difference between the length of subjective words and objective words is significant. Second, In the field of image metadata characteristics, when it comes to the frequency of objective features in the picture, users pays more attention to the semantic layer of the picture, followed by the syntactic layer and the Non-visual layer. The frequency of the generic features are more than the specific features, the local features are more than the global features, and the color is more than the shape and texture. When it comes to the subjective features of the picture, users describes the style layer more, followed by the physical layer and the emotional layer, and the aesthetic layer are the least concerned. In the description order, for the objective features of the picture, the semantic layer is described first and then the syntactic layer is described, the specific feature followed by the generic feature, the global feature followed by the local feature, the shape followed by the color. For the subjective feature, the physical layer is described first, followed by the aesthetic layer, the style layer and the emotional layer are described later. In physical layer, bright and dark, shape, texture followed by cool and warm, saturation and hue; In emotional layer, positive emotion followed by negative emotion. Third, According to the characteristics of user tagging, two types of image tagging behavior patterns are summarized, that is description/categorization and widening/narrowing /parallel/ salutatory. The findings is that users tend to use description pattern and parallel pattern to tagging images. Besides, this study explores the relationship between the influence factors and the basic tagging behavior and the tagging pattern behavior from three aspects of the user attribute, the image type and the number of AOIs in the picture. The findings is that, First, image types have significant influence on the basic tagging behavior. There is a significant correlation between the image with different theme and content and the frequency of the subjective features. In the mean time, There is a significant correlation between the image with different ways of expression and the number of the objective word tags, the length of the word tags and the order of the objective features. Second, the number of AOIs have significant influence on the basic pattern behavior. The number of AOIs in the picture is closely related to the gradually narrowing tagging pattern, gradually widening tagging pattern and salutatory tagging pattern. Moreover, the AOI region is highly consistent with the tags added by the user. Finally, the user attributes are hardly significant correlated with the basic tagging behavior and the tagging pattern behavior.