基于深度学习的图像语义理解和分类研究-计算机科学与技术专业论文.docxVIP

AbstractWith Abstract With the vigorous development of Internet and multimedia technology,as well as the rising popularity of social media，more and more people begin to upload，downlo ad and share images resources in the Interact．Image data gradually become an important medium of information exchange and it’S growing at all astonishing rate．Therefore，how to classify and management the massive image data is not only an urgent problem，but also a challenging and rewarding project． The traditional image classification methods also obtain features by preprocessing the original image data with some different rules defined by people．But the deep learning algorithm automatically learning features through the training process of the computers．Therefore，this paper proposes an image classification algorithm based on deep neural network of linear decoder and softmax regression model．First，we learn features of some small image patches with linear decodeL and by convolving and pooling the large images with the learned features，then we obtain the pooled convolved features； secondly,extract the image color and texture features；thirdly,we use softmax regression model to train the features；finally,with color,texture features and distance metric，we can obtain the final classification results． In the evaluation of the experimental results，we use precision and recall as standards．Experimental results show that this method can not only achieve good performance，but also the classification accuracy is higher than traditional image classification algorithm． Keywords：Image classification；Deep learning；Linear decoder；Softmax regression model Il 万方数据 目 目 录 摘 要 ．．I ABSTRACT ．II 第1章绪论 1 1．1研究背景 1 1．2国内外研究现状 2 1．3本文的主要工作 4 1．4本文的结构安排 4 第2章图像分类研究综述 6 2．1图像分类流程 6 2．2传统神经网络 7 2．2．1基本思想 ．7 2．2．2 BP神经网络 8 2．3深度学习简介 ．．12 2．3．1基本思想 12 2．3．2训练过程概述 13 2．4图像分类中的相关性度量 ．14 2．5本章小结 ．．15 第3章图像视觉特征分析与深度学习 ．16 3．1图像传统特征提取 ．16 3．1．1颜色特征 16 3．1．2纹理特征 18 3．2基于深度学习的图像特征分析 ．2l 3．2．1基于线性解码神经网络的特征分析 一2l 3．2．2基于部分联通网络的卷