基于颜色特征的家居设计图像情感语义分类-计算机科学与技术专业论文.docx

II II Abstract With the development of the modern maturity, the house-designing has become an important part of the social culture. Home is not longer as a residence, people request higher demands of household environment and atmosphere. In order to satisfy people’s personalized pursuit of household design, more and more house-designing images with various styles appear on the Internet and daily life. This has made it is urgency to classify the house-designing image for users search. The classification of house-designing image based on the people’s perceptual of the indoor environment reaction mainly. Therefore, it is necessary to analyze and extract the image emotional semantic in the house-designing image classification and retrieval field. The paper main work about the emotional classification of the house-designing image includes: First, because the color features of the indoor environment affects people’s psychological activities directly, and the extraction of texture and shape features is difficult in the complex household design background environment. So the paper selects the color feature as the image perceptual feature. And the paper come up with the method is extracting the global and local color feature which improve the shortage of global color feature, which is made the local color information lost. Second, on the base of the visual perceptual system, the paper analyze and count the people’s perceptual cognition and understanding of house-designing image’s color features through reading literature, investigation and experiment, then a relationship model is built between the image color features and emotional semantic. Third, the paper selects the radial basis function neural network (RBFNN) as a classifier, and optimizing the RBFNN with the orthogonal least square algorithm, to overcome the defects and shortage of the large network scale and training much slower. At the end, the paper develops a house-designing image classification system based on the