场景图像分类的相关技术分析-计算机应用技术专业论文.docxVIP

摘要 摘要 摘要 如今，网络多媒体技术正在迅猛地发展，各种图像数据库中的图像数量呈现 爆炸式增长。如何对这些图像进行有效的管理与快速的检索是一个巨大的挑战。 场景图像的分类正是解决这个问题的关键技术。本文针对图像特征提取、场景图 像内容表述及其分类方面展开了研究。 在基于视觉词汇的图像表述上，本文以词袋模型(Bag Of Words, BOW)在图像 处理中的应用为背景，较为深入地研究了空间金字塔匹配模型(Spatial Pyramid Matching, SPM)和潜在狄利克雷分布模型(Latent Dirichlet Allocation, LDA)的相关 理论。在图像分类阶段，本文利用支持向量机(Support Vector Machine, SVM)分类 器对场景图像进行了分类实验，实验结果显示，这三种模型都是基于单一的视觉 特征来提取特征，并且在分类准确率、图像特征维数和图像分类时间上均有着不 同的缺陷。 针对上述三种模型的缺点，本文提出了一种融合颜色特征和 SIFT 特征的概率 主题模型(Multi-Features LDA, MF-LDA)进行场景图像的分类。通过实验我们发现 本文提出的模型在提高分类准确率，降低图像特征维数和图像分类时间方面均取 得了理想的效果，为大规模地处理图像奠定了基础。 关键词：场景图像分类 视觉词汇 特征提取 概率主题模型 Abstract Abstract Abstract Today, with fast development of network multimedia technology, the number of images in a variety of image database showing explosive growth. It is a tremendous challenge to manage effectively and retrial images rapidly. Classifying scene images is the key technology to solve this problem. This paper focuses on image features extraction, to image scene representation with visual words and image classification. Based on the visual word representation of images, this paper is based on the study of Bag of Words and its application in image processing. With deeply study on spatial pyramid matching and on latent semantic topic model, LDA. For classification, support vector machine (SVM) is employed to learn the models of mass classes. The experimental results show that, these three models are based on a single visual feature to extracts feature, and have a different defect in the accuracy of classification, the dimension of the image feature and the image classification time. For disadvantage of the above three models, this paper presents a probabilistic topic models (Multi-Features LDA, MF-LDA) of a fusion of color features and SIFT features to classify scene images. Through experiments, we found that the model proposed in this paper to improve the classification accuracy and reduce the image feature dimension and image classification time achieved the desired resu