Content-based Emotional Semantic Recognition And Retrieval Of Male T-shirt Images

At present,with the rapid development of computer technology,an increasing number of images appear in the humanity's vision.Image recognition technology has become an important research goal in the field of recognition.Data mining and knowledge discovery in image emotion semantic level has always attracted the attention of the industry,how to correctly identify the emotional information contained in the image is also a challenging task in technology.Image recognition has risen to the emotional semantic level that it is achieved by computer technology.Emotional semantic analysis of images based on human vision is subjective,so the realization of image emotional semantic analysis needs to draw on psychology,image processing and pattern recognition and other fields of theory and technology.Image recognition based on emotional semantics is an important research to achieve image emotional retrieval.Clothing image as one of the images,to achieve image emotional recognition and retrieval is an important research field of this paper based using emotional semantic analysis based on visual feature extraction.After subjective evaluation of emotional semantic description of male T-shirt images,eight pairs of descriptors were identified.Subsequently,the participants voted on 8 pairs of emotional descriptors of each T-shirt image.After a certain number of participants voted,the average score of emotional descriptors of each image could be calculated and counted.Then,using factor analysis method,an image affective semantic space composed of three independent factor vectors is established.The relationship between emotional factors and low-level features of male T-shirts was analyzed.The images of male T-shirts were processed in HSV color space and the results were obtained.According to this,it is concluded that the emotional semantics of the first factor can be represented by 11-dimensional features(10-dimensional saturation-cold-warm fuzzy histogram plus 1-dimensional color contrast);the emotional semantics of the second factor can be represented by 257-dimensional(256-dimensional gray histogram plus 1-dimensional color contrast);factor 3 represents the emotional semantics of the third factor with a4-dimensional(3-dimensional Tamura texture feature extraction algorithm's element parameters plus 1-dimensional average hue)feature.Finally,a mapping model between three affective factors and low-level features of the image is established by using support vector machine(SVM).The model can automatically calculate the three affective factors according to the low-level features of the male T-shirt image,and then calculate of eight pairs of emotional semantic descriptions,thus realizing the emotional semantic recognition of the male T-shirt image.According to the three emotion factor vectors of male T-shirt images,the content-based emotional semantic retrieval of male T-shirt images is realized by using similarity measure algorithm.Better recognition and retrieval results are obtained in the experimental stage.Combining the knowledge of clothing field and emotional psychology,this paper finds the relationship between the emotional semantic descriptions of T-shirt images and their low-level features.Through the research of this subject,it is shown that the combination of images of professional field and knowledge can help to realize emotional semantic recognition and retrieval content-based.The research on emotional semantic recognition and retrieval of male T-shirt images based on content is carried out.The research results can be applied to the evaluation,design,e-commerce and other aspects of male T-shirt,which has application value.